Traffic Sign Recognition Based On Compressed Sensing

Traffic sign recognition is currently a hot topic in the research of driverless and automotive assistance systems.The traffic sign recognition can be used by drivers or unmanned vehicles to provide real-time and accurate information on traffic conditions and traffic rules,traffic signs also help drivers to make decisions.With these information,traffic safety factors can be improved,and traffic accidents can be reduced or avoided.Traffic sign recognition which is based on compressive sensing has the advantages of high recognition rate,strong robustness,etc.It has also attracts more and more attention in recent years.First reprocess the images in the GTSRB.After the image is grayed,we use the bilinear interpolation method to normalize the dimensions and use the Contrast Limited adaptive histgram equalization to enhance the image.Then the original image is sparsely transformed by the discrete wavelet transform(DWT).The Gaussian random matrix is used as the measurement matrix,the orthogonal matching pursuit(OMP)algorithm is used to reconstruct the image.By comparing the reconstruction effects of traffic sign images under different sampling rates,the best one is selected,the original data can be recovered well with a little data.Finally,the sparse representation based algorithm is used to identify traffic signs,The experimental results show that the orthogonal matching pursuit algorithm has a higher recognition rate and stronger robustness,but the recognition rate and recognition time need to be further improved.For this reason,a sparse representation algorithm based on two phases is implemented,this algorithm can obtain good experimental results when the training sample is small,But when training sample increases,the computer will have the problem of memory overflow.In order to solve this problem,a two-stage sparse representation algorithm based on local redundant dictionary is designed and implemented,the redundant dictionary is divided into many local dictionaries,and the categories of the training samples are separated into each local redundant dictionary,also the time complexity of this algorithm is decreased by this paper.A two-stage sparse representation algorithm based on local dictionary is used to test theGerman Traffic Sign Recognition Benchmark.The recognition rate of the two-stage sparse representation algorithm based on local dictionary is 94.61%,and the average recognition time of each training sample is 0.989 seconds.The recognition rate is 4.42% higher than that of the orthogonal matching pursuit algorithm,and the average recognition time is shortened by 1.23 seconds.The experimental results show the effectiveness of this algorithm.