Enhancing small parcel sorting accuracy: Robot machine vision in stacking target image experiment

As robots are increasingly being used in various fields of production and daily life, research in this area has become a hot topic among researchers. In the chemical industry, the use of robots instead of manual labor in hazardous environments is a viable solution to reduce the risk of accidents. However, there are challenges in terms of low efficiency, poor accuracy, and inaccurate positioning in robot sorting operations. To overcome these challenges, a method based on visual information perception has been proposed. A prototype of a sorting robot verification experiment platform has been developed, which accurately completes sorting actions through parcel graphic recognition and positioning. The robot parcel sorting experiment has been conducted, yielding promising results. The adaptive recognition rate of miscellaneous feature package images is 93.34%, with an average processing time of 0.12 seconds per image. The successful parcel sorting rate reached 91.03%, with an average sorting time of 1.12 seconds per parcel. The average recognition accuracy of Im-AlexNet, LeNet, AlexNet, and VGG16 models were compared, and Im-AlexNet had higher recognition accuracy. These results demonstrate the successful implementation of robot sorting test in hazardous environments decreases the risk of accidents. The suggested approach, which relies on visual information perception, has demonstrated its effectiveness in enhancing the efficiency, accuracy, and positioning of robot sorting tasks.