Research on Cluttered Object Recognition Algorithm Based on Improved YOLO V5

Aiming at the problems of slow speed and low accuracy in the automatic recognition of traditional robot workpieces, this paper proposes an improved workpiece recognition method based on YOLO V5. First, this paper collects photos of common workpieces by camera, creates data set independently and performs data enhancement to fill the gap of workpiece data set. Since YOLO V5 has the problems of low recognition accuracy and missed detection for small target objects, the feature extraction capability of the network is enhanced by incorporating an attention module to its backbone network. Secondly, the SIoU loss function is used to speed up the convergence speed and improve the regression accuracy, which enhances the network accuracy. Finally, the trained model is compared and evaluated with the original model. The test results show that the improved YOLO V5 target detection algorithm improves 4.58% and 3.40% in accuracy and recall, respectively, compared with the original algorithm, which can effectively avoid the phenomena of missed detection, false detection and overlap, and has significant advantages over existing methods and is more suitable for the detection needs in practical production applications.