Search

As an emerging concept, metaverse has attracted extensive attention from industry, academia and scientific research field. The combination of agriculture and metaverse will greatly promote the development of agricultural informatization and agricultural intelligence, provide new impetus for the transformation and upgrading of agricultural intelligence. Firstly, to expound feasibility of the application research of metaverse in agriculture, the basic principle and key technologies of agriculture metaverse were briefly described, such as blockchain, non-fungible token, 5G/6G, artificial intelligence, Internet of Things, 3D reconstruction, cloud computing, edge computing, augmented reality, virtual reality, mixed reality, brain computer interface, digital twins and parallel system. Then, the main scenarios of three agricultural applications of metaverse in the fields of virtual farm, agricultural teaching system and agricultural product traceability system were discussed. Among them, virtual farm is one of the most important applications of agricultural metaverse. Agricultural metaverse can help the growth of crops and the raising of livestock and poultry in the field of agricultural production, provide a three-dimensional and visual virtual leisure agricultural experience, provide virtual characters in the field of agricultural product promotion. The agricultural metaverse teaching system can provide virtual agricultural teaching similar to natural scenes, save training time and improve training efficiency by means of fragmentation. Traceability of agricultural products can let consumers know the production information of agricultural products and feel more confident about enterprises and products. Finally, the challenges in the development of agricultural metaverse were summarized in the aspects of difficulties in establishing agricultural metaverse system, weak communication foundation of agricultural metaverse, immature agricultural metaverse hardware equipment and uncertain agricultural meta universe operation, and the future development directions of agricultural metaverse were prospected. In the future, researches on the application of metaverse, agricultural growth mechanism, and low power wireless communication technologies are suggested to be carried out. A rural broadband network covering households can be established. The industrialization application of agricultural meta universe can be promoted. This review can provide theoretical references and technical supports for the development of metaverse in the field of agriculture.

In aquaculture, feeding intensity can directly reflect the appetite of fish, which is of great significance for guiding feeding and productive practice. However, most of the existing fish feeding intensity evaluation methods have problems of low observation efficiency and low objectivity. In this study, a fish feeding intensity evaluation method based on near-infrared machine vision was proposed to achieve an automatic objective evaluation of fish appetite. Firstly, a near-infrared image acquisition system was built by using near-infrared industrial camera. After a series of image processing steps, the gray level co-occurrence matrix was used to extract the texture feature variable information of the image, including contrast, energy, correlation, inverse gap and entropy. Then the data set were constructed by using these five feature variables as input vectors, and the support vector machine classifier was trained. Among them, the optimal penalty coefficient c and kernel function parameter g were selected by grid search. Finally, the trained images were used to classify the feeding images of fish. And ultimately, the evaluation of fish feeding intensity was realized. The results show that the accuracy of the evaluation could reach 87.78%. In addition, this method does not need to consider the impact of reflections, sprays and other factors on image processing results, so it has strong adaptability and can be used for automatic and objective evaluation of fish appetite, thus provide theoretical basis and methodological support for subsequent feeding decisions.

In intensive aquaculture, feeding level is directly related to production efficiency and breeding cost. How to achieve precise feeding is the key issue of aquaculture. In this paper, a prediction method of feeding quantity based on adaptive network fuzzy inference system was proposed. Taking the tilapia as research object, 2 factors of water temperature and average weight of fish were selected as input variables, and an adaptive network fuzzy inference system prediction model was established. The temperature sensor and weight scale were used to record the water temperature and the average weight of the fish, respectively. The experiment was carried out for two months. Results showed that the root mean square error (RMSE), the mean absolute error (MAE) and mean absolute percentage error (MAPE) from the NFIS feeding prediction models were 1.18, 0.74 and 0.003 1, respectively, which were far less than the index values of the fuzzy reasoning feeding amount prediction model. The validity and rationality of the feeding forecasting model were proved. It was clear that the ANFIS network forecasting ability was better than the fuzzy reasoning forecasting model. This model not only saved labor costs, but also provided technical support and theoretical support for scientific and reasonable feeding. ulture [ABSTRACT FROM AUTHOR]

In aquaculture, feeding is the primary factor determining efficiency and cost, so it is important to know when to stop feeding to maximize efficiency. Until now, fish feeding has been mostly based on artificial discrimination, which is usually time‐consuming and laborious. In recent years, intelligent feeding control according to changes in behaviour and growth status has gained increasing attention. This approach involves many methods as well as monitoring and feedback equipment and can automatically determine the feeding demands of fish. This review summarizes the development of intelligent feeding control methods, such as mathematical models, acoustic methods and computer vision, in aquaculture over the past three decades. All methods have unique application scenarios and models for the culture to which they are most suitable, and the advantages and disadvantages of each method in the laboratory as well as in pond, cage and recirculating aquaculture systems are analysed. Studies show that improvements in sensor accuracy and hardware and software processing speed have promoted the development of new technologies and methods, providing effective or potential support for intelligent feeding control. However, its accuracy and intelligent are still need to be improved to meet the needs of actual feeding scenarios. Through close collaborations between engineers and fish behaviourists, the feeding machine and system will be more elaborate and precise on the basis of the above methods, and the level of intelligence will be further improved. [ABSTRACT FROM AUTHOR]

