Research On Image Quality Improvement Based On Line-scan CCD

In recent years,line-scan CCD have been widely used in remote sensing imaging,medical imaging,industrial detection,high-speed contactless measurement and other fields.The premise of these practical applications is that the quality of linear array image can meet the application standard,but the distortion of linear array image,uneven brightness and insufficient brightness will affect the quality of linear array image,which is not conducive to the practical application.In practical application,it is an important research topic to control the quality error of linear array image within an acceptable range.In this paper,line-scan CCD and linear array image are studied,including advantages and disadvantages of line-scan CCD imaging process of linear array image,characteristics of linear array image and so on.At the same time,the existing problems of linear array image quality are pointed out.Firstly,the application status of line-scan CCD is summarized.In industrial applications,line-scan CCD has significant advantages compared with planar array camera.Then,the imaging model of the line-scan CCD is analyzed,and the line frequency of the line-scan CCD changes with the speed of the freight train.If the line frequency of the line-array camera does not match the speed of the freight train,the line array image will be distorted.Secondly,the line frequency of line-scan CCD is inversely proportional to the exposure time,and the exposure time is linear with the brightness of the image.The constant change of line frequency will cause the uneven brightness of the line array image.At the same time,the grayscale value of linear array image is generally low due to the influence of the fast running speed of freight train,high line frequency of line-scan CCD,short exposure time of line-scan CCD and other factors.Finally,the linear array image has the characteristic of significant difference between relative stationary and relative moving regions,which is beneficial to segment the linear array image and remove the redundant information in the linear array image.Aiming at the problem that the linear array image quality is not high in the intelligent detection of freight vehicles in the field of rail transit,this paper proposes a concrete method to improve the linear array image quality.1.In order to evaluate the quality of linear array image quickly,a rapid evaluation system of linear array image is constructed.Based on the general evaluation index of image quality,this paper proposes the specific evaluation index of linear array image combining with the characteristics of linear array image.Among them,the general evaluation indexes of image quality include image brightness,image brightness uniformity,image sharpness and structural similarity.The specific evaluation index of linear array image is the distortion degree of linear array image.The rapid evaluation system of line array image quality is constructed according to the above quantitative evaluation indexes,which can realize the rapid evaluation of line array image and meet the demand of real-time evaluation of line array image quality in industry.2.Aiming at the distortion problem of linear array image,a distortion correction method based on velocity prediction is proposed.By predicting the freight train speed and adjusting the line frequency of the line-scan CCD in advance,the line frequency can be matched with the speed,the aspect ratio error of the line array image can be reduced,and the distortion degree of the line array image can be reduced.The linear speed compensation method assumes that the train moves at uniform speed,while the second-order exponential smoothing method considers the time dimension trend of freight speed.On this basis,the Nonparametric Prediction Method method is proposed.By establishing the historical sample database,using Fast DTW speed data similarity metric,using the current speed and speed as state vector of the first three time,matching samples and the influence factors of structural consistency highest data in the database is the most similar to the speed of data,using the weighted assignment method to estimate the speed of the train the next moment.The linear velocity compensation method,the second-order exponential smoothing method and the non-parametric regression method are compared and analyzed.The non-parametric regression method has less prediction error and stronger robustness.Especially,when the train speed varies widely,the mean square mean error predicted by the non-parametric regression method is reduced by 51% compared with the linear speed compensation method,and the mean square mean error predicted by the proposed method is reduced by 56% compared with the second-order exponential smoothing method.The results of sports car experiment and field data collection in freight station show that the distortion correction method based on velocity prediction can reduce the aspect ratio error of line array image,and the absolute value of the average aspect ratio error of line array image is less than 3%.3.Aiming at the problem of low brightness and uneven brightness distortion in linear array images,a brightness enhancement method based on multi-scale Retinex theory was proposed.The redundant information in linear array image will affect the result of brightness enhancement algorithm.Therefore,it is necessary to remove the redundant information in linear array image before image brightness enhancement,which can reduce the spatial complexity of subsequent linear array image brightness enhancement algorithm and reduce the interference of redundant information to the result.According to the significant difference between relative static and relative motion of line array images,a line-array image segmentation algorithm based on regional features is proposed.Compared with frame difference method,the average time consuming of line array image segmentation algorithm based on region feature is reduced by 2.689 ms,and the probability of correct segmentation increased by 7.78%.Based on the MSR algorithm,an MSR optimization algorithm is proposed to enhance the brightness of linear array images.This method optimizes MSR algorithm from two aspects.On the one hand,the use of bilateral filtering as the center surround function can effectively improve the MSR algorithm prone to halo,overenhancement and other problems.On the other hand,the adaptive filter can be improved by using the change of gray mean to estimate the noise difference.The linear array images with uneven brightness and insufficient brightness were analyzed empirically,and 100 linear array images with uneven brightness and low brightness were screened by using the rapid evaluation system of linear array image quality.The average brightness uniformity of the image was improved by 25.6% and the average brightness of the image was improved by 19.1%.In this paper,the image brightness enhancement method can effectively improve the low brightness and uneven linear array image.In this paper,the existing quality problems of linear array images are analyzed,and the improvement methods for different problems are put forward.Experimental results show that the proposed method can correct distorted images and enhance the brightness of linear array images.