Railway Recognition Based On Prior Knowledge

In order to accelerate development of the railways and level ofinformation, the ministry of railways had made the "fifteen" plan ofrailways information, set the aim and the requests of the railwaysconstruction. Information system is a management which aim at thoseprocess such as producing, managing, transferring…Three applicationfields of the railways information are transferring organization,transferring managing and operation managing. The new transfer lineput the disaster prevention system into the design request, the disasterprevention system is the main support system to the transferringmanaging, not the assistant equipment. Recently, the natural disasterbring more impact on the railways security, so the ministry of railwaysset a project: Galaxy Disaster auto-Prevention System for Railways in2005 to monitor the security of railways.Recent years, image recognition technology had got a greatimprovement in traffic system, but it is still a gap in railway monitoring.Now, the application of the image recognition makes more contributionin traffic system, such as license plate number recognition, drivingassistant system…All of this play the key role on intelligent trafficsystem, and provide theoretical support for auto-monitor system inhigh-dangerous roads.In this paper, we proposed a railways recognition technique basedon pre-testing knowledge, the crucial parts are the model building andmodel matching. The model building is contained by computermorphology field, here we use the curve fitting method to build themodel, and get the locations of the railways in present image by modelmatching, to avoid the error of the camera vibrating. This method usesmany image processing methods to recognize the railways efficiently inmany weather conditions, and be applied in GPRS-Rv1.0. The main usage of this block is static monitoring in many stable conditions, andto provide analysis-basis for the entire system. It plays an importantrole in alarm resetting, background rebuilding and system resetting.In this paper, we also proposed a edge based block detecting andrecognition technique. This method uses the edge information to solvethe block detecting and recognition. Finally, we use the grayinformation to deal with the error alarm and miss alarm.At last, we carried out the test on samples and test the performanceof the method. The experiments show the advantage of our approach.