Image segmentation is the precondition of feature extraction and recognition. In order to improve segmentation accuracy of touching objects in pest identification and counting system, an image segmentation algorithm based on shape factor and separation point location was presented. In this method, a shape factor that was defined using area and perimeter of a region was used to be a parameter to justify whether the region was one touching region or not. In this paper, the threshold of shape factor was set to be 0.50. And then, if a region was a touching one, its contour was stripped layer by layer. In each contour, it was necessary to check whether a local segmentation point existed or not. There were two types of local segmentation points. The first type was a point that was found twice in one contour at the same time, whose traversal sequence number satisfied the determined threshold condition. The second type was one point that could be found in one contour with its four connected region points at the same time, and the difference between their traversal sequence numbers satisfied the same threshold condition. Once the local segmentation point was found, two separating points of this touching region were searched and located in its original contour. The search method was based on the shortest distance between the local segmentation and the background pixel points. At last, the segmentation lines were plotted between the local segmentation and the two separating points. In order to verify the validity of the proposed algorithm, three types of touching images, such as serial connection, loop connection and hybrid connection images were used. The results showed that the proposed method could locate the local segmentation points and separating points more accurately than the watershed method. In addition, the lab and field images were used to test reliability of the proposed method. In the lab experiment, 100 yellow peach moth (Conogethes punctiferalis(Guenée) ) were collected and divided into two independent groups with 50 individuals in each one. In the field experiment, two sticky trap images of the Oriental fruit moth (Grapholitha molesta (Busck) ) were used. In this paper, three criteria such as SR (segmentation rate), SERR (segmentation error rate), and SEFR (segmentation efficiency rate) were used to evaluate the segmentation results between the proposed method and the watershed method. The results showed that, in the lab experiment, the mean SR of watershed method was more than the proposed method, but the average segmentation error rate of the proposed segmentation method was 7%, which was reduced by 6 percentage points than the watershed method. The average segmentation efficiency rate of the proposed segmentation method was 92.65%, which was more than watershed method by 5.7 percentage points. In the field experiment, the average segmentation error rate of the proposed segmentation method was 2.24%, which was reduced by 4.29 percentage points than the one of the watershed method. The average segmentation efficiency rate of the proposed segmentation method was 97.8%, which was more than the one of watershed method by 3.95 percentage points.The series of data showed that the proposed segmentation algorithm located points accurately and its invalid segmentation rate was low. The presented segmentation method for touching pest image could improve the segmentation performance and had a remarkable significance for the feature extraction and pest identification. [ABSTRACT FROM AUTHOR]

Pest identification and classification is time-consuming work that requires expert knowledge for integrated pest management. Automation, including machine vision combined with pattern recognition, has achieved some applications in areas such as fruit sorting, robotic harvesting, and quality detection, etc. Automatic classification and counting of pests using machine vision is still a challenge because of variable and uncertain poses of trapped pests. Therefore, using Pseudaletia separata, Conogethes punctiferalis, Helicoverpa armigera, Agrotis ypsilon with different poses as research objects, this paper presents a novel classification method for multi-pose pests based on color and texture feature groups and using a multi-class support vector machine. 320 images were taken using field samples with an original resolution of 4 288×2 848. The subimages of pests with 640×640 pixel size were obtained from original images for computational efficiency. Color features in RGB and HSV spaces, statistical texture features, and wavelet-based texture features were extracted. Six feature vector groups were constructed using those features. In order to select effective feature parameters of each group, a genetic algorithm was designed to optimize feature vectors based on 10-fold cross-validation. Finally, the one-against-one DAGMSVM (acronym as yet undefined) algorithm was applied to classify and recognize the four kinds of target pests and to find the best feature group. 80 images (60 for the training set and 20 for the testing set) were adopted for each species. Parameter numbers were calculated and analyzed after optimization, thus the best parameters were selected for each group. The training time of the SVM model and classification accuracy, which contains false negative and false positive details, were compared between pre-optimization and post-optimization. The results showed that the highest parameter optimization ratio is from the sixth feature group with a dimension reduction rate of 61.11%. Compared with the RGB and statistical texture feature group F2, the optimization ratio of HSV and statistical texture feature group F3 is much better; that is, the latter one is more suitable to pest classification. Analysis and comparison between the optimization results of feature group F5 and F6 shows that the latter one is more suitable for multi-pose pest classification. The modeling time of each group has been greatly decreased, especially the one of group F6 (about 8 s), which is the minimum time of all groups with a decreased rate of 74.5%. Average accuracies of all groups have been improved beyond 97%. The sixth group has the highest accuracy (100%). Consequently, the sixth feature group, the feature vector of the wavelet filter in HSV color space, is an effective feature set for use in the classification of multi-pose pests. In addition, we have found that the feature parameters are similar among the misclassification pest sets, which may be improved by increasing the number of sample images in the training set. [ABSTRACT FROM AUTHOR]