Search

• Cybernetics systems have been developed to monitor animal production. • Machine vision methods are key for precision poultry welfare detection. • Robotics is an emerging method for poultry farm management. • Sustainable poultry farming requires continuous innovation in cybernetics. As global demands on the poultry production and welfare both intensify, the precision poultry farming technologies such as computer vision-based cybernetics system is becoming important in addressing the current issues related to animal welfare and production efficiencies. The integration of computer vision technology has become a catalyst for transformative change in precision farming, particularly concerning productivity and welfare. This review paper delineates the central role of computer vision in precision poultry farming, focusing on its applications in non-contact monitoring methods that employ advanced sensors and cameras to enhance farm biosecurity and bird observation without disturbance. We delved into the multifaceted advancements such as the utilization of convolutional neural networks (CNNs) for behavior analysis and health monitoring, evidenced by the high accuracy sorting of eggs and identification of health concerns within target-dense farm environments. The review paper underscores advancements in precision agriculture, including accurate egg weight estimation and egg classification within cage-free systems, paralleling the poultry sector's evolution towards more ethical farming practices. Moreover, it addresses the progress in poultry growth monitoring and examines case studies of commercial farms, showcasing how these innovations are being practically applied to enhance productivity and animal welfare. Challenges remain, particularly in terms of environmental variability and data annotation for deep learning models. Nevertheless, the review emphasizes the scope for future innovations like voice-controlled robotics and virtual reality applications, which have the potential to enhance poultry farming to new levels of efficiency, humanity, and sustainability. The insights assert that the continued exploration and development in computer vision technologies are not only instrumental for the poultry sector but also offer a blueprint for agricultural enhancement at large. [ABSTRACT FROM AUTHOR]

• Microwave detection and image processing have great potential for livestock farming. • Continuous detection and early disease and aberrant behavior identification are made possible by non-contact sensing technology. • Integration of non-contact sensing technologies in smart farms to enhance precision livestock farming practices and improve efficiency. • This paper also summarizes the advantages, challenges and prospects of non-contact sensing technology. More efficient, sustainable, and precise agri-sensing technologies will be adopted in the future to promote precision livestock farming (PLF). Through a detailed survey of research related to sensing technology, we believe that non-contact sensing technology (infrared detection technology, Microwave radiometry technology, Image processing and machine vision, and acoustic detection technology) hold promise for application in intensive livestock production system. Animal welfare has always been a concern of the scientific community, as well as improving production efficiency has been the goal of the intensive livestock industry. With the help of the Internet of Things, big data, artificial intelligence, etc., non-contact sensing technology can better play its role. In this paper, we analyzed the background and technical characteristics of non-contact sensing technology and proposed a new application for applying non-contact sensing technology to intensive livestock farming. These technologies are now gradually being applied and practiced in the livestock industry and will facilitate physiological detection in the intensive livestock industry, potentially protecting animal welfare. The precise adoption of management measures based on better knowledge of animal status can also improve production efficiency. In addition, this paper discusses the advantages, challenges, and prospects of non-contact sensing technology in livestock farming. [ABSTRACT FROM AUTHOR]

• The model can detect abnormal manure and locate the cage of sick hens. • The YOLOv5 model was used to detect abnormal manure. • SuperPoint-SuperGlue was used for distance estimation. • A method for reducing cumulative errors in distance estimation. • The backtracking accuracy of the model reaches 91.6%. Digestive diseases are common poultry diseases that significantly affect production and animal welfare. Early warning of digestive disease outbreaks is essential in poultry breeding. In conventional cage laying hen systems, breeders usually spend 1–2 days randomly selecting hens for health assessment. This process is time-consuming, labor-intensive, and may cause stress to the hens, which goes against animal welfare principles. In this paper, a model structure was proposed to locate the cage of sick hens based on manure images using digital image processing and deep learning. Abnormal features of poultry manure was extracted using the YOLOv5 and combined with the motion information of the manure belt estimated by the SuperPoint-SuperGlue model to construct an early warning and traceability system for diseases in a layered cage farming mode. This system enables the automatic detection and traceability of digestive diseases in layer-raised hens. The results showed that YOLOv5 can effectively recognize abnormal chicken manure with precision rate, recall rate, mAP@0.5 and mAP@[0.5:0.95] of 95.7 %, 95.4 %, 98.1 % and 84.3 %. The SuperPoint-SuperGlue model demonstrated strong data fitting capabilities and remarkable robustness in long-distance monitoring of manure belt movement, with MAE, RMSE, and MAPE values of 0.19 m, 0.14 m, and 0.34 %, respectively. After combining the SuperPoint-SuperGlue with a self-designed error correction module, the traceability accuracy of the model was validated through 48 field tests, achieving a correct tracing rate of 91.6 %. These research results can be used for providing early warnings of digestive system diseases in caged chickens, providing a prerequisite for the development of intelligent disease diagnostic equipment. [ABSTRACT FROM AUTHOR]

• Pixel-wise facial landmark segmentation was done in cattle infrared images. • Ghost modules were introduced to reduce computational complexity. • ECA modules further improved segmentation performance. • Temperature parameters obtained on the segmentations were close to the ground truth. • Automated temperature collection will promote relevant uses of infrared thermography. In cattle, facial temperatures captured by infrared thermography provide useful information from physiological aspects for researchers and local practitioners. Traditional temperature collection requires massive manual operations on relevant software. Therefore, this paper aimed to propose a tool for automated temperature collection from cattle facial landmarks (i.e., eyes, muzzle, nostrils, ears, and horns). An improved UNet was designed by replacing the traditional convolutional layers in the decoder with Ghost modules and adding Efficient Channel Attention (ECA) modules. The improved model was trained on our open-source cattle infrared image dataset. The results show that Ghost modules reduced computational complexity and ECA modules further improved segmentation performance. The improved UNet outperformed other comparable models on the testing set, with the highest mean Intersection of Union of 80.76% and a slightly slower but still good inference speed of 32.7 frames per second. Further agreement analysis reveals small to negligible differences between the temperatures obtained automatically in the areas of eyes and ears and the ground truth. Collectively, this study demonstrates the capacity of the proposed method for automated facial temperature collection in cattle infrared images. Further modelling and correction with data collected in more complex conditions are required before it can be integrated into on-farm monitoring of animal health and welfare. [ABSTRACT FROM AUTHOR]

