Vehicle Detection And Track Based On Structural Feature

With the rapid growth of the gross domestic product(GDP) and improving living standard, vehicle has entered into thousands of Chinese families. At the end of 2013, the number of vehicles in China has more than 250,000,000, including 85,000,000 private cars, and is still increasing more than 14,000,000 annually. Therefore, the maintaining of the traffic order and ensuring the traffic safety become very important. Consequently, the real-time traffic surveillance will help administrator obtain important information easily and make proper decision.This paper mainly studies multi-scale vehicle identification and tracking framework based on digital image processing and structural features vehicle detection and tracking techniques from the following aspects:1, In image pre-processing procedure, extracting the region of interest(ROI) of the image is difficult since the shape, color and size are complex to the detection scenario. By utilizing RGB color space, multi-color channel is enhanced to obtain the key part information such as license plate and rear lamps.2, Aiming to extract the multi-scale vehicles structural information in a complex scene, a method based on the Maximally Stable Extremal Regions(MSER) is proposed to achieve fast detection of images' nodes.3, In order to solve the problem of efficient identification to the correct vehicle information, a support vector machine(SVM) classifier is proposed to identify the right combination. Compared with some existing literature, the effectiveness and the robustness of the proposed method is verified on both CPU time and detection probability through real world traffic monitoring video clips.4, A Kalman filter algorithm is utilized to achieve fast vehicles tracking in complex urban traffic surveillance. Simulations and experiments demonstrate the robustness of the algorithm in various scenarios.Simulations and experiments show the proposed method with high robustness, detection and tracking rate as well as low false alarm rate.