Research On Railway Cargo Inspection Image Quality Improvement Technology And Evaluation Method For Intelligent Application

Railway cargo inspection operation is an important part of the safety inspection of wagons and cargo loading status in transit and timely elimination of safety hazards,which is mainly divided into human inspection and machine inspection,human inspection refers to manual on-site inspection,and machine inspection refers to inspection through video,image and other collection equipment.In 2020 China State Railway Group Company Limited issued the "Optimization of cargo inspection workflow implementation plan" to promote cargo inspection operations From "human inspection as the main,machine inspection as a supplement" to "arrival operation machine inspection as the main,human inspection as a supplement,departure operation human inspection as the main,human-machine combination" conversion.At present,the railway intelligent cargo inspection application has been gradually promoted in the national railway,through intelligent identification technology,to a large extent,machine vision instead of human eye observation.However,due to the huge difference in the technical capability of the railway cargo inspection image acquisition system in different stations nationwide,the quality of the formed cargo inspection image varies greatly and the overall quality is not high,which seriously affects the human eye judgment and also causes resistance to the promotion of intelligent identification work,and finally affects the transformation of the railway cargo inspection operation process.In this paper,we study the image acquisition system of railway cargo inspection,analyze the advantages and disadvantages of key component equipment,combine the imaging principle of line array camera,the characteristics of railway cargo inspection images,etc.,focus on improving image quality,start from image acquisition,image processing,image evaluation of the whole process key links,research on equipment selection methods,image quality improvement methods,image quality evaluation methods and other key technologies.The main work is summarized as follows:(1)The research proposes the selection scheme of image acquisition equipment for railway cargo inspection.In view of the backwardness of the current technical standards of the railway cargo inspection image acquisition system,combined with the development needs of intelligent cargo inspection applications,the optimization direction of detection speed,image accuracy,data transmission and equipment reliability is summarized,and the selection scheme of key equipment for image acquisition module,optical module,speed measurement module and data transmission module is established to realize a high combination of hardware equipment and performance requirements and reasonably improve the overall performance of the system.(2)The study proposes an image quality improvement scheme.For the situation that there are differences in the detection of wagon models,the mathematical relationship between the diameter of the tolerance dispersion circle,aperture,focal length,object distance and depth of field is established,and the optimal model of camera parameters of the railway cargo inspection image acquisition system is formed to maximize the depth of field range covering various models and clear imaging;for the problem that the speed measurement scheme of the wheel sensor has insufficient real time,the speed formed by using the speed that has occurred during the train operation The method of using the short-term series to predict the train speed at the next moment can effectively improve the matching between the camera line frequency and the train speed and help reduce the image distortion;the histogram equalization algorithm is used to pre-process the image to achieve a good brightness enhancement effect while causing basically no impact on the existing cargo inspection work.(3)The study proposes a method for evaluating the quality of railway cargo inspection images.In order to establish the system performance evaluation mechanism,a comprehensive evaluation method based on the multi-dimension of sharpness,distortion and brightness is proposed;for the problem of noise sensitivity in the NRSS algorithm for sharpness evaluation,the Canny operator double-threshold algorithm is adopted for edge information extraction to improve the accuracy of the evaluation results of high-noise images;the coordinates of the target object based on manual measurement and deep learning are proposed inductively,and the information extraction method is proposed accordingly.Information extraction methods based on manual measurement and deep learning are proposed,and the distortion evaluation method is proposed accordingly;in order to overcome the problems of grayscale value,standard deviation evaluation value method degree evaluation and normalization difficulties in the brightness evaluation process,the histogram-based brightness information evaluation value method is proposed.This paper analyzes the problems affecting the image quality of railway cargo inspection from multiple dimensions,studies the countermeasures to solve them one by one,and verifies and analyzes the relevant theories from practice,which enriches the research perspective of railway wagon image utilization management and provides support for promoting the intelligence and less personalization of railway cargo inspection operation.