| Haze reduces the visibility of the visual system and the image contrast,which seriously affects traffic monitoring,so the research of image hazing methods has important theoretical and application value for traffic monitoring systems.The difficulty with current image hazing algorithms is how to make effective judgments about haze areas by effective means,while ensuring that the effect meets realistic requirements after defogging,and improves the clarity and contrast of the visual system or image,and restores a clear and realistic scene.The research in this paper is as follows:(1)The current dark channel a priori dehazng method does not take into account the influence of the difference in haze concentration in different regions.This paper introduces the relationship between sharpness index and depth information,This paper classifies the haze concentration in different regions of the image,This paper analyses the relationship between depth and transmittance to determine the refined transmittance,and proposes an improved algorithm for image dehazing based on the dark channel a priori.Finally,the haze removal effect of the improved algorithm is analysed by combining the image quality evaluation index with the outdoor real scene dataset for the haze removal process.(2)In this paper,we apply deep learning to image dehazing and propose a dehazing algorithm that fuses attention mechanisms with multi-scale features.The feature extraction network of the algorithm introduces an attention mechanism to fuse the contextual information of the shallow and deep features.We hope that the algorithm can make the network more focused on the dense fog region.To address the problem of dark luminance after network dehazing,We introduced an enhancement module for the deep network to reconstruct images at different scales,We solves the problem of brightness and low contrast and improving the robustness of the dehazing method to outdoor scenes.We combines image dehazing technology with vehicle detection,A traffic monitoring system is designed to address the problems of low recognition rate of traffic images in haze weather,The weather results in reduced efficiency of the traffic monitoring system.The system will first judge the image with haze,then send the fogged image to the two innovative image defogging methods in this paper for defogging,and finally output through the vehicle detection module.The image defogging technique can be effectively ensured to play an effective role in the traffic monitoring system.The effectiveness and practicality of the system is verified through model comparison experiments. |
