Research Of The Rapid Detection Method On Missing Railway Fastener Based On Computer Vision

Railway fastener is important to maintain the railway transport safety and missing fasteners may lead to accidents like train derailment. Currently, the rapid development of the railway, in particular high-speed rail, makes the needs of automated detection of fastener become increasingly prominent. How to use modern technology to achieve rapid and accurate detection of fasteners becomes an important issue. Using computer vision technology automatically detects fastener is commonly used in the current international programs. Computer vision acquires image information from the measurement object and realize the measurement process through image processing, which has the advantage of non-contact, high speed, high precision, large amount of information, highly intelligent, highly adaptive, etc.Based on the current research on the methods about fastener detection, with the project Detection of Missing High-speed Railway Fasteners as background, this essay researched on the rapid detection method of missing fastener based on computer vision. The main content of the essay is the overall design of the detection system, and the algorithm for detecting fasteners.The essay presents the hardware and software design of the detection system. It uses an IPC as kernel hardware part in order to control a high-speed camera to grab images, with a lighting system. Normally, positioning system sends a trigger signal when the fastener locates in the middle of vision field. The camera gets the signal, grabs an image, send the image to IPC. The IPC processes the image. If the fastener is lost, IPC stores the image and location. At the same time, IPC adjust the parameters of poisoning system.The software is also designed in the essay. It is divided into three layers, interface layer, control layer and module layer. The module layer contains several modules. The essay analyzes the demand function and implementation of each layer.The essay focuses on image processing algorithms. Based on existing algorithms, it presents a new algorithm about detecting parallel lines, which is used to detect fasteners and get a good result when the fastener edges are obvious. At last, it uses the oriental field algorithm, which is based on statistical information, to detect fasteners, and finally achieves quickly and efficiently identifying the target fastener even using a low quality image.The essay also researches on implementing the oriental field algorithm on GPU, including selecting GPU memories and dividing blocks and threads of GPU kernel functions.At last, the essay tests correctness of the algorithm and several speed of the system, and validate the detection system to work effectively.