Abstract:

A new method of arrow traffic light detection and recognition is proposed for the complicated traffic conditions of intersections.Intensity segmentation and morphological filtering are used to locate the traffic lights,then combined with the HIS color space and the relative position of the traffic light and the lamp board to confirm the color of the traffic lights.2D-Gabor Wavelets are used to extract the features of the candidate regions of traffic lights.PCA is used to compress the images,and ICA to further extract the feature of PCA.Finally template matching is used for classification.Experimental results show that the correct recognition rate of this method reaches 96.7%.

Key words: traffic lights recognition;D Gabor wavelet transform;PCA;ICA