Study On Low-illumination Image Enhancement Technology At Night

With the continuous growth of the number of cars and the current safety problems in vehicle driving,unmanned driving and its auxiliary technology have attracted more and more attention in reducing traffic accidents,solving road congestion,and expanding travel modes.It has become an important direction for the future development of automobile intelligence.This paper takes unmanned driving as the starting point to study the lowillumination image enhancement technology of machine vision in night scenes.Unmanned driving technology is currently facing a variety of difficulties and complex factors from its proposal to its implementation,and low-illumination images during night driving are also one of the complex issues that need to be faced.In order to effectively improve the problems of poor image contrast,low brightness and noise at night,this paper analyzes the characteristics of low-illumination images at night and the existing algorithms,proposes and implements related image enhancement algorithms.The main research contents of this article are as follows:1.Analyzed the significance of the subject research,based on the research on the development status at home and abroad,put forward the difficulties and problems in the night low-illumination image enhancement technology,and conducted an experimental analysis on some image enhancement technologies and related methods,and designed the overall research plan.2.An improved adaptive image enhancement algorithm based on gamma correction by adjusting image brightness information is proposed.On the basis of analyzing the theory and shortcomings of the MSR algorithm,by extracting the illuminance component of the image V channel,combining the illuminance component information and designing a parameter-adaptive gamma correction function,the brightness distribution of the image is improved and uniformized.By combining the illuminance component information and designing a parameter-adaptive gamma correction function to achieve the improvement and uniformity of the image brightness distribution,at the same time,the guided filtered image and residual image of the reflected component are weighted and merged,so as to achieve image denoising and edge sharpening,and finally the image is converted from HSV to RGB color space.Combined with the analysis of subjective and objective evaluation methods,it is verified that the algorithm can effectively improve low brightness and suppress high brightness,and retain richer image information.3.An improved low-illumination image enhancement algorithm based on dark channel defogging is proposed.By analyzing the problems of dark channel defogging algorithm in night images such as halo and inaccurate brightness adjustment,a method of obtaining the edge of the dark channel is adopted,and the atmospheric light value is accurately estimated according to the characteristics of the image,combined with the sampling method to improve the processing efficiency,so as to achieve the enhancement of low-illuminance images.After experimental analysis,the algorithm effectively prevents the halo problem,improves the image noise,and the overall image is brighter and more natural.4.A set of software system for image enhancement and result analysis is realized.The algorithm proposed in this article was successfully transplanted to the Qt platform,through Matlab,C++ mixed programming and using the Open CV framework,the development of related functions was completed,and the debugging,simulation and comparative analysis of related algorithms were realized.