Design of moving object detection algorithm based on RoboMaster

Abstract

RoboMaster is a global high-level robot competition, in which the vision system of infantry robot is studied in this paper. In complex environment, the existing vision detection algorithms have some problems, such as unstable recognition, low detection accuracy and long running time. The improvement and optimization of the algorithm are as follows: image quality detection is carried out when brightness is bright or brightness is abnormal; gamma transform is used in the pre-processing process; maximum class difference method is used in image segmentation; SVM algorithm is used to train the selection of light strips; directional gradient histogram is used to extract features; and PNP algorithm is used to calculate the single-visual distance measurement and camera attitude. The experimental results show that the vision detection algorithm designed in the complex scene can realize the recognition of moving objects.