Research on Motion Planning Model and Control Strategy of Corn Inspection Robot

Abstract

Abstract In order to understand the growth condition of corn crop in real time, this paper designs an inspection robot for corn growth information collection, and proposes a path planning method for corn growth information collection robot based on Yolo v4. Firstly, the maize inspection robot adopts Kinect v2.0 camera to collect images of maize rootstalk information, and processes the images through the IPC to build a sample library of maize rootstalk, on which the Yolo v4 network model is used to train the maize rootstalk images. The accuracy of Yolo v4 is found to be 10.48% higher than that of the Faster R-CNN model. After that, we fit the robot's walking path based on the recognition results, and convert the image pixel coordinates into spatial coordinates through coordinate conversion to obtain the robot's walking path in the actual field environment. Finally, a prototype inspection robot is built and tested in a corn field. The results showed that the inspection robot is stable and did not lose control, and at the same time, it could effectively collect various information in the process of corn growth, which provided a means to realize human-robot separation.