Urban Road And Vehicles Information Collecting Research Based On The Remote Sensing Images

As the rapid development of remote sensing technology, an increasingimprovement was emerging in a variety of remote sensing images, and so as thecarriable information.Urban traffic are composed of three elements, including vehicles, roads andpeople. So roads and vehicles are two important aspects of the urban transport. Inorder to exact the information of both, based on imaging processing method and fullyanalysis of the remote sensing images, the paper conducts the research focus on theexaction of road edge and applying it on the study of traffic flow at the same time. Adual-threshold SSDA (Sequential Similarity Detection Algorithm) template matchingmethod is proposed in an image processing model, while with an optimization methodbased on the lane-road area to identify the information of vehicles. The maincontents include the following aspects:1. The removal of remote sensing images’ background. Though the analysis ofremote sensing images, a variety of ways mask are applied to separated vegetationlike trees, shadow parts that cover the road, and also, on the basis of GIS, the methodswith a polygon corresponding to road mask are used to achieve the rough edge of theroads.2. In terms of the extraction of roads image edge, pointing at the questionscoming from template matching method like low efficiency, narrow flexible space andsome other deficiencies., a dual-threshold SSDA template matching method isproposed to improve the speed of the template matching method, while expanding theflexible space based on the variable threshold. Applying the method to the extractionof the road edge, and the results compared with some other edge extraction methods.3. For the questions in the vesicles features exacting such as too much thedimensions needed, the overly complexity of classifiers, an optimization methodbased on road areas in lane lines was proposed to reduce the features in extraction ofvehicles identification, and lower the complexity of vehicles extraction, so thedetection efficiency would be improved.4. In response to the classifier which the extracting features were reduced, themethod combined with temple matching was adopt to achieve the extraction ofvehicles, with that the detection accuracy was further enhanced. Experimental resultsshow that the method achieved a test time dramatically reduced and without sacrificing too much accuracy, therefore it can be considered as the detectionefficiency was improved.By discussing the extracting processes and the key technologies particularly inthe remote sensing images and the information of traffic flow, this paper summarizesthe contents and point out the next research directions at the end.