Video Of Models Identify The Key Technology-based Applications

At present,intelligent transportation system(ITS) has become a hot research task which is applied to the electronics in the traffic field in the world.Along with the study on application of ITS, image processing and pattern recognition stand on the front research fields of ITS,and have great theory and application value.With the video equipment greatly used in the traffic system,how to get more traffic data and improve the equipment capacity factor have become an important topic in the world.And more,for overcoming the shortcomings of the traditional detection and classification,such as the difficulty of installation and maintenance,subjection factors in the manual identification,this paper mainly study on this topic.Vehicle Classification based on video sequences image obtain the information of ITS by using the digital image processing and computer vision.It collects,identifies and classifies the vehicles at the specific location and time,and takes it as the traffic management,fees,scheduling and statistics.Based on the research status of the image processing and pattern recognition,this thesis deeply analyzes the previous algorithms of objects detection and pattern matching,take the sequence of video image by the ordinary video monitoring as the main research objects,and describe the two key technics--vehicle detection and identification,especially study the algorithms of object detection, image segmentation,feature extraction and recognition method.Then,this thesis analyzed the results with the images taken in different angles,and introduced into many basic operations,such as dilation,erosion etc.But,this thesis takes more attention on the connected component labeling and gives an algorithm,connected component labeling based on contour tracking.In the process of video obtaining,the variation of outside environment including little dithering of vidicon and weather changing will affect the precision of moving target detection.Aiming at these problems,this paper presents a method of background reconstruction,in which there are moving targets in scene.The background can be dynamically changed.It can decrease the influence of outside environment variation and shorten the time of measurement.In addition,this paper modified the traditional approach,moving vehicle detection based on virtual wirefram.The vehicle classification takes the features extracted from the vehicle lateral image,and takes the whole contour shape as the recognized features.This paper gives an approach,which is vehicle classification based on characteristics of contour.Eigenvector has been first structured by using length to width ratio,posture ratio,rectangle degree,elongation degree and roundness degree.Then the vehicle type recognition and classification is successfully realized by nearest neighbor method.The technique mentioned above is implemented by experiments with VC++6.0.The experimental results show that the technique has the advantages of high precision and less calculation.