• Movesense enables continuous calf heart rate monitoring for extended periods. • Weak correlation between rectal and superficial armpit temperature in calves. • Movesense system attains a high level of posture accuracy compared manually visually labeling of video recordings of individual calves. • Study prospect: calf health & welfare tracking by heart, move & temperature data. In the last decade, development of continuous sensor based monitoring systems for farm animals (Precision Livestock Farming (PLF)) has markedly increased with the aim to automate and improve disease detection. In calves, multiple technologies exist, but most detect only a single parameter, and none are available for continuous heart rate and electrocardiogram (ECG). The objective of this paper was to develop and validate a sensor system, derived from a commercially available sensor for humans, for heart (ECG, heart rate), movement (inertial measurement unit with nine degrees of freedom (IMU9)) and temperature data. Data is sent over Bluetooth Low Energy (BLE) to a gateway equipped with a built-in camera for monitoring individual calves. The gateway transmits the data using the Message Queuing Telemetry Transport (MQTT) protocol to a server for long term storage. The data can also be shown on a real-time dashboard. Results indicate a very high correlation for heart rate compared with a widely used veterinary sensor (Pearson correlation coefficient (r) = 0.98), and a high level of accuracy for the movement data for lying versus standing detection (94.3 %). However, a lower correlation between sensor temperature and rectal temperature measurements was observed (r = 0.56). In conclusion, the adapted Movesense system exhibits the capability to accurately measure heart rate and detect the position (lying – standing) of calves but was insufficient as such for temperature monitoring. This sensor system has potential not only for disease detection, but equally for continuous, non-invasive heart rate monitoring, which can be used as an indicator (besides others) for animal welfare. [ABSTRACT FROM AUTHOR]

In intensive farming systems the facilities have a central role on both animal welfare and animal production all this paving the way of researching new housing systems and management strategies for reducing the impacts. In particular, in the dairy cattle sector, the early detection of irregular productions is fundamental for animal health and safety. On the other hand, despite the growing interest concerning the modelling and forecasting daily production data, there is lack of studies devoted to identification of anomalous data. To this regard, in this work, a data driven approach for detecting milk production and behaviour anomalies is presented and applied to three farms selected as case study. The DAIRY CHAOS procedure proposed in this paper bases on two numerical algorithms having the scope of separately detect anomalies daily data for a single cow. Both the algorithms presented hereinafter have statistical foundations and take in input daily resting time, milk yield and climate data respectively recorded by pedometer worn by the cow, automatic milking robot and a thermo-hygrometer data logger installed in each barn. The first algorithm takes into consideration three indicators, namely Relative Yield Difference, Relative Laying time Difference and Cumulative Discomfort Index. An anomaly, i.e. a deviation from a normal value, is determined, for a single cow, for a specific day, if the three conditions assessing a noticeable deviation from the normal values of the three indicators above are contemporary verified. The second algorithm, by means of a multifit procedure, introduces the concept of reliability of robust statistics and provides statistically solid, since not affected by outlier values, milk yield and laying time trends for each animal. The application, in a production context, of the procedure proposed here can result extremely useful for the identification of animals suffering heat stress and therefore can become a support to the farmer's decisions for the mitigation of the heat stress effects and a more efficient management of the animals. • A data driven approach for detecting milk production anomalies is proposed. • The DAIRY CHAOS procedure represents a robust procedure for heat stress detection. • An anomaly in the daily indicators labels the cow as suffering heat stress. [ABSTRACT FROM AUTHOR]

• Cows and their abdomen areas were located using YOLOv5. • Individual cows were measured remotely, non-invasively, and simultaneously. • Heat stress was detected when the respiration rate reached 60 breaths per minute. In cattle, respiration rate (RR) provides researchers and practitioners with valuable physiological information. However, traditional visual observation requires massive labour and is impractical for large-scale commercial farms. Recently developed vision-based RR measurement methods are highly limited to measuring a small number of animals in a controlled environment. Therefore, this paper aimed to propose a vision-based multi-object RR measurement method for dairy cows lying in free stalls. An RGB camera was aimed at the lying zone and was able to cover about 16 stalls. The proposed framework first utilised two YOLOv5-based networks to segment cow instances and detect cow flank objects. Next, an object tracker was used to link the predictions of each cow throughout the video clip. The Lucas-Kanade optical flow was then calculated specifically on the overlapped area of the cow mask and the flank bounding box. Finally, RR was extracted using Fast Fourier Transform. The results show that the proposed method had a precise RR measurement with a correlation coefficient of 0.944, a root mean square error of 5.35 breaths per minute, and an intraclass correlation coefficient of 0.974 when compared to visual observation. The piecewise regression models identified a change in RR when the ambient temperature reached 23.6 °C or the temperature-humidity index reached 72. The corresponding RR thresholds were 60.9 and 60.2 breaths per minute, respectively. Collectively, these results can be used to inform an automated local cooling system, e.g., fans in the lying area. However, more experiments and calibration with data collected using more cost-effective video recording systems are required before this technology can be applied on farms. [ABSTRACT FROM AUTHOR]

