Real-time detection of formation head vehicles based on improved YOLOv5s network
Zhigang Ren, Guoquan Ren, Dinhai Wu · 2022
With the development of military modernization in various countries, mechanized troops are working towards unmanned combat. The control and command of formation operations is an important part of unmanned operations. How to achieve stable formation autonomously following people in a complex battlefield environment has great technical difficulties. In this paper, the monocular vision sensor is used for acquisition, and the YOLOv5s neural network structure is modified to realize the rapid identification of the target vehicle. The simulation experiment is used to simplify the model, and the local path planning is designed for the complex environment. Better experimental results were obtained. In this paper, the network structure is optimized based on the existing YOLOv5s, which reduces the computational difficulty, improves the network solution efficiency, and improves the network real-time detection.