Deep Learning for Pink Bollworm Detection and Management in Organic Cotton Farming Practices

Deep learning (DL) for the organic cotton pest detection and management using Agri-Bot has introduced a new practice of pest detection and pest management. The robotic pest detection for pink bollworm with its characterization is carried out in YOLO v8 software by trained model. A mean Average Precision (mAP) value is observed as 67.1%. The peripherals such as Intel Real-Sense camera, Data Storage System (DSS) mounted on Agri-Bot for field experiment and the health condition monitoring system are developed for organic cotton production. The accuracy and performance of the model for Rasi variety was decreasing at 70cm height from 87.50% when speed varies from 1 to 3.6 km/h, whereas for Bt-cotton species at 50 cm height and 1 km/h speed exhibits a model accuracy up to 78.12%, and for PDKV variety at a range 50–70 cm shows 65–54%, respectively. Additionally, the study delves into agricultural machinery's operational parameters, showcasing diverse theoretical capacities ranging from 0.067 to 0.2412 ha/h at speeds of 1, 2.5, and 3.6 km/h. Actual machine performance capacities vary from 0.058 to 0.177 ha/h, with average accuracy range between 74 and 81%. This DL model can be more trained for high accuracy level management of overall crop health condition monitoring.