[Display omitted] • A camera system is presented to monitor broiler activity using 2D computer vision. • A convolutional neural network-based chicken detector was developed. • Individual birds were tracked using a minimum cost network flow based tracker. • A clustering algorithm was applied to group trajectories into 3 activity clusters. • The potential of the system to monitor short-term changes in activity is discussed. To monitor changes in broiler behaviour related to animal health and welfare, farmers typically observe their flocks using manual observation. However, due to the labour intensive and continuous aspect of this task, the analysis of broiler behaviour could be automated using camera technology. This paper proposes a proof-of-concept camera surveillance system based on the automated detection and tracking of broilers to monitor activity bouts using unsupervised 2D trajectory clustering. Firstly, a convolutional neural network-based detector was trained and tested on our labelled dataset which resulted in a precision, recall and f score of 0.98, 0.90 and 0.94, respectively. Using a tracking-by-detection approach, the proposed system was able to track chickens across video frames with a multi-object tracking accuracy of 74.7%. A component-based feature saliency Gaussian mixture model (CFSGMM) was subsequently generated and applied to objectively cluster the trajectories based on their spatiotemporal information. Nineteen features were extracted from the trajectories, representing both static and dynamic characteristics of broiler movement, and three activity classes were identified: 'least active/resting,' 'active' and 'highly-active.' The proposed method was validated on one-minute monocular video sequences. CFSGMM was applied to cluster 2D trajectories relating to broiler activity bouts within the commercial rearing environment with an agreement ranging from 6.0 to 99.7% when compared to human observation. We demonstrate the potential of the computer vision system to monitor overt, short-term changes in broiler activity associated with on-farm events and discuss the opportunities of leveraging the technology to monitor longer term changes in welfare state. It is anticipated that further development of the detection and tracking systems will improve the performance of the trajectory clustering method. [ABSTRACT FROM AUTHOR]

This paper presents a novel approach to automated classification and quantification of sow postures and posture transitions that may enable large scale and accurate continuous behaviour assessment on farm. Automatic classification and quantification of postures and posture transitions in domestic animals has substantial potential to enhance their welfare and productivity. Analysis of such behaviours in farrowing sows can highlight the need for human intervention or lead to the prediction of movement patterns that are potentially dangerous for their piglets, such as crushing when the sow lies down. Data were recorded by a tri-axial accelerometer secured to the hind-end of each sow, in a deployment that involved six sows over the period around parturition. The posture state (standing, sitting, lateral and sternal lying) was automatically classified for the full dataset with a mean F 1 score (a measure of predictive performance between 0 and 1) of 0.78. Sitting was shown to present a greater challenge to classification with a F 1 score of 0.54, compared to the lateral lying postures, which were classified with an average F 1 score of 0.91. Posture transitions were detected with a F 1 score of 0.79. We automatically extracted and visualized a range of features that characterise the manner in which the sows changed posture in order to provide comparative descriptors of sow activity and lying style that can be used to assess the influence of genetics or housing design. The methodology presented in this paper can be applied in large scale deployments with potential for enhancing animal welfare and productivity on farm. [ABSTRACT FROM AUTHOR]

• A multiple sub-bands poultry vocalization endpoints detection method was proposed. • Sliding window detection was used for different duration of poultry vocalization. • Bi-LSTM used to detect Newcastle disease based on poultry vocalization. • Detection of Newcastle disease in the early stage is possible. Newcastle disease (ND) is a common disease in poultry that has a great impact on poultry health and production. ND has destructive effects on the respiratory system, such as altering the acoustic features of bird vocalizations. For this reason, this research proposed a new method, the deep poultry vocalization network (DPVN), for the early detection of ND based on poultry vocalization. The method combined multiwindow spectral subtraction and high-pass filtering to reduce the influence of noise. In order to detect poultry vocalizations automatically, a multiple subband poultry vocalization endpoint detection method was proposed in this paper. The performance of the detection method was evaluated using the intersection-over-union (IOU) between the detected vocalizations and ground truth vocalizations. The recall of the detection method was 95.11%, and the precision was 96.54%. The audio features of poultry vocalizations are extracted by sound technology and used as the input of a deep learning network to recognize the vocalizations of poultry with Newcastle disease. Five different models were compared in the experiments. The method used in this paper achieves the best performance and the highest accuracy, recall and F1-score of 98.50%, 96.60% and 97.33%, respectively. The accuracies within the first, second, third and fourth days after infection were 82.15%, 90.00%, 93.60% and 98.50%, respectively. The experimental results show that the method proposed in this paper can be used to detect Newcastle disease in the early stage. It will be significant for improving animal welfare and the automated monitoring of poultry production. [ABSTRACT FROM AUTHOR]

• We introduce a goat image dataset for face detection and recognition. • Presenting an automatic goat face and facial landmark extraction from source image along with face alignment. • We report goat face recognition using convolutional neural network approach with 96.4% accuracy. Automatic identification of individual animals is an important step towards achieving accurate breeding histories, significant contributions to breeding and genetic management programmers. Currently, a different type of tags, tattoos, paint brands and microchips are used to uniquely identify livestock animals. However, the manual identification system is time-consuming, expensive and unreliable. In this paper, we present a deep learning approach that aims to fully automated pipeline for face detection and recognition of goats. Due to the high similarity and the lack of adequate dataset this problem is more complex than human face recognition. We composed two different publicly available datasets for detection and recognition. State-of-the-art convolutional neural networks (CNN) model are trained on this dataset. To evaluate the robustness of our approach, we compared it with different face recognition methods. The results show better performance with an accuracy of 96.4%. Furthermore, this paper reports 93%, 83%, 92% and 85% detection accuracy (average precision) for face, eye, nose and ear, respectively. The findings of this research could be helpful to improve animal health and welfare, individual monitoring, activity monitoring and phenotypic data collection. All the dataset and the related outcome are publicly available (https://doi.org/10.17632/4skwhnrscr.2). [ABSTRACT FROM AUTHOR]

