| Nowadays,traffic monitoring has been widely deployed in modern transportation systems for vehicle information collection,traffic command,road monitoring,accident detection,and so on.The use effect of the monitoring system is closely related to the quality of the acquired image.Due to the complex outdoor environment,the images obtained by the monitoring equipment are often polluted.Therefore,it is necessary to enhance the images obtained by the monitoring equipment by using image enhancement technology.This paper focuses on the enhancement algorithm of the monitoring image obtained in fog and low illumination and completes the simulation verification.The main work and innovations of this paper are as follows:(1)When facing the fog traffic monitoring image,an adaptive guidance filter is proposed to deal with the fog image with different visibility,and the improved total variational Retinex algorithm is used to optimize the illumination component to avoid the halo effect.The fog concentration distribution of different images is different,which leads to the problem of incomplete fog removal or excessive enhancement in the process of fog removal by using the algorithm of fixed scale filter.Aiming at this problem,an adaptive guidance filter is designed,and the peak signal-to-noise ratio is used as the optimal criterion to construct an adaptive function,which solves the problem of incomplete fog removal caused by inaccurate estimation of illumination component,and the problem of halo effect at the sudden change of depth of field,The illumination component optimization algorithm based on structure and texture perception is designed.Through the iterative optimization of the illumination component,the halo effect is avoided.(2)For the low illumination traffic monitoring image,a dual-channel a priori algorithm is designed to improve the brightness of the image,and the noise is suppressed through image decomposition and texture compensation.At present,the mainstream algorithm is prone to the excessive enhancement of brightness distortion after improving the brightness of the image.This paper studies the characteristics of traffic monitoring images under low illumination,analyzes the causes of brightness distortion in the current mainstream algorithm,designs a dual-channel a priori algorithm,weights and fuses the transmittance images of the dark channel and bright channel,optimizes the calculation method of transmittance,and solves the problem of brightness distortion.At the same time,aiming at the difficulty of noise suppression in low illumination image enhancement,an image decomposition,and texture compensation algorithm is designed.The image is decomposed into structural layer and texture layer,the brightness of the structural layer is improved,the texture of the image is estimated twice to eliminate noise interference,and the texture of the structural layer after brightness improvement is compensated.The experimental results show that the method proposed in this paper can effectively improve the brightness of traffic monitoring images under low illumination,And can suppress the noise.(3)In this paper,a traffic monitoring image classification method based on Resnet is proposed,and the two image enhancement algorithms proposed in this paper are applied to traffic flow detection.At present,there are many types of traffic monitoring image pollution and a lack of detection and classification process.On the basis of residual network,using the data set constructed in this paper and combined with migration learning,a traffic monitoring image classification method based on residual network is designed.The accuracy of this method for traffic monitoring image classification is verified by experimental simulation.The real polluted traffic surveillance video is used for preprocessing combined with the two algorithms proposed in this paper.Then the background difference method is used to detect the vehicle,which verifies the effectiveness of the proposed method.To sum up,the two image enhancement algorithms designed in this paper are used in vehicle detection,which improves the accuracy of vehicle detection,which reflects that the two algorithms proposed in this paper have high theoretical significance and application value. |
