Traffic Light Recognition For Intelligent Vehicle

As the urban roads is increasing, the traffic signal and the road sign informationis more and more, more and more multifarious that makes the vehicle environmentcomplicated. If they want to make intelligent vehicles safe in urban road environment,they have to abide by the complexity of the traffic signal. Traffic lights is the one ofimportant information in the traffic signal indication, also is a typical feature of theurban roads, to identify the traffic lights in advance for intelligent vehicle route planahead, and may be according to the specified destination path choice of drivingdriveway or parking or go straight choice. So the traffic lights recognition technologyhas the important significance and good application promotion prospects. This papermainly studies the content summarized as follows:Analyzing the standard of the traffic lights can solve the lights for the successof the localization testing to identify provides prior knowledge of support, in additionto the present study situation analysis, leads to the key technology research problems.Through the color space based on threshold segmentation and morphologicaloperation of traffic lights processing, comparing the different color space modelfeatures and application environment, the selection of the HSV color space modelcolor segmentation. Through to the different light and shade of light color signaldegree unit weight statistics, statistical curves from gain has statistical properties ofcolor component threshold, reduce the color segmentation introduced by the noise.Using the expansion of the corrosion and morphological processing methodeffectively will be scattered noise removal, the use of regional labeling generates asteady reliable region of interest. AdaBoost algorithm is put forward based on the rapid traffic light detectionand based on prior knowledge of traffic lights identification method, using verticallights and horizontal type lights classifier for the interested region within the targetdetection and localization, shorten the target detection search time. Of vehicle lightsand direction for the light signal unit area of statistics, according to the statistical dataset the normalized shine unit area, the division of the threshold, combined with theposition of the light prior knowledge get the instructions of the direction type, finallycomplete the type of traffic lights identification. Through the traffic lights recognitionexperiment in the crossing that has a good experimental results.Tracking method of traffic lights is put forward based on the wave doorSettings and CAMSHIFT algorithm. Through the multi frame after identification ofthe existence of the traffic lights confirmation that can set the wave of testing orcolour gate component calibration, complete the light track. In addition because of thetraffic light color component has good statistical characteristics, using CAMSHIFTalgorithm looking for light color component of the biggest statistical distribution toachieve robust tracking. The introduction of the signal tracking algorithm greatlyshorten the recognition of the algorithm processing time and enhance the systemrobustness.