The Research On Highway Visibility Detection Technology Based On Image Processing

With the rapid development of social economy and the continuous improvement of scientific and technological level,the application of Machine vision in the intelligent era is more and more widely.Intelligent traffic has become an important part of the development of social intelligence.But With industrial development,the haze become a terrible impact in people's daily travel.So the visibility detection method which based on machine vision has become a very popular research topic.In order to improve the accuracy and the real-time of atmospheric visibility testing,a method to detect the highway visibility level based on Gray Level Co-occurrence Matrix(GLCM)-Speeded-Up Robust Features(SURF)Matching features-Bat Particle Swarm Optimization(BAPSO)-Incremental Probabilistic Neural Network(IPNN)is proposed in this thesis.In this paper,it conducts the samples as experimental data which were collected in the same location at different time for two months.In order to improve the real-time performance of visibility detection,a step-by-step approach is adopted: calculate the standard deviation of the grayscale to forecast visibility by choosing the method of calculating the simple grey histogram.If it doesn't satisfy the end condition,the image feature extraction method will be adopted for further visibility detection.Traditional image feature extraction method has poor stability in low visibility conditions and the effect of feature extraction is not good.In order to solve this problem,it combines SURF Matching features with the contrast parameter and energy of GLCM feature parameters to describe the image samples features,and gives the visibility rating label for the new combination feature vectors as the input data for the neural network.Adopted the IPNN to classify the visibility rating.At the same time,to improve the slow convergence and make it easy to get into local optimal state of PSO,the neural network adopted BAPSO to optimize the parameters and enhance the classification accuracy of neural network and reduce the complexity of network structure.The theoretical research and experimental results show that the proposed method based on image processing of GLCM-SURF-BAPSO-IPNN has good robustness in highway visibility detection.It also has the advantages of real-time compared with the traditional transmission method.At the same time,the accuracy of test results is high and meet the requirements of the World Meteorological Organization on atmospheric visibility measurement that the error is lower than 20%.