Computer Vision Assisted Autonomous Intra-Row Weeder

Abstract

Vegetable crop production is the rudimentary process for the sustainability of mankind on this planet. Technological advancements have been shaping the crop production in a commendable way nonetheless, weeds play a deleterious role for the growth of the crops. Weed management has hence become the alarming solution for increasing the yield. Even though manual weeding is being practiced by many nonetheless, labor costs, time and tedium have become the major constraints for the crop production. The introduction of the chemical methods of weed control has affected the growth of crops. However, the continuous use of herbicide-resistant weedicides has a serious impact on crops and the environment. So an increasing demand for chemical-free food has led to the investigation of alternative methods of weed control. Most of the methods of mechanical weeding doesn't produce the accurate result and moreover, existing intra-row weeders have limitations. This paper proposes the construction and working of an autonomous mechanical weeder. It consists of an actuator which is developed to mechanically remove intra-row weeds. The mechanically weeding actuator consists of an integrated servo motor which is combined with the computer vision assisted system for detecting the crop plant locations and guiding the weeding actuator to execute mechanical weeding operations without damaging the crops. The image extraction is based on a novel algorithm which effectively works with the encoding system of the robot movement with a precision of plus or minus 1cm. The accuracy of the system is found to be 96.3% using haar cascade classifier using Open-CV open source framework.