Research on Coded Target Recognition Based on Improved YOLOv5

To deal with the problem that the recognition of coded target in visual measurement is easily affected by factors such as angle, light intensity, noise, an improved YOLOv5-based algorithm is proposed to realize the detection of circular coded target. First, the acquired images are preprocessed to recover useful and real information; Secondly, improved YOLOv5, add Convolutional Block Attention Module to improve detection accuracy, a multi-scale detection layer is added to enhance detection of small-scale targets, and a SIoU loss function is introduced to help training. Finally, verification of the algorithm feasibility by experiments.