Robot Vision Recognition System Based on Improved YOLOv3 Algorithm

With the rapid development of the logistics business, a visual recognition system is being developed to speed up the transportation, packaging, and categorization of goods. This system can enhance identification speed and accuracy. First, the experimental platform was set up, and the experimentwas performed on a six-axis robot with a camera lens, light source, and acquisition card. The receptive field module was then used to construct an updated YOLOv3 model based on the original YOLOv3 model to enhance the in-depth features and increase the feature representation ability. Experiments were conducted on VOC data sets. The mAP values of twelve different types of objects, the visual recognition effects of five different groups of pictures, and the visual recognition values of the original and revised models were compared. The results showed that the improved YOLOv3 model outperformed the original model in terms ofmPA, visual recognition image, and recognition speed.