• A decision support system for sustainable animal farming is proposed. • It relies on a data-driven learning framework based on quantile regression forests. • Fully comprehensive yet accurate models are obtained and deployed in the cloud. • The system is tested on real data concerning 20 different flocks of broilers. • Obtained global accuracy in growth, welfare and mortality models is over 81%. An efficient and sustainable animal production requires fine-tuning and control of all the parameters involved. But this is not a simple task. Animal farming is a complex biological system in which environmental parameters and management practices interact in a dynamic way. In addition, the typical non-linear response of biological processes implies that relationships across parameters that are critical to assure animal welfare and performance are difficult to determine. In this paper a novel decision support system based on environmental indicators and on weights, leg problems and mortality rates is proposed to address this issue. The data-driven modeling process is performed by a quantile regression forests approach that allows estimating growth, welfare and mortality parameters on the basis of environmental deviations from optimal farm conditions. Resulting models also provide confidence intervals able to deal with uncertainty. They are deployed in farm, offering an accessible tool for farmers, veterinarians and technical personnel. Experimental results involving 20 flocks of broiler meat chickens from different farms show the validity of the system, obtaining robust prediction intervals and high accuracy, namely over 81% for every model. The in-field use of the proposed approach will facilitate an efficient and animal welfare-friendly production management. [ABSTRACT FROM AUTHOR]

• A comprehensive review of deep learning- and wearable sensor-based methods for AAR. • Cover the most widely used sensors for the most important domesticated animal species and activities. • Challenges associated with developing deep learning models for AAR are discussed. • Potential solutions and future research directions for mentioned challenges are proposed. Animal behavior, as one of the most crucial indicators of animal health and welfare, provides rich insights into animal physical and mental states. Automated animal activity recognition (AAR) allows caretakers to monitor animal behavioral variations in real time, significantly reducing workloads and costs in veterinary clinics and promoting livestock management efficiency. With recent advances in sensing technologies and smart computing techniques, automated AAR has been increasingly studied, and tremendous successes have been achieved. This paper provides a comprehensive summary of recent research on AAR based on wearable sensors and deep learning algorithms. First, the commonly used sensor types and frequently studied animal species and activities are described. Then, an extensive overview of deep learning-based methods for wearable sensor-aided AAR is presented, according to the taxonomy of deep learning algorithms. We also provide a comprehensive list of publicly available datasets collected via wearable sensor-aided AAR over the past five years. This list can serve as a valuable resource for readers who wish to further explore the field of AAR. In addition, we discuss potential challenges associated with the development of deep learning models for AAR and suggest potential solutions and future research directions for these challenges. In conclusion, this review work provides rich inspiration for the future advancement of robust AAR systems based on wearable sensors and deep learning techniques. When combined with qualitative assessments of veterinary specialists, the accurate and quantitative results obtained by automated AAR systems hold the potential to significantly improve animal health and welfare. [ABSTRACT FROM AUTHOR]

Cattle detection is an important task in precision livestock farming, but it remains challenging due to the varying appearance and poses of cattle in different scenarios. In this paper, we propose a novel approach for fast cattle detection using deformable convolution and coordinate attention within YOLOv8, a SOTA object detection model. Our proposed method enhances the YOLOv8 architecture by introducing deformable convolution to capture more fine-grained spatial information and coordinate attention to emphasize important features in the detection process. We evaluate our method on a cattle dataset collected in a cattle farm and achieve superior performance compared to the baseline YOLOv8 and several SOTA object detection models. Specifically, our approach achieves a mean average precision (mAP) of 72.9% at 62.5 frames per second (FPS), which demonstrates its effectiveness and efficiency for fast cattle detection. By deploying our method on the farm's monitoring computer, our proposed approach has the potential to facilitate the development of automated cattle monitoring systems for improving animal welfare and farm management. • The DCA-YOLOv8 outperforms several SOTA object detection models. • DCA-YOLOv8 is capable of effectively dealing with various corruptions in the farm. • DCA-YOLOv8 is able to learn better interpretable features. [ABSTRACT FROM AUTHOR]

• Computer vision methods for livestock behaviour recognition were evaluated. • Journey from computer vision to deep learning was elaborated. • This paper involves with segmentation, identification and behaviour recognition. • Research trend of livestock behaviour recognition was illustrated from four aspects. • This review could provide researchers and producers with technical references. The increasing demand for sustainable livestock products also demands new considerations in animal breeding. Breeding programs are now seeking to integrate animal behavioural phenotypes, as these relate to the productivity, health and welfare of the animals and thereby can influence yield and economic benefits in the industry. Traditional manual observation of pig behaviour is time-consuming, laborious, subjective, and difficult to achieve in continuous and large-scale operations. It is not surprising that computer vision technology with the advantages of being objective, non-invasive and continuous has been widely researched for its use in the recognition of livestock behaviours over recent years. Nevertheless, in studies of livestock behaviour recognition, computer vision technology faces some challenges, e.g., complex scenes, variable illumination, occlusion, touching and overlapping between livestock, which has limited the fast translation of technology to industry. On the other hand, deep learning technology has proven to solve these difficulties to a certain extent and is being adopted to recognise livestock behaviours. This paper mainly evaluates the recent developments in computer vision methods for recognition of these behaviours in pigs and cattle. The focus on these species is made possible by the number of studies exist quantifying behaviours that are of importance for their health, welfare and productivity such as aggression, drinking, feeding, lameness, mounting, posture, tail-biting and nursing. This review paper especially analyses the development of image segmentation, identification and behaviour recognition using tradition computer vision and more recent deep learning methods, and evaluates the evolution of key research in the field. We elaborate the research trend of livestock behaviour recognition from four aspects, i.e., development of robust livestock identification algorithms, recognition of livestock behaviours for different growth stages, further quantification of the results of behaviour recognition, and building evaluation system of growth status, health and welfare. [ABSTRACT FROM AUTHOR]

