Research Of Vehicle Type Recognition Based On Bayonet Images

With the increasingly serious of traffic problems, the development of Intelligent Transportation System(ITS) technology is particularly impendency. As an important part of ITS,vehicle type recognition system plays a key role in ITS. This paper studies the vehicle type recognition of algorithms and techniques in the Intelligent Transportation System, mainly including three important part of the vehicle type recognition, namely vehicle detection and location, vehicle feature extraction and vehicle classification.(1) In terms of vehicle detection and localization, this paper proposes a vehicle detection method that combining preliminary detection with precise localization. Firstly, extracting HOG(Histogram of Oriented Gradient) feature of image, combining with SVM (Support Vector Machine) classifier to detect vehicle preliminary, and then using mathematical morphology and its opening and closing reconstruction method to eliminate noise and unrelated edge, extracting vehicle edge information, and proceed edge of image projection, proposing to convert the projection image into a binary image, and then processing the binary image to eliminate noise and partial projection glitches using mathematical morphology method, finally locating the vehicle boundaries.(2) In terms of vehicle feature extraction, this paper analyzes and compares the characteristics of the vehicle extraction methods, including invariant moment features, texture features, HOG features, SURF (speed up robust features) features and Integral Channel Features, considering the features of the performance, we use the SURF features and Integral Channel Features of the vehicle as input features of the vehicle identification classifier.(3) In terms of Vehicle classification, this paper introduces the classification performance and characteristic of SVM classifier, as well as introduces the method use two class classifier for multi-classification. comparing and analyzing the vehicle identification effect of different features by SVM classifier, verified the effectiveness of choice the SURF features and Integral Channel Features as vehicle features of this vehicle recognition system.Vehicle type recognition system in this paper use Visual Studio 2010 of Microsoft as a development environment, using C/C++ programming, and with the open source image library OpenCV for implementation correlation algorithm, using industrial camera collect vehicle images. The experiments indicate that the method in this paper has better recognition performance, and good robustness, it has a certain value and significance for the research and development of automatic vehicle recognition.