Search

This research addresses challenges in capsicum peduncle detection in night-time greenhouse environments, including low light, uneven illumination, and shadows, using advanced computer vision models. A dataset of 200 images was curated, capturing diverse distances, heights, occlusion levels, and lighting conditions, and was rigorously pre-processed and augmented. Two YOLOv9 instance segmentation variants, YOLOv9c-seg and YOLOv9e-seg, were custom-trained and fine-tuned using Google Colaboratory. YOLOv9c-seg (56.3 MB) achieved superior mean Average Precision (mAP) scores of 0.751 (box) and 0.725 (mask), outperforming YOLOv9e-seg (121.9 MB) with mAP scores of 0.674 (box) and 0.658 (mask). Grounded SAM, a zero-shot segmentation model, achieved maximum peduncle detection confidences of 59% and 49% with positional prompts. Comparative testing on 50 images containing 70 capsicums showed YOLOv9c-seg achieving mean precision, recall, and F1-scores of 0.93, 0.86, and 0.89, respectively, outperforming Grounded SAM (0.86, 0.70, and 0.77). This study highlights the efficacy of single-shot versus zero-shot segmentation models for automated capsicum peduncle detection in controlled agricultural environments, offering insights into model performance and future research directions for model optimization and dataset expansion. [ABSTRACT FROM AUTHOR]

Tractor engine performance is critical for optimizing energy efficiency in tillage operations across various soil conditions. This study introduces an artificial neural network (ANN) based novel approach for predicting and optimizing specific fuel consumption (SFC) and fuel consumption per tilled area (FCA), during tillage operations. Field tests were conducted to collect the data on rear axle torque, forward velocity, fuel consumption, throttle position, engine speed, and depth of operation under soft, medium, and hard soil conditions while using a reversible MB plough, offset disc harrow, and rigid tyne cultivator. Tractor engine performance was predicted using ANN and optimized by matching the required rear axle power with the available rear axle power. Optimal operating parameters for three distinct tillage implements, each associated with varying soil types and tillage depths, were determined based on minimizing SFC and FCA. An integrated ANN-based user advisory system was developed to provide tractor engine performance predictions and recommendations for optimal operating parameters to maximize fuel efficiency. The recommendations were validated with actual field data, resulting in Mean Absolute Percentage Errors of 7.56% and 8.98% for MB plough, 11.19% and 9.36% for offset disc Harrow, and 10.90% and 10.57% for cultivator, respectively, for SFC and FCA. Highlights: ANN-based novel approach is proposed to optimize fuel efficiency during tillage Advisory system is developed to recommend parameters for optimum tillage Suggestions for minimum SFC and FCA are provided in soft, medium and hard soil The accuracy and reliability of advisory system are proven by field validation [ABSTRACT FROM AUTHOR]

Knowing tractor drawbar pull is crucial to ensure the tractor can handle the required workload efficiently and safely, preventing soil damage and optimising field productivity. The present study proposes a novel approach for tractor drawbar pull prediction by utilising the tractor's geometric parameters and forward speed to develop a cloud-infused, server-less, machine learning-based real-time generalised tractor drawbar pull prediction model for any tractor between the 6-58 kW power range. The drawbar pull prediction models from ANN and six ML algorithms were developed, and the data analysis with hyperparameter tuning concluded that the Extreme Gradient Boosting (XGB) ML model outperformed the other ML models. A reasonable accuracy with R2 = 0.93 and MAPE = 6.77% was achieved using the XGB ML model for a separate validation dataset, which was not used for training. Furthermore, a cloud-based serverless Android App integrated with the XGB ML-based drawbar pull prediction model was developed for real-time tractor drawbar pull prediction and monitoring during tillage operations. The field validation demonstrated the XGB ML model's generalisation ability and effectiveness, with R2 = 0.90 and maximum MAPE of 9.86%. It can be used to simulate and optimize tractor performance, guiding manufacturers in selecting geometric parameters for tractor design. [ABSTRACT FROM AUTHOR]

