High-resolution Image Acquisition And Processing Based On GigE Camera

At present,China’s railway transportation industry is developing rapidly.In order to meet the requirements of social progress and national economic development,the railway network expands the scale and improves the structure continuously,making the total mileage of railway business increases rapidly year by year.As one of the important ground infrastructure in the railway system,the rail fastener undertakes the important task of fixing the rail,ensuring the stability and reliability of the track.However,due to the adverse weather and complex external environment,abnormal phenomena such as missing fasteners,loose fasteners and broken fasteners occur easily,which affect the stability and reliability of the track.Furthermore,the abnormal fasteners described above will affect the efficiency of railway transportation and even endanger the life safety of passengers.Therefore,it’s important for railway department to ensure the high reliability of the rail fastener.Under the current situation,the traditional operation and maintenance mode have not adapted to demand due to poor detection accuracy and low efficiency.With the development of computer and image processing technology,machine vision system has also attracted the railway industry.In addition,there is a higher requirement for the quality and transmission rate of the collected images.Therefore,the detection of the rail fastener no longer relies on manual and inspection train.Instead,it is gradually systematic and intelligent,and is committed to realizing real-time image acquisition and processing on the way.In order to realize the above objectives,this dissertation studies the image acquisition and processing of rail fastener.The main work of this dissertation can be summarized as follows:(1)In the work of image acquisition,in order to meet the requirements of high-speed and large throughput,the Gig E camera and its interface are selected to obtain images of this dissertation.In addition,the interface program that controls the Gig E camera is designed in the Linux operating system.the program completes the configuration of parameters belonging to the Gig E camera,the functions of acquisition and network transmission,and finally implements the storage and display of the fastener image in the host computer.(2)In the work of image processing,a standard data set about rail fastener is constructed.In addition,according to the actual application scenario of the dissertation,under the condition of a certain operation speed,it is proposed to use YOLOv4-Tiny network as the detection model of rail fastener,and the data set is used to complete the training and testing of several different target detection networks.By evaluating various indicators,the applicability of YOLOv4-Tiny network structure of this dissertation is verified.the Tensor RT is used to realize the optimal deployment of the model on the embedded platform,and the detection accuracy has reached 99.01%.(3)Finally,the HMI(Human Machine Interface)of the system is designed based on Py Qt5.The functions of the software are tested with real data,and the results show that the software meets the requirements of actual application.Figure 48,Table 16,References 68...