• A novel occlusion-resistant instance segmentation method was developed for lactating piglets. • Predicted object centers were clustered and traced back to achieve piglet instance segmentation. • Mean average precision of 84.1 was achieved on the test set under heavy occlusion. • The predicted center is the occlusion-resistant representation of the location of an occluded object. Computer vision enables the development of new approaches to monitor the behavior, health, and welfare of animals. Instance segmentation is a high-precision method in computer vision for detecting individual animals of interest. This method can be used for in-depth analysis of animals, such as examining their subtle interactive behaviors, from videos and images. However, existing deep-learning-based instance segmentation methods have been mostly developed based on public datasets, which largely omit heavy occlusion problems; therefore, these methods have limitations in real-world applications involving object occlusions, such as farrowing pen systems used on pig farms in which the farrowing crates often impede the sow and piglets. In this paper, we adapt a Center Clustering Network originally designed for counting to achieve instance segmentation, dubbed as CClusnet-Inseg. Specifically, CClusnet-Inseg uses each pixel to predict object centers and trace these centers to form masks based on clustering results, which consists of a network for segmentation and center offset vector map, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, Centers-to-Mask (C2M), and Remain-Centers-to-Mask (RC2M) algorithms. In all, 4,600 images were extracted from six videos collected from three closed and three half-open farrowing crates to train and validate our method. CClusnet-Inseg achieves a mean average precision (mAP) of 84.1 and outperforms all other methods compared in this study. We conduct comprehensive ablation studies to demonstrate the advantages and effectiveness of core modules of our method. In addition, we apply CClusnet-Inseg to multi-object tracking for animal monitoring, and the predicted object center that is a conjunct output could serve as an occlusion-resistant representation of the location of an object. [ABSTRACT FROM AUTHOR]

Most farmers do not have the skills and time to utilize new Precision Livestock Farming (PLF) technologies effectively. It is time consuming to combine and analyse the data coming from sensors in different formats and frequencies. As part of the EU-PLF project, the authors have developed a visualisation tool to bring together and analyse the scattered data, and present them in an easy to use format to the end user. The correct use of these data might improve animal welfare, and reduce emissions through the application of PLF techniques. Data were collected at five broiler farms and ten pig farms across Europe. At the farms, a number of variables were automatically measured including climate data, production data, environmental data, and data on animal behaviour coming from cameras and microphones. Simultaneously, the welfare of the animals was assessed by trained assessors on a regular basis by using the standardized Welfare Quality protocol. All data were gathered, stored and processed on a daily basis, and visualised on a web-based tool. End-users of the tool were trained on how to interpret the available information on the visualisation tool. This paper presents the development of this PLF data visualisation tool. The farmer’s use of this tool and the early warning capabilities are described by six case studies. The selected farmers participated actively in evaluating its usefulness, resulting in a web-based visualisation tool that is practical and useful for both the farmer and other stakeholders (e.g. vets, advisors, researchers, etc.). [ABSTRACT FROM AUTHOR]

Changes in the feeding pattern of a pig may indicate disease or welfare issues and can also be strongly related to productivity changes. Automatic measurements of the feeding patterns of group-housed growing-finishing pigs can enable the detection of such feeding related changes. In this paper, a high frequency (HF) Radio Frequency Identification (RFID) system was used to register pigs’ feeding patterns at a commercial-type feed trough. The raw RFID registrations had to be translated into feeding visits, as there were irregular time gaps between the registrations that were inherent to the system for this application. Data from two experiments were used: one used behavioural observations of six pigs during three days with the aim of finding the best method for visit construction (experiment 2, E2), whilst the other experiment used feeding pattern observations of 20 pigs on one day as an independent validation data-set (experiment 1, E1). Three methods for visit construction were proposed: (1) visit criteria (a bout criterion and a minimum duration criterion), (2) a digital filter (using a well-defined time delay and a threshold for feeding) and (3) erosion and dilation operations from mathematical morphology. The best results were obtained using a bout criterion equal to 10 s for two RFID tags per pig or 20 s for one RFID tag per pig. Removal of short visits was found to be unnecessary. This optimal method is equal to a closing operation (a dilation followed by an erosion). For the RFID system with the optimal bout criterion, using two tags per pig, average sensitivity was 83%, specificity 98%, accuracy 97% and precision was 75% for E2. Applying the same bout criterion on the independent test set (E1), resulted in an average sensitivity of 80%, a specificity of 99%, an accuracy of 98% and a precision of 78%. Performance of the method tended to be lower when using one tag per pig. Regression analysis revealed that the observed feeding duration was well-correlated with RFID-based feeding duration (R 2 = 0.86 for two tags per pig, E1 + E2). Other observed variables (number of feeding visits and inter-visit interval) were harder to predict by the RFID system. From these results, there was concluded that feeding patterns can be recorded accurately in an automatic way if the registrations from the RFID system are analysed using an appropriate bout criterion. [ABSTRACT FROM AUTHOR]

