Traffic signal lamp recognition method based on Gabor and sparse representation

Abstract

The invention relates to a traffic signal lamp recognition method based on Gabor and sparse representation, comprising the following steps: (1) carrying out 2D Gabor wavelet filtering on a region of interest, and acquiring a filtered Gabor image as a test sample, wherein the region of interest is a traffic signal lamp region; (2) based on a trained over-complete dictionary, calculating the sparse coefficient of the test sample using an orthogonal matching pursuit algorithm, wherein the over-complete dictionary is trained based on a K-SVD algorithm; and (3) carrying out image reconstruction using the sparse coefficient and the over-complete dictionary to get reconstructed images corresponding to the traffic signal lamp categories, and judging to which traffic signal lamp category the test sample belongs according to the residual error between the test sample and each reconstructed image. Compared with the prior art, traffic signal lamp recognition is carried out based on Gabor and sparse representation, big error caused by arrow signal lamp direction recognition is overcome, and small round and arrow signal lamps in an image can be recognized correctly.

Description

technical field [0001] The invention relates to an image recognition method, in particular to a traffic signal light recognition method based on Gabor and sparse representation. Background technique [0002] The detection and recognition of traffic lights is an important part of intelligent transportation and an important part of assisted driving. For the detection and recognition of traffic lights, researchers at home and abroad have proposed many methods in recent years. [0003] Miao Xiaodong et al. proposed to use the improved multi-scale Log Gabor wavelet to extract multi-resolution features of traffic signs, calculate the phase consistency according to the feature information at different scales, and extract the target phase information that can effectively overcome the influence of exposure. Multi-class support vector machine (SVM) for multi-object classification (reference "Real-time recognition method of traffic signs in complex environments"). [0004] Wang Yucha...

AI Technical Summary

Problems solved by technology

[0008] However, there are still many problems in the field of traffic signal detection and recognition that have not been well resolved, especially in the area of ​​traffic signal recognition. , the recognition accuracy rate is not high; (2) For signal light images collected under different lighting environments (such as strong sunlight during the day and night, etc.), the pixel color is distorted, and the edge of the signal light is unclear. The recognition of circular signal

Images