Research And Application Of Detection And Recognition Method Of Road Traffic Signs Based On Video

With the rapid development of the national economy,family cars are becoming more and more popular,which puts forward more challenging requirements for preventing traffic accidents and solving traffic jams.In this context,the intelligent transportation system as a new scientific and technological innovation has attracted everyone’s attention.Video-based road traffic sign recognition,as an important part of the intelligent traffic system,can improve driving safety and reduce the occurrence of traffic accidents.It is an important part of intelligent traffic research.For complex natural scenes,the immaturity of recognition technology,how to accurately and quickly identify the traffic sign images in the scene is a research hotspot of intelligent traffic technology,and has certain research and application value.This paper mainly studies the preprocessing of traffic sign images,the detection of traffic sign images and the identification of traffic sign images in complex scenes.The research contents are as follows:1.Image Preprocessing: In order to improve the detection and recognition speed,this paper first uses the downsampling method to sample the real-time image of the collected traffic signs under the premise of ensuring the integrity of the sign information,which reduces the image resolution and reduces the processing.The amount of data,while the image is normalized to facilitate the subsequent detection and identification;Secondly,in order to solve the problem of insufficient contrast of the traffic signs obtained,based on the analysis and analysis of commonly used image enhancement algorithms,the use of histogram equalization The method of brightness enhancement of the image processing;Finally,the denoising method of the image is studied.Through contrast experiments,the median filter algorithm is finally used to denoise the real-time image of the traffic sign,which effectively reduces the noise of the image.2.Image detection: In order to improve the detection efficiency of traffic signs,research and analysis of traffic sign detection algorithms based on color models and traffic signdetection algorithms based on contour shapes were conducted.Based on this,a fusion was proposed.Color and shape features based on the HSV color model and corner point shape traffic sign detection method.The principle of this method is to first convert the acquired RGB color model image into an HSV color model representation,perform coarse detection of color features in the HSV color space,and then establish a shape corner point mask according to the traffic sign shape feature and use the mask to perform the roll The product operation determines the corner points,and finally uses the corner points for shape detection and segmentation.The experimental results show that this method effectively improves the detection efficiency.3.In order to improve the recognition rate and recognition speed of traffic signs,traditional template matching recognition algorithms are studied and experimented.Based on the analysis of experimental results,a template recognition algorithm based on Pearson correlation coefficient is proposed.In this method,Pearson correlation coefficient matching recognition is performed by extracting the grayscale distributions in the four directions of horizontal,vertical,45°,and 135° of the standard image and the image to be recognized.Experiments show that this method has higher accuracy and faster speed than the traditional de-correlation coefficient matching method.4.On the basis of analyzing the characteristics of the SIFT feature recognition algorithm,the SURF feature recognition algorithm of the optimization algorithm is studied emphatically.This algorithm generates all interest feature points by constructing the Hessian matrix,and then in the constructed scale space,the Hessian matrix.The key feature points are located,and feature point descriptors are generated for the main points of the feature points.Finally,the SURF feature matching is used to complete the recognition.Based on the algorithm principle,a traffic sign recognition algorithm is programmed.The algorithm first performs rough classification on traffic signs detected by segmentation,and then carries out SURF feature extraction and description on traffic signs.Finally,the rough classification results are matched with feature vector sets of template bases to perform feature point matching to complete recognition.The experimental results show that the recognition of traffic signsbased on SURF feature recognition is better.5.The BP neural network recognition algorithm is studied.According to the shortcomings of long training time and slow convergence speed of BP neural network algorithm,an improved BP neural network recognition algorithm is proposed.The algorithm firstly adjusts the learning rate by an adaptive method instead of a manual method.At the same time,it increases the setting of the momentum item and compresses the range of the S function.Secondly,it classifies and refines the sample library.Experiments show that the improved algorithm accelerates the convergence speed of neural network recognition algorithm,reduces the training time,and significantly improves the recognition rate and real-time performance of traffic signs.6.Based on the above research results,using Microsoft Visual Studio 2010 combined with OpenCV open source visual library designed and implemented a traffic sign recognition prototype system based on MFC framework.The system mainly integrates and implements functions such as video reading,image preprocessing,logo detection,logo identification,and other functional modules.