Manual observation and classification of animal behaviors is laborious, time-consuming, and of limited ability to process large amount of data. A computer vision-based system was developed that automatically recognizes sow behaviors (lying, sitting, standing, kneeling, feeding, drinking, and shifting) in farrowing crate. The system consisted of a low-cost 3D camera that simultaneously acquires digital and depth images and a software program that detects and identifies the sow’s behaviors. This paper describes the computational algorithm for the analysis of depth images and presents its performance in recognizing the sow’s behaviors as compared to manual recognition. The images were acquired at 6 s intervals on three days of a 21-day lactation period. Based on analysis of the 6 s interval images, the algorithm had the following accuracy of behavioral classification: 99.9% in lying, 96.4% in sitting, 99.2% in standing, 78.1% in kneeling, 97.4% in feeding, 92.7% in drinking, and 63.9% in transitioning between behaviors. The lower classification accuracy for the transitioning category presumably stemmed from insufficient frequency of the image acquisition which can be readily improved. Hence the reported system provides an effective way to automatically process and classify the sow’s behavioral images. This tool is conducive to investigating behavioral responses and time budget of lactating sows and their litters to farrowing crate designs and management practices. [ABSTRACT FROM AUTHOR]

This paper illustrates three different methods which can be used to quantify passive ventilation inside small livestock trailers. Accomplishing this is crucial to determine whether animal welfare ventilation standards are met during road transport. High-fidelity Computational Fluid Dynamics (CFD) simulations are conducted for a detailed livestock trailer containing 48 sheep and coupled to a generic towing vehicle. An initial correlation study shows that the observed airflow patterns share the general flow structure with an equivalent but simplified geometry which has previously been validated. Further analysis of the detailed geometry shows that twice as much air flows through the upper deck of the trailer compared to the lower deck which is due to a tailboard vent allowing easier passage to the low-pressure wake. Ventilation is analysed using the net flow rate through vent openings along the side of the trailer; it is found to be below the regulatory threshold for vehicle speeds of up to 17.9 m/s on the lower deck and 8.9 m/s on the upper one. One drawback of this analysis method is that it does not consider ventilation efficacy within the interior. In order to address this, two additional methods are proposed to identify ventilation rates using (i) passive scalars and (ii) a particle tracking algorithm. The former method suggests that the greatest residence time is 30.0 s with the latter indicating a value of 19.5 s; both of these occur in the vicinity of one animal muzzle, in a stagnant airflow region of the lower deck. All three analysis methods are compared to EU-specified regulatory ventilation rates in the form of a ventilation coefficient. Overall, ventilation is shown to be significantly better on the upper deck of the trailer, satisfying current livestock transport regulations at moderate to high vehicle speeds, whereas ventilation is markedly less on the lower deck and potentially below the regulatory threshold. [ABSTRACT FROM AUTHOR]

Humane handling and slaughter of livestock are of major concern in modern societies. Monitoring animal wellbeing in slaughterhouses is critical in preventing unnecessary stress and physical damage to livestock, which can also affect the meat quality. The goal of this study is to monitor pig herds at the slaughterhouse and identify undesirable events such as pigs tripping or stepping on each other. In this paper, we monitor pig behavior in color videos recorded during unloading from transportation trucks. We monitor the movement of a pig herd where the pigs enter and leave a surveyed area. The method is based on optical flow, which is not well explored for monitoring all types of animals, but is the method of choice for human crowd monitoring. We recommend using modified angular histograms to summarize the optical flow vectors. We show that the classification rate based on support vector machines is 93% of all frames. The sensitivity of the model is 93.5% with 90% specificity and 6.5% false alarm rate. The radial lens distortion and camera position required for convenient surveillance make the recordings highly distorted. Therefore, we also propose a new approach to correct lens and foreshortening distortions by using moving reference points. The method can be applied real-time during the actual unloading operations of pigs. In addition, we present a method for identification of the causes leading to undesirable events, which currently only runs off-line. The comparative analysis of three drivers, which performed the unloading of the pigs from the trucks in the available datasets, indicates that the drivers perform significantly differently. Driver 1 has 2.95 times higher odds to have pigs tripping and stepping on each other than the two others, and Driver 2 has 1.11 times higher odds than Driver 3. [ABSTRACT FROM AUTHOR]

• Proposed interdisciplinary terms and concepts in precision management of animals. • Sources of uncertainty in animal behavior labeling for computer vision applications. • Directions of advanced computer vision algorithms in pig production facilities. • Image dataset construction workflows for computer vision algorithm development. Surveillance and analysis of behavior can be used to detect and characterize health disruption and welfare status in animals. The accurate identification of changes in behavior is a time-consuming task for caretakers in large, commercial pig production systems and requires strong observational skills and a working knowledge of animal husbandry and livestock systems operations. In recent years, many studies have explored the use of various technologies and sensors to assist animal caretakers in monitoring animal activity and behavior. Of these technologies, computer vision offers the most consistent promise as an effective aid in animal care, and yet, a systematic review of the state of application of this technology indicates that there are many significant barriers to its widespread adoption and successful utilization in commercial production system settings. One of the most important of these barriers is the recognition of the sources of errors from objective behavior labeling that are not measurable by current algorithm performance evaluations. Additionally, there is a significant disconnect between the remarkable advances in computer vision research interests and the integration of advances and practical needs being instituted by scientific experts working in commercial animal production partnerships. This lack of synergy between experts in the computer vision and animal health and production sectors means that existing and emerging datasets tend to have a very particular focus that cannot be easily pivoted or extended for use in other contexts, resulting in a generality versus particularity conundrum. This goal of this paper is to help catalogue and consider the major obstacles and impediments to the effective use of computer vision associated technologies in the swine industry by offering a systematic analysis of computer vision applications specific to commercial pig management by reviewing and summarizing the following: (i) the purpose and associated challenges of computer vision applications in pig behavior analysis; (ii) the use of computer vision algorithms and datasets for pig husbandry and management tasks; (iii) the process of dataset construction for computer vision algorithm development. In this appraisal, we outline common difficulties and challenges associated with each of these themes and suggest possible solutions. Finally, we highlight the opportunities for future research in computer vision applications that can build upon existing knowledge of pig management by extending our capability to interpret pig behaviors and thereby overcome the current barriers to applying computer vision technologies to pig production systems. In conclusion, we believe productive collaboration between animal-based scientists and computer-based scientists may accelerate animal behavior studies and lead the computer vision technologies to commercial applications in pig production facilities. [ABSTRACT FROM AUTHOR]

