Research On Traffic Image Defogging Method And Applications

The fog leads to low visibility of the road,which seriously affects the traffic monitoring.Therefore,the research of image defogging algorithm has important value in theory and application for traffic video monitoring.According to the requirement of the traffic video monitoring project,this paper focuses on the method of traffic image defogging and application in traffic monitoring,and the simulations are completed.The main research contents as follows:(1)The lack of detecting whether the images contain fog or not in the existing defogging methods,which leads to the introduction of more noise after the no fog images defogging,and it is not conductive to the image post-processing.According to the research of Alexnet network and the transfer knowledge,an image classification althorithm based on Alexnet transfer network to classify and recognize fog traffic images is designed,which can improve the image defogging efficiency.The classification accuracy of the method is verified by simulation experiments.(2)To solve the problem of color distortion in sky or white areas after image defogging,this paper designs an adaptive processing method for sky or white areas.The corresponding adaptive correction function is constructed after segmenting the image,which improves the restoration effect of the region and enhances the contrast of the whole image.In this paper,an improved gradient bilateral filter transmission optimization method is designed to handle the problem that the blur details and “noise patches” are easy to produce in the close scene after image defogging.By refining the transmittance,the effect of detail processing in close scene is improved.(3)In order to improve the robustness and adaptability of the image defogging method,this paper proposes an adaptive defogging method based on improved gradient similarity nuclear,which is fused the adaptive processing algorithm for sky or white areas and the improved gradient bilateral filter transmission optimization algorithm.Apply it to the traffic fog images processing,and the simulation results show that the method can effectively improve the defogging quality of traffic fog images,and it has good robustness and reliability.(4)This paper studies an improved SURF traffic image matching method,which can better tackle the low accuracy about traffic images matching results in the decline of traffic monitoring efficiency in fog weather.Firstly,defogging the fog image is served to extract more image features better,and then fast match these feature points.Furthermore,two-way matching is used to match the image accurately,and improved RANSAC algorithm is employed to eliminate mismatching points,which can improve the accuracy of image matching.The simulation results illustrate the effectiveness and practicability of the new SURF method.