Research On Image Enhancement Method Of Rain And Fog Weather For Intelligent Transportation

Image transmission technology is developing rapidly because of its large amount of information,easy storage,rich content and many other advantages.At present,it has been widely used in all areas of society.The first vision of intelligent traffic real-time monitoring system and automatic navigation system is the most intuitive expression in people's eyes.However,in real life,these systems are often affected by rain and fog weather,causing blurry images under the vision.How to use the algorithm to make the fuzzy image clear is a challenge in the field of intelligent transportation.This thesis focuses on the algorithm research of two weather environments,rainy day and foggy day.The main contents are as follows:(1)In view of image quality reduction caused by raindrops in rainy days,a visual model optimization algorithm is proposed in this thesis.Traditional raindrop removal algorithms usually realize raindrop recognition by a given prior,and then remove.Raindrops are transparent,and their shapes and sizes are ever-changing,so there are certain limitations in raindrop identification.In order to overcome the above defects,the generated adversarial network is organically combined with and the super-resolution network to achieve the effect of removing raindrops in this thesis.Then,proposed algorithm is compared with the Pix2 Pix algorithm and the Attention Gan algorithm by experiments.Experimental results show that the visual model optimization algorithm presented in this thesis has certain advantages in the structure similarity and the peak signal-to-noise ratio for different raindrops in the images.(2)In view of image quality reduction caused by foggy images,a hybrid transmission optimization algorithm is proposed in this thesis by sufficient reliability consideration of the real-time performance of the monitoring system.The dark channel prior algorithm is one of the most popular image defogging algorithms nowadays.However,the real-time performance may not be guaranteed when this algorithm is applied to the outdoor intelligent traffic monitoring system.In order to overcome the above defects,the guided filtering,the median filtering and the color attenuation model are combined to remove fog in this thesis.And the transmission was optimized to achieve the fogging effect.The proposed algorithm is compared with the MSRCR algorithm and the He algorithm by experiments.Experimental results show that the hybrid transmission optimization algorithm proposed in this thesis has certain advantages in the image restoration effect and the algorithm speed for images with different concentrations of haze.