Research On Stereo Vision Technology Under High Intensity Illumination

With the policy of "Made in China 2025" and the popularization of "Industrial 4.0" technology,binocular stereo vision system works as an important component of computer vision,has attracted more and more attention.Most algorithms in computer vision,such as image segmentation,image restoration,target recognition and so on,do not take into account the influence of illumination in the process of implementation.However,there are various heterogeneous materials in the real world.When the intensity of illumination is high,the targets' surface will appear mirror reflection,resulting in the image texture color information is covered by light,has a great influence on the stereo matching results of later images,so it is necessary to restore the images captured in high intensity illumination environment,which help to highlight the color and texture details of the original image.In this paper,an improved high-light removal method is proposed and applied to binocular stereo vision.The main work is as follows:In view of the problem of large error existing in the mature camera calibration method at present,the corner detection of Zhang's calibration method is optimized by using the sub-pixel corner detection algorithm in the opencv visual library to get a more accurate interior of the stereo camera and external parameter values.Traditional image highlight removal method is easy to cause color distortion or edge loss in image restoration.In view of these shortcomings,an improved method for removing high light which based on bilateral filtering is proposed.Firstly,the image pixels are divided into pixels with diffuse reflection component only and pixels with specular reflection component by threshold value,and the maximum diffuse reflection chromaticity of the two kinds of pixels is estimated respectively.Then,the estimated maximum diffuse reflectance chromaticity similarity is used as the range of bilateral filters,and the maximum chromaticity map of the image is used as the guide map of bilateral filters to preserve edges and denoise,eventually achieve the goal of removing specular component.During the experiment,50 high-gloss images were processed by the improved bilateral filtering high-gloss algorithm,and the algorithm was compared with the Yang's method and the Shen's method respectively.The results showed that the peak signal-to-noise ratio(PSNR)of the algorithm increased by 4.17% and 8.40% on average,respectively.The processing effect of the algorithm was more consistent with human vision,and the obtained image quality was better.Using image pairs with different illumination intensity as processing objects,KAZE registration algorithm with robustness to illumination was selected through experimental comparison.Finally,by quantitative analysis of the experimental results,it is verified that the improved specular removal method proposed in this paper improves the stereo matching accuracy under high-intensity illumination.