Research On Traffic Indication Sign Interpretation Method

Traffic sign interpretation is the core component of intelligent transportation system,and the main purpose is to interpret the signs quickly and accurately. Then remind thedrivers the existence of the sign and its meaning, reduce the traffic accidents and finallyensure the safe and reliable driving. Therefore, the research for the traffic signinterpretation method is highly necessary and has important value for practicalapplications.At present, the general phenomena of the traffic indication sign interpretation in thetechnology and practical applications are shown as follows. Firstly, the imbalance of thealgorithm’s robust and real-time is general. The key points and difficulties of theinterpretation algorithms research are how to make them reach equilibrium. Secondly, thegenerality and universality of the system is poor. This paper has established two-tiersystem, including the context layer and general processing layer, is based on informationentropy and knowledge representation. The general processing layer which is theunderlying based on the digital image processing technology to complete the targets oftraffic signs image preprocessing, feature extraction and selection, feature representation.The context layer which is the top layer uses the knowledge base guidance of theunderlying to extract features and formalize description, then using the interpretationresults to update the information of knowledge base. The details of research are shown asfollows.Firstly, research the method of feature selection which based on information entropy.Getting effective image features is both the key to solve the problems of traffic signinterpretation and the important research contents in this paper. The direction of featureselection in pattern recognition is pointed out by the information entropy theory. Theauthor has selected the most effective multiple-feature set as the basis of feature selectionaccording to the study and research of the theory. And it keeps the whole operation processin the low-dimensional space which enhances operation efficiency and real-time of system.Secondly, research on feature extraction method. The features of the traffic signsimages can be divided into contour shape and internal detail which leads to two aspects forthe extraction of image features in the study. Firstly, the paper proposes an improved edgedetection method for implementing the extraction of contour features. And then, themultiple features of internal graph are obtained according to projection curve.Finally, the research on formal description of the image features and the establishmentof the simulation model which based on the above. The reasonable representation of features has a powerful description and discrimination for distinguishing between differentcategories of images or images with interferences. Meanwhile, it is the key point toimprove the universality in the traffic sign indicating interpretation system. The authorintegrates the predicate structure of knowledge representation and the set theory tohierarchically describe features, and then constructs the rule base for interpretation system.This can achieve the communication between the top and bottom so as to the universalityof the system.The experimental result shows that the traffic indication sign interpretation model hasconstructed in this paper which successfully implements any indication signs accuratejudgment and recognition. The model has a certain robustness, real-time and universality.