Evaluation on Vision Intelligent Control and Image Target Location and Recognition Based on Deep Learning

Abstract

Abstract Vision is the most important way for human beings to obtain information. Under the constant evolution of electronic imaging technology, visual images are extensively applied to the production and life of people. The analysis of visual image information can achieve intelligent control and complete specific tasks in industrial production. For example, in the logistics parcel sorting, the traditional manual parcel sorting is slow, inefficient and costly. For the logistics parcel sorting system, the machine vision was used to obtain the parcel image information, and the depth learning algorithm was used to locate and recognize the parcel image. In this paper, the depth confidence network algorithm and the convolution neural network algorithm were compared in image positioning and recognition experiments. After several groups of iterative experiments, the results showed that in large package images, the average image recognition accuracy of the depth confidence network algorithm and the convolution neural network algorithm was 94.42% and 96.09% respectively. In the small package image, the average image recognition accuracy of the depth confidence network algorithm and the convolution neural network algorithm were 96.53% and 97.64%, respectively. Therefore, applying convolution neural network to the object recognition of logistics package image can effectively improve the accuracy of image recognition and improve the efficiency of logistics package sorting.