Mango harvesting remains predominantly manual despite technological advances in agriculture. Traditional manual harvesting methods are labour-intensive and often result in mechanical injuries, bruising, and sap-induced infections, leading to reduced fruit quality and market value. This study presents the development of a specialized end effector for robotic mango harvesting. The developed end effector is capable of simultaneously cutting and grasping the mango peduncle, maintaining fruit integrity while minimizing mechanical injuries, bruising, and sap-induced infections. Its compact and modular design enhances maneuverability and operational efficiency, providing a secure grip on mangoes during cutting and transport until release in the conveying unit, ensuring the harvested fruit remains undamaged and market-ready. Initially inspired by handheld tools, the design evolved from using a net bag to finger grippers, and finally, a compact unit with integrated pads, capable of cutting and holding the fruit simultaneously. The design process involved conceptualizing and evaluating three designs, with the most efficient one being fabricated and tested. The end effector operates using a single servo motor for both cutting and grasping the peduncle. In lab trials, it successfully picked 17 mangoes at a rotational speed of 105 rad/s. The developed end effector achieved a conveying efficiency of approximately 96.7% with only 2 out of 60 mangoes falling from it during the lab trials. The performance evaluation demonstrated the end effector's ability to consistently harvest mangoes with minimal failures, validating its practicality and robustness. [ABSTRACT FROM AUTHOR]

The fuel consumption model serves as a valuable tool for estimating real-time fuel consumption levels. An accurate fuel consumption model is crucial in providing precise information regarding fuel utilisation and motivating experts to assess the vehicle's economy. The goal of this study was to develop a novel approach by eliminating the measurement of real-time tractor PTO (power take-off) power for fuel consumption prediction and to develop a cloud-infused, server-less, machine learning (ML) based real-time generalised tractor fuel consumption prediction model for any tractor between the power range of 8–48 hp. The fuel consumption prediction models from 18 Machine Learning algorithms were developed, and the extensive data analysis with hyperparameter tuning concluded that the Gradient Boosting Regressor Machine Learning model outperformed the other Machine Learning models with reasonable accuracy (R2 = 0.999 for training and 0.914 for testing). Cloud-based serverless Web App and Android App integrated with the Gradient Boosting Regressor based fuel consumption prediction model were developed for the real-time fuel consumption prediction and monitoring of a tractor during field operations. The developed Machine Learning model predicted fuel consumption with a Mean Absolute Percentage Error of 11.43% during real-time field experiments with Mouldboard plough operation. The field validation showed the generalisation ability and efficacy of the developed model, and it can be implemented as a user advisory system for real-time energy-efficient agricultural operations. [ABSTRACT FROM AUTHOR]

Hand-pushed cone penetrometer has been utilized to assess soil compaction based on soil penetration resistance (SPR). However, the manual operation of these devices can be quite challenging and erroneous. The accuracy of the data gathered using such devices can be hindered by the operator's limitation in maintaining a constant-rate of penetration into the soil, thus preventing the attainment of a steady, predefined penetration rate. To address this issue, a constant-rate, IoT-enabled cone penetrometer is designed and constructed. The developed penetrometer measures the SPR across the soil profile with a designed lead screw arrangement that ensures a constant-rate penetration. The device comprises a mechanical system and an electronic control unit that utilizes IoT framework. The mechanical system includes a primary frame, lead screw arrangement and a standard penetrometer probe. The SPR, and depth along with coordinates of location are measured and stored on a web server. A user-friendly customizable web interface and mobile application enable real-time access. The developed penetrometer measures the SPR up to 2000 kPa up to 400 mm depth. Comparative analysis shows that the constant-rate penetrometer (CRP) provides more reliable and less variable resistance measurements compared to the hand-pushed penetrometer (HPP), with significant differences observed at multiple depths. This penetrometer maintains a constant-rate of penetration, ensuring adherence to predefined standards for accurate measurements. [ABSTRACT FROM AUTHOR]