Research On Grid Extraction Algorithm Based On Railway Guardrail Image

With the rapid development of information society and the popularity of safetyconsciousness,Each countries has placed great importance on the detection and maintenance ofhigh speed railway infrastructure, a lot of infrastructure detection and repair equipment hasdeveloped, but all the research to the rail system detection device were focuses on the train andtrack itself. Railway line environmental monitoring is one of the important parts of theinfrastructure detection, it includes the detection about the guardrail, catenary, foreign bodiesintrusion etc. Because of complex natural environment and huge amount of data, rail barrierintegrity detection is still in a blank of study so far. This paper is a research for grid extractionalgorithm based on the railway guardrail image, aimed at strengthening the daily maintenance ofthe rail fence netting in the safe operation of the train, provide technical support for the integrityjudgment of rail fence netting. This paper studies from the simple threshold segmentationalgorithm firstly; then research the effect of the Mean Shift algorithm for railway guardrail imagesmoothing and segmentation; and then in order to fix the over segmentation of the image, basedon the area weighted regional similarity criteria and the minimum area threshold merger stoppingcriteria for region merging were adopted; finally extract the grid of the rail fence netting.According to this series content, this thesis expounds from the following four aspects in turn.1. In order to take the processing speed of the algorithm into account, This paper studiesfrom the simple threshold segmentation algorithm firstly, discussed the different segmentationeffect of the rail fence netting by several classical threshold segmentation method Such asbetween-cluster variance method (Otsu method), the maximum entropy threshold segmentationmethod, local threshold segmentation Niblack method. Although the result was ineffective, thecolor range of the grid area, and the direction for follow-up study were got.2.In the study of the classical Mean Shift image filtering algorithm, found that the MeanShift algorithm not only filter and smooth out the small differences in image but also can protectand highlight the grid area, and has precedent in image segmentation. This paper use Mean Shiftalgorithm in domain of space and chromaticity to segment the rail fence image, and through theexperiment got the best space bandwidth and chrominance bandwidth.3. After rail fence image segmentation by fixed bandwidth Mean Shift algorithm, and then inorder to fix the over segmentation of the image, based on the area weighted regional similaritycriteria and the minimum area threshold stopping criteria for region merging were adopted. Andin the area after the region merging, use chromaticity value range of the grid area to get the earlyextraction binary image. 4. After binary image analysis to geometric features, through straight line detection torealize grid area location. And remove the non grid lines and complementary grid line. The resultwas rail fence grid image...