Traffic road sign recognition method based on image saliency and depth learning

Abstract

The invention discloses a traffic road sign recognition method based on image saliency and depth learning. According to the method, a network (Alexnet) is generated on the basis of a convolutional neural network (CNN); traffic road sign pictures are obtained from an ImageNet data set; normalization pre-processing is performed on the training images; and the pre-processed images are inputted to the network; the Alexnet is trained; RC saliency extraction is performed on a test picture, an obtained picture is inputted into the trained Alexnet convolutional neural network; and the test classification of a traffic road sign is performed. According to the traffic road sign recognition method, a salient region can be efficiently extracted based on image saliency, and the advantages of depth learning in picture recognition are utilized in combination, and therefore, the traffic road sign can be accurately classified.

Description

technical field [0001] The invention belongs to the technical field of image recognition, in particular to an image recognition method of traffic road signs based on a convolutional neural network model. Background technique [0002] With the development of the economy, modern transportation has been very developed, while road traffic still occupies a dominant position. However, traffic safety and traffic congestion have become increasingly serious social problems, and have also caused staggering economic losses. Coupled with environmental pollution and energy issues, the solution to road traffic problems has to resort to intelligent technology. The research field of intelligent transportation system came into being and developed rapidly. [0003] Road traffic sign recognition is one of the unresolved problems in the field of intelligent transportation system research, and it is also one of the more difficult real-scene graphics recognition problems. Although the research...

AI Technical Summary

Problems solved by technology

Many theories and methods are proposed based on standard graphs or certain local conditions, without considering the needs of more practical applications

[0005] 2. The processing method is relatively simple, there are few intelligent methods, and the combination of intelligent methods and other methods is seldom

The current processing methods mainly use intelligent methods such as neural networks and genetic algorithms, many of which are improvements or applications of traditional methods, but rarely combine intelligent methods with other methods

[0006] 3. Most of the experimental objects are based on standard images, and there are few studies on real-world images

Images