Key Technology Research In Transportation Object Recognition

Object recognition is an important branch of computer vision, object recognition in intelligentrobots, intelligent traffic monitoring, and smart home has important applications. Transportationobject recognition is important and difficult intelligent transportation and unmanned research inthe field, it is mainly to complete the identification of vehicles and pedestrians on the road throughcomputer vision. Because traffic object recognition system when solid is relatively high,especially when the vehicle is blocked greatly increases the difficulty of identification.Based on the research study identified several key technical traffic object recognition, themain contents are as follows:The first chapter outlines the research focused on the purpose and significance of thisresearch project, and from both domestic and foreign elaborated research status of traffic objectrecognition.The second chapter studies the basic image processing algorithms, including several commonprinciples and extraction algorithm of color space model, spatial and frequency filtering, andimage features two school Sift with LBP features and characteristics.The third chapter studies the vehicle identification algorithm based on Haar classifier, whichfocuses on the principles of composition principle strength classification algorithms and Haarclassifier and using integral image method to quickly calculate Haar-like features of the image,and finally to out of the test results Haar classifier to identify the vehicle.The fourth chapter studies the video motion detection algorithm of the vehicle, whichintroduces the single Gaussian background modeling method and Gaussian mixture backgroundmodeling method, and pointed out some problems Gaussian mixture background modelingmethod gives improved measures, in the final analysis and comparison of the results given the useof Gaussian mixture background modeling and improved Gaussian mixture background modelingconducted to detect moving vehicles and vehicle-related data.The fifth chapter studies the object recognition algorithm pedestrian traffic, mainly to studythe principles of SVM Hog features and algorithms, and propose the use of a dual classificationidentifies pedestrian approach Finally, in recognition of the dual classification results and OpencvHog comes comparing the results to identify pedestrians.The sixth chapter summarizes the full text of innovation research and proposed shortcomingsand prospects methods.