The Study Of Target Three-Dimensional Feature Extraction Based On Monocular Image

Traffic flow detection has been the base of traffic management departments to monitor, manage and statistics the state of highways. Whether can we objectively, comprehensively and timely get the indicators of traffic flow will determine the ability of effective grasp, prediction and timely control.Traditional traffic flow detection technique is based of electromagnetic and infrared detection, Such as burying electromagnetic induction coil, ultrasonic velocity detector, infrared detection devices, etc. However, all of these techniques mostly confined to the negative factors such as limited testing scope, cumbersome setting and transportation facilities transformation, etc. Instead, the traffic flow parameters detection method based video detection can be convenient and efficient detecting the traffic conditions and it has gradually become an effective tool to traffic control department.Based on computer vision and digital image processing technology, this article proposed an algorithm using two-dimension plane detection to build a three-dimensional space vehicle detection. We can complete three-dimensional information detection about vehicles in the view field under the condition of installation of single surveillance camera. Thereby, it will reduce the costs of transport infrastructure and improve the detection accuracy. The paper built a compatible calibration space of calibration plane and the world coordinate system. With it, we can complete a comprehensive and stereo detection of three-dimensional features of the target vehicle in the space. Moreover on this basis, using digital graphics matching algorithm, we has improved the previous tracking algorithm with three-dimensional characteristics and proposed the new tracking algorithm based on three-dimensional characteristics. In addition, we studied and implemented the algorithms such as background extraction, target detection, target feature extraction and target modeling. Moreover, we construct the Traffic Flow Detection CCTFSD which based single camera three-dimensional object detection.By the detection system constructed by the paper, we improved the object detecting and tracing arithmetic. After testing, the traffic flow detection system constructed by the paper archives a higher accuracy of indicators of traffic flow parameters, fulfilling practical requirement.