A Vision System for Autonomous Weed Detection Robot

Abstract

In this paper, a vision guidance system for automated weed detection robot is presented. The developed vision system use series of image processing techniques to detect the inter-row space between the crops and then calculate the current pose and orientation with the help of Hough transform. The dynamic model is used for evolution of values over time and to predict the changes in pose and orientation from frame to frame. The vision system is implemented and simulated in Matlab, and it is observed that the developed system successfully detects and calculates the pose and orientation of the crop boundaries on both real and synthetic images.