Research On Extraction Method Of Road Signs Information Based On Internet Street View Data

As one of the most important parts of the Intelligent Transportation System,road signs plays an important role in automatic driving,traffic signs maintenance etc.An efficient and fast detection and recognition method for road signs is a critical technology in ITS.However,because of the effects of the environmental factors,the real-time detection efficiency is low in real.And most of the road signs are recognized by their color or shapes’ classification result,which leads to the signs recognition is incomplete and cant guide the traffic.In view of the above problems,this paper puts forward the information extraction method of road signs based on the Internet street view to get the detail contents of signs.In order to improve the efficiency of road traffic and reduce road traffic accidents.The main work of this article includes the following:(1)Constructing the data model of sign recognitionIn this paper,I get the concept model of the sign recognition by interpreting the road sign relevant standards of China and USA,and also analyzing the signs relevant elements.By comparing the two countries concept models,put forward the universal concept models for sign recognition.(2)Road signs detection methodThis paper use the SSD method of object detection in the image to detect street signs;then using HOG feature and SVM to reclassify the targets,and combine the targets confidence with the result of the street view image clustering segmentation results to filter the wrong background image region,and training the segmentation results with SVM.Finally,the scene background of the sign are segmented by clustering image,and the area that may appear in the sign is reclassified,to reduce the leak detection rate.The experimental results show that the method can effectively improve the detection efficiency of the sign.(3)Information extraction method of road signFirstly,this paper rectify the slanted signs,secondly,the structure segmentation is carried out by image segmentation and scanning line algorithm.Then,based on related text database,symbol library,etc.,template matching method is used to recognize the signs,and reconstruct signs by the recognition results.The experimental results show that the detailed extraction of the sign information can be identified by the above methodsThe research results show that the road sign detection method in this paper can effectively improve the accuracy and robustness of road sign detection by adding different constraints to data sources.By analyzing and identifying the characters,colors,symbols and other characteristics of signs,we can effectively get the detailed information of the signs,and reproduce the signs.