Research Of Modern Tram Railway Limit Recognition And Real-time Dynamic Tracing System

Modern tram is a new rail transit mode based on the traditional tram after the overall technical transformation and the management model innovation, which has important significance for developing the urban public transportation system and relieving urban congestion. At present, the tram mainly relies on human vision to analyze the front road status. But drivers’ vision has a limited range. The judgment on the surrounding is influenced by the people themselves and is not stable and accurate. The vehicular real-time monitoring system based on machine vision can assist drivers to collect and analyze the information of the road ahead so as to ensure the safe operation. Railway limit recognition is the basis for the monitoring system. Railway limit can effectively identify the region of interest, reduce the searching range, shorten the processing time and improve the recognition accuracy. In this paper, we have researched the theory of modern tram railway limit recognition and real-time dynamic tracing system. The study mainly includes the following contents:1. Track region segmentation. The advantages and disadvantages of the commonly used traditional methods based on threshold segmentation or edge detection are firstly analyzed. According to the actual characteristics of the tram track, track region segmentation based on multiple thresholds is proposed. The threshold parameters are selected adaptively based on local accumulation histogram.2. Railway linear recognition. On the basis of binary image, according to the general rules of extrapolation, combined with related theories of region growing and skeleton extraction, track feature points extraction based on pixel tracing is proposed. Analyzing and comparing the advantages and disadvantages of some different fitting methods, we decide to use least square piecewise polynomial fitting method to establish the equation.3. The accuracy evaluation index for railway recognition. According to the analysis of the whole process of railway recognition method, we innovatively connect the accuracy evaluation index for railway recognition with time series similarity metric. Euclidean distance and dynamic time warping(DTW) distance are taken as the accuracy evaluation index for railway recognition and can prove of each other.4. The research for real-time dynamic tracing system. The way that dynamically increases or lessens work threads in a thread pool is introduced in the design of the system to solve the problem that it takes different time for the single-frame algorithm. The robustness and practicability of the system are obviously improved. In the process of the single-frame recognition, we use multi-thread parallel processing for the left and right railway recognition to meet the real-time requirement.5. The establishment of video library for the tram railway. In the field of tram and high-speed railway recognition, there is no video library at home and abroad. In this paper, we collect all kinds of videos from Suzhou High-Tech Zone Tram Line 1 to establish a standardized video library which covers different weather, scenes, track trends. The establishment of video library provides help for the research and test.