Quadruped Robot Platform for Selective Pesticide Spraying

Abstract

Effective control of disease and pest infection is vital for maximizing crop yields, and pesticide spraying is a commonly used method for achieving this goal. This study proposes a novel approach to selective pesticide spraying using a quadruped robot platform, which we tested in a broccoli field. We developed an algorithm to detect and track worms based on our proposed Histogram of Oriented Gradients and Support Vector Machine (HOG-SVM) techniques, integrated with the recent object detection and tracking methods. Our platform was tested by traversing the furrows between the broccoli crop lines and continuously scanning to detect cabbage worms. Our experiments demonstrate that the proposed HOG-SVM algorithm successfully reduced the false positive rate of real-time worm detection by reducing around 90% for the imitation environments and around 60% for the actual field.