• A deep learning model based on YOLO v5 was used for sheep behavior recognition. • Two types of data sets were constructed for YOLO v5 model training. • The characteristics of different training set data were analyzed and compared. • The magnitude of the training set used for model training was determined. The behavior of animals can reflect animal health status and physiological stages. Automatic recognition of animal behavior can provide a powerful tool for improving the breeding management level and ensuring animal welfare. Although the image-based deep learning algorithms can be used to recognize animal behavior automatically, there has been no unified and clear conclusive definition of the characteristics and amount of training data of the deep learning model. To address this issue, this paper proposes a deep learning model based on the YOLO v5 network for sheep behavior recognition. The proposed model is trained using various types of datasets divided into two categories based on whether the training data have high similarity data characteristics with the test data. The model training included several rounds with different training data amounts. The experimental results show that if the training and testing data have the same characteristics, only 1,125 images per behavior type are required to achieve the recognition precision of 0.967 and recall of 0.965. However, when training and test data have different characteristics, it is challenging to achieve such high precision and recall values, even when using many datasets. These results demonstrate that in a structured scenario, when training data and data generated in the practical application have consistent characteristics, there is no need to use a large amount of training data. As a result, deep learning deployment efficiency in practical applications can be improved. [ABSTRACT FROM AUTHOR]

• Both acoustic and deep features were investigated for pig cough. • A novel acoustic and deep feature fusion framework for pig cough recognition was proposed. • The proposed method reached an accuracy of 97.35%. The recognition of pig cough sound is a prerequisite for early warning of respiratory diseases in pig houses, which is essential for detecting animal welfare and predicting productivity. With respect to pig cough recognition, it is a highly crucial step to create representative pig sound characteristics. To this end, this paper proposed a feature fusion method by combining acoustic and deep features from audio segments. First, a set of acoustic features from different domains were extracted from sound signals, and recursive feature elimination based on random forest (RF-RFE) was adopted to conduct feature selection. Second, time-frequency representations (TFRs) involving constant-Q transform (CQT) and short-time Fourier transform (STFT) were employed to extract visual features from a fine-tuned convolutional neural network (CNN) model. Finally, the ensemble of the two kinds of features was fed into support vector machine (SVM) by early fusion to identify pig cough sounds. This work investigated the performance of the proposed acoustic and deep features fusion, which achieved 97.35% accuracy for pig cough recognition. The results provide further evidence for the effectiveness of combining acoustic and deep spectrum features as a robust feature representation for pig cough recognition. [ABSTRACT FROM AUTHOR]

• It is possible to identify individual cows based on their morphological appearance. • A real-time computer vision system for automatic recognition of cows is presented. • Indications about quantity and characteristics of the images were assessed. Precision Livestock Farming relies on several technological approaches to acquire in the most efficient way precise and up-to-date data concerning individual animals. In dairy farming, particular attention is paid to the automatic cow detection and tracking, as such information is closely related to animal welfare and thus to possible health issues. Computer vision represents a suitable and promising method for this purpose. This paper describes the first step for the development of a computer vision system, based on deep learning, aiming to recognize in real-time the individual cows, detect their positions, actions and movements and record the time history outputs for each animal. Specifically, a neural network based on deep learning techniques has been trained and validated on a case study farm, for the automatic recognition of individual cows in videos recorded in the barn. Four cows were selected to train and validate a YOLO neural network able to recognize a cow starting from the coat pattern. Then, precision-recall curves of the identification of individual cows were elaborated for both the specific target classes and the whole dataset in order to assess the performances of the network. By means of data augmentation techniques, an enlarged dataset has been created and considered in order to improve the performance of the network and to provide indications to increase detection efficiency in those cases where data acquisition is not easy to be carried out for long periods. The mean average precision of the detection, ranging from 0.64 to 0.66, showed that it is possible to properly identify individual cows based on their morphological appearance and that the piebald spotting pattern of a cow's coat represents a clearly distinguishable object for a computer vision network. The results also led to obtain indications about the quantity and the characteristics of the images to be used for the network training in order to achieve efficient detections when facing with applications involving animals. [ABSTRACT FROM AUTHOR]

• Deep learning-based system for cattle behavior recognition in video. • Our framework involves appearance features and spatio-temporal information. • We introduce our cattle behavior dataset with videos recorded on different farms. • Our system can recognize 15 types of hierarchical cattle activities in realtime. • Our results can facilitate management operations by understanding cattle behavior. Behavior is an important indicator for understanding the well-being of animals. This process has been frequently carried out by observing video records to detect changes with statistical analysis, or by using portable devices to monitor animal movements. However, regarding animal welfare, the use of such devices could affect the normal behavior of the animal, and present limitations in its applicability. This paper introduces an approach for hierarchical cattle behavior recognition with spatio-temporal information based on deep learning. Our research extends the idea of activity recognition in video and focuses specifically on cattle behavior. Our framework involves appearance features at frame-level and spatio-temporal information that incorporates more context-temporal features. The system can detect (class) and localize (bounding box) regions containing multiple cattle behaviors in the video frames. Additionally, we introduce our cattle behavior dataset that includes videos recorded with RGB cameras on different livestock farms during day and night environments. Experimental results show that our system can effectively recognize 15 different types of hierarchical activities divided into individual and group activities, and also part actions. Qualitative and quantitative evaluation evidence the performance of our framework as an effective method to monitor cattle behavior. [ABSTRACT FROM AUTHOR]

