In order to solve the problems existing in the process of commodity bag aging image recognition,an im-proved commodity bag aging image recognition technology based on YOLOV5 was proposed,and the FasterNet lightweight network was introduced to form the C3-Faster module,which further reduced the number of redundant computation and memory access,and reduced the model size.The experimental results showed that the absorption peaks gradually increased with the prolongation of irradiation time,and the content of functional groups gradually in-creased after aging.And when the number of iterations was greater than 200,the missed detection rate showed a downward trend.The leakage rate of Yolov5 model tended to converge after 800 iterations,the leakage rate of the improved Yolov5 model tended to converge after 600 iterations,and the average leakage rate was about 2.66%,which was much lower than that of Yolov5 model.The detection time and detection accuracy of unmanned supermar-ket packaging bags were better,with a detection time of 25 ms,which was 30.55% and 51.92% lower than bag cracking and discoloration,respectively.
维普
万方数据