Research On Railway Video Surveillance Image Enhancement Technology Under Low Visibility

Images of outdoor scenes taken in bad weather such as fog and haze are usually affected by low visibility and poor contrast.The image clarity is low,which seriously affects the effectiveness of computer vision recognition systems and intelligent target detection.It is useful for transportation,aerospace,military Real-time monitoring in other fields brings problems.At the same time,image dehazing is a comprehensive problem involving many fields such as atmospherics,optics,and computer vision.It has great academic value.In recent years,new algorithms have also been proposed.In this thesis,in terms of image defogging,the dark channel prior defogging method proposed by He et al.Was first analyzed.For the problem that it could not meet the real-time performance of railway video surveillance,an appropriate scaling of the dark channel image size was proposed to sacrifice some image fineness The cost of the degree reduces the calculation time for obtaining the transmittance and improves the efficiency of the algorithm.Through Matlab simulation test,the experimental results show that the proposed improved algorithm has certain practical significance in the application of railway video surveillance image dehazing.At the same time,this paper presents a method for removing haze and noise in the frequency domain using efficient multi-scale coherent wavelets.Through attempts to find that in natural pictures,fog is usually distributed in the low-frequency spectrum of its multi-scale wavelet decomposition.Thanks to this decomposition,this paper first based on an open dark channel model(ODCM),refines the transmittance through constraints,and eliminates the smog effect at low frequencies.Considering that there is a coefficient relationship between the low-frequency part and the high-frequency part,the soft threshold is used to reduce noise,and the open dark channel model is used to estimate the transmittance to further enhance the high-frequency adaptive part.Then wavelet reconstruction is performed on the low-frequency part and the relevant high-frequency enhancement part,which can restore the fog-free image well.Experiments prove that the method presented in this paper can retain more texture details and reduce noise effect,and has better performance.Due to the different imaging mechanisms of day and night images,human eye recognition cannot meet the requirements of nighttime automation and continuous recognition of railway fog images.At the same time,the imaging of night fog images is relatively complicated.Most of theexisting image defog methods are designed for day image fog removal.This paper uses the similarities and characteristics of day and night fog images,combined with atmospheric scattering models,to give night fog image models and corresponding Fog algorithm.Using the similarity between the night fog image and the day fog image after inversion,combined with the dark channel prior algorithm for preliminary defogging,refine the transmittance through constraints,and then use halo removal and color correction to improve the defogging image vision Results,and the effectiveness of the algorithm is verified by experimental simulation and data comparison.