Traffic Sign Recognition Based On Vehicle Vision

Traffic sign recognition (TSR) is one of the most important part of intelligent driverassistance system. It is based on pattern recognition, image processing, artificialintelligence and computer vision. With the rapid development of China’s automobileindustry, traffic safety problems have become increasingly prominent. The application ofintelligent transport plays an important role in protection of personnel and property security.Traffic sign recognition system is based on vehicle vision and plays an important rolein the intelligent transportation. The vast majority of published traffic sign recognitionapproaches utilize at least two steps, one aiming at detection, the other one at classification,that is, the task of mapping the detected sign image into its semantic category.Regarding the detection problem, Threshold in HSV color space, as an efficient way, isused to find the areas of interest which may contains a road sign, because HSV color spaceis similarity to human perception of colors and this model is more robust against lightingconditions variations. After an analysis of the connected component some candidate blocksare pre-segmentated. Then, we use a Viola-Jones-type detector based on Haar-like featuresand the improved Haar-like feature to identify the content of the extracted traffic signs. Forthe classification task, the improved convolutional networks are applied in this paper, whichare biologically inspired multi-stage architectures that automatically learn hierarchies ofinvariant features.In this paper, The AdaBoost algorithm and convolutional networks are used to solvethe problem of traffic sign recognition, and experimented on the Matlab platform. Finally,experimental results show that the method in this paper are effective regarding the varietyof driving and weather conditions.