Detection And Recognition Algorithm Of Traffic Signs Based On The Characteristics Of HOG

Traffic sign detection and recognition is an important part of the intelligent vehicle system, which is also one of the effective ways to solve the traffic safety,In reality, due to the low resolution, different climatic conditions and illumination intensity, Motion blur and so on,Traffic image quality is very poor, which is difficult to be detected, So these factors make the traffic sign detection and recognition is still an urgent subject in this fieldTraffic sign detection and recognition mainly includes two parts: how to find the traffic signs accurately in the image and how to extract the traffic sign image to achieve accurate identification or classification. This article from the two direction :one is traffic sign detection and another is traffic sign recognition.In this paper, the algorithm of the two stages were studied and realized respectively, for the traffic sign detection, we use traffic sign detection, which combines based on the RGB color space and Based on the normalized correlation coefficient. Firstly, we need to preprocess the captured image, mainly using the image denoising based on median filtering and histogram equalization based on the R, G, B three channels, It can enhance the quality of image. Secondly, the detection process fuses the color and shape features, extracted twice, the crude extraction is based on color channel of image segmentation, then using morphology to process, highlighting the target of interest region. The fine extraction aims to get the accurate ROI, based on calculating the template image and normalized cross correlation coefficient got by the crude extraction detection of image, the detection process fusion of color and shape features, two times of extraction.For the recognition of traffic signs, this paper extracts the histograms of oriented gradients(HOG) feature of interest region(ROI), then delivered to optimized classifier to classify color and shape features. For the SVM training, the paper is based on the database of GTSDB, and optimize the parameters of the classifier model, to get the final PR curve.To test the algorithm based on MATLAB, the test results show that traffic sign detection and recognition algorithm in this paper has good effect At the end of the paper, we test the real image, the results show that the algorithm has good accuracy, and has good robustness under different real circumstances.