• Sheep face corpus dataset of normal and abnormal sheep. • Different pre-trained architectures of deep learning performance for small datasets. • State-of-the-art models and their usages for protection of animal life's. • Layer freezing, regularization of architectures and augmentations of datasets to achieve higher accuracy. Digital image recognition has been used in the different aspects of life, mostly in object classification and detections. Monitoring of animal life with image recognition in natural habitats is essential for animal health and production. Currently, Sheep Pain Facial Expression Scale (SPFES) has become the focus of monitoring sheep from facial expression. In contrast, pain level estimation from facial expression is an efficient and reliable mark of animal life. However, the manual assessment is lack of accuracy, time-consuming, and monotonous. Hence, the recent advancement of deep learning in computer vision helps to classify facial expression as fast and accurate. In this paper, we proposed a sheep face dataset and framework that uses transfer learning with fine-tuning for automating the classification of normal (no pain) and abnormal (pain) sheep face images. Current state-of-the-art convolutional neural networks (CNN) based architectures are used to train the sheep face dataset. The data augmentation, L2 regularization, and fine-tuning has been used to prepare the models. The experimental results related to the sheep facial expression dataset achieved 100% training, 99.69% validation, and 100% testing accuracy using the VGG16 model. While employing other pre-trained models, we gained 93.10% to 98.4% accuracy. Thus, it shows that our proposed model is optimal for high-precision classification of normal and abnormal sheep faces and can check on a comprehensive dataset. It can also be used to assist other animal life with high accuracy, save time and expenses. [ABSTRACT FROM AUTHOR]

• The data acquisition system collects point clouds when pigs freely move on it. • The accuracy of non-contact pigs' body measurement is highly improved. • A registration method using cuboid is applied on reconstructing the pig contour. • A novel method is applied on curve-fitting to effectively estimate abdominal girth. Body measurement is an important technique to acquire the livestock body parameters for precision agriculture and effective management of livestock. Traditionally, the livestock body parameters are often manually measured, which suffers from several drawbacks, including unstable accuracy, low efficiency, difficulty in measuring complex parameters, and potentially negative influence on animal welfare. Considering this, this paper presents a 3D surface reconstruction and body size measurement system based on multi-view RGB-D cameras. We use Kinect depth cameras to obtain the point clouds of the freely walking pigs from three different views (i.e., upper-view, left-view and right-view). The registration parameters are obtained by a rectangular cuboid, which can be used to reconstruct three local point clouds. Thereafter, the distribution of the point cloud projection in different directions is used to locate the measuring positions, which then contribute to the precise measurement of several key parameters, such as body length, body height, body width and abdominal girth. Moreover, the polar coordinate transform is developed to improve the measurement accuracy of the abdominal girth. The experiments performed on forty pigs have shown that our average relative errors in the body length, height, width and abdominal girth are 2.97%, 3.35%, 4.13% and 4.67%, respectively, which demonstrate the feasibility and accuracy of the proposed system. [ABSTRACT FROM AUTHOR]

• Ammonia determination with low cost sensor. • Use of IoT to increase productivity in biosystems. • Low cost embedded technology in rural properties. • Precision animal science helps to increase production and well-being. Animal welfare is a major concern these days, as well as the factors that have impact on animal productivity. Constant monitoring of environmental parameters such as temperature, humidity, levels of ammonia and luminosity is crucial because these variables directly impact poultry performance and welfare. Along these lines, it is necessary to look for solutions that allow the monitoring of these variables in a practical, reliable and cheap way. In the agribusiness, the scenario for the IoT (Internet of Things) is very promising, because it has emerged to fill this gap and deliver sensed data, widely increasing the connectivity of the devices. Therefore, this paper describes a low cost, hardware and software proposal for monitoring these environmental parameters in poultry farms. In addition, experiments were done to compare the proposed prototype with commercial devices. The measured variables showed a correlation above 0.90 between the sensors used in the prototype and the calibrated equipment, also having the cost of only 13% of that of the conventional equipment. The prototype displayed favorable results, indicating that its implementation is viable. [ABSTRACT FROM AUTHOR]

• A deep learning method based on CNN and LSTM was used to recognise pig aggression. • This study directly process video episodes rather than individual frames. • CNN features were input into LSTM framework to extract spatial-temporal features. • The proposed method could recognise aggressive episodes with an accuracy of 97.2%. • A frame skipping approach can improve the accuracy to 98.4% and halve running time. Aggression is considered as a major animal welfare problem in commercial pig farming. The aim of this study is to develop a deep learning method based on convolutional neural network (CNN) and long short-term memory (LSTM) to recognise aggressive episodes of pigs. Compared to previous studies of pig behaviours based on deep learning, this study directly process video episodes rather than individual frames. In the experiment, nursery pigs (8/pen) were mixed for 3 days and then 8 h of video was recorded in each day. From these videos, 600 aggressive 2 s-episodes were manually selected and then augmented into 2400 episodes by using horizontal, vertical and diagonal mirroring. From the videos, 2400 non-aggressive 2 s-episodes were also manually selected. 80% of the data were randomly allocated as training set and the remaining 20% as validation set. Firstly, the CNN architecture VGG-16 was used to extract spatial features. These features were then input into LSTM framework to further extract temporal features. Through fully connected layer, the prediction function Softmax was finally used to determine if the current episode is aggression or non-aggression. Using the proposed method, aggressive episodes could be recognised with an accuracy of 97.2%. This result indicates that this method can be used to recognise aggressive episodes of pigs. Additionally, this paper further investigates the validity of this method under the conditions of skipping frames and reducing the episode length. The results show that a frame skipping approach whereby 30 fps is reduced into 15 fps within each 2 s-episode can improve the accuracy into 98.4% and halve the total running time. [ABSTRACT FROM AUTHOR]