Research On Traffic Sign Recognition Based On Machine Learning

With the rapid development of social economy,the continuous improvement of science and technology,the traffic environment becomes more and more complex,it is expected to be able to appear an intelligent visual auxiliary navigation equipment,can provide the environment traffic signs information,regulate the operation of the driver,or the auxiliary control of the vehicle to ensure the safety of traffic.However,the actual situation in the scene is very complex,the researchers after years of effort,has not yet made the identification system to the practical application of the degree,still need further research and improvement.Based on the study of road traffic sign recognition system,this paper focuses on the correctness and efficiency of recognition.The main contents of this paper are as follows:The road traffic signs in the actual scene to detect.Based on the histogram of Oriented Gradient(HOG)feature and the support vector machine(SVM)classifier,First to detect the image of the traffic signs unique shape of the rough detection,And then for traffic signs unique color components for fine detection,To achieve a higher detection results,excluding false detection,Save the coordinates of the target area for the next recognition task to proceed smoothly.Traffic Mark Recognition Method Based on HOG and SVM.First,the training samples are obtained by a series of synthetic methods,The HOG feature of the extracted sample library is then fed into the SVM for training to obtain a one-to-many classifier,Using the cross-validation method to continuously adjust the parameters,The correct rate can reach more than 80%,To achieve a variety of traffic signs fast and accurate judgments.Traffic Recognition Method Based on Depth Learning.The main reason why the method of deep learning is widely recognized is that the model can autonomously learn from the training samples the deep-seated features within the image,especially for many who do not know how to design the feature extractor,such as expression recognition,gesture Estimate and so on.In this paper,the open-source Caffe framework uses the depth convolution neural network model to train the traffic sign sample library,so that the machine can learn the characteristics of self-learning,so as to better improve the accuracy of recognition.The correctness of the test set is 96 %,which is much better than the combination of HOG and SVM.