Research On Traffic Light Identification In The Micro-traffic Environment

Unmanned intelligent vehicles have improved the safety performance driving and the accurate identification is of vital importance. As the high cost and long testing period of researches on traffic light identification, software simulation is the ideal experimenting way. Researches conducted in micro-traffic conditions can not only reflect real results, but also save the experiment cost. This thesis mainly introduces ways of identifying circular traffic lights and conducts researches on identifying arrow traffic lights.First, this paper searches interest areas, detects circular transportation and chooses Gaussian modeling in HIS color space to identify traffic light colors on the basis of detailed analysis of RGB, HSV and Lab color space. After processing the color segmentation of images, this paper adopts HOUGH and circular degree detection respectively to detect the external shape of circular traffic lights. Match traffic light models in accordance with models in database to determine circular traffic lights with a higher identifying rate as 98.8%.In terms of arrow traffic lights, this thesis adapts the CANNY operator to detect external shape of traffic lights after image pre-processing. Outline features of traffic lights can be obtained. Then identify directions of traffic lights on the basis of the improved HU invariant moment. At last, this paper categorizes arrow traffic lights in terms of the Euclidean distance. Experimental results show that this is an effective method to identify arrow traffic lights.