Development of a CNN Approach Classification Model for Soybean (Glycine max., L.) and its Common Weeds to Support Precision Agriculture

Weeds significantly interfere with soybean production, leading to an 84% reduction in yield when uncontrolled. This study explores the application of the YOLOv8 deep learning algorithm for the effective identification and classification of soybean crops and common weeds, particularly focusing on the CLSoy 1 variety. By utilizing a low-cost 1080p webcam, images of soybeans and weeds were captured for analysis. The research implemented machine vision and image processing techniques to enhance weed management strategies, thus minimizing the reliance on chemical herbicides and addressing ecological concerns. A comprehensive dataset of 523 annotated images was processed using YOLOv8, resulting an accuracy of 83% and a precision of 88%. The Intersection over Union (IoU) score of 0.71 indicates a satisfactory overlap between predicted and actual results, although the mean Average Precision (mAP) of 0.48 recommends areas for improvement. The key findings highlight the potential of integrating advanced technologies in precision agriculture, underscoring the need for sustainable and efficient weed management solutions. Future work will aim to enhance detection capabilities, possibly with higher-quality imaging equipment and advanced data augmentation techniques, to further improve the model's performance in diverse agricultural environments.