Agricultural Science Digest

India faces a critical shortage of farm labour. Out of basic farming operations, weeding operations are not fully mechanized in India. Traditional weed control methods are tedious, labour-intensive and costly.

The objective of this study was to develop a prototype of an autonomous weeder for inter and intra-row weeding. The materials were used include a DC motor operating on electric rechargeable batteries, a camera for taking photographs of the intrarow space, an inter-and intra-row weeding tool assembly, a Raspberry Pi to deploy a machine learning object detection model to detect weeds and crops, to control a motor and to control an intra-row weeding tool through the pneumatic actuator. While the intensive image dataset was prepared from a cultivated spring onion farm on a test soil bin over the University campus, moreover the small-sized YOLOv8n model worked accurately on the Raspberry Pi.

YOLOv8n achieved a mean average precision (mAP) of 0.85 at the Intersection of the Union (IoU) threshold of 50%. The results indicated that the prototype removed both types of weeds with an overall weeding efficiency of 86%, crop damage of 4%, a field capacity of 0.023 hectares per hour and a performance index 1726.26 using 0.11 horsepower. It was concluded that the study contributed a spring onion-weed dataset and a Weeding Agrobot prototype to test the described concept, which would help farmers reduce labour dependency and the cost of cultivation.