A Method of Traffic Lights Detection Based on Visual Selective Attention

In this paper, the candidate target traffic lights under different backgrounds were automatically detected with the visual selective attention mechanism. The region of interest in color image was projected into the gray, and the gray image enhancement was done through top-hat algorithm. The binary image was obtained by using Otsu's method. And finally the logical 'and' operation was carried out for the binary image and the binary image which was obtained according to threshold segmentation in HSI space. The detection on depth information feature of the lights surrounded by the dark boundary was applied to the interested target candidate region. Experimental results show that this method was robust.