Based On A Single View Camera Calibration And Its Application

In recent years, a new application has aroused in computer vision field--intelligent visual surveillance, which automatically analyzes the image sequences to achieve locating, identifying and tracking of moving objects without any assistance of human intervention. Then the behaviors of moving objects are obtained to accomplish daily management as well as make immediate respond to abnormal situations in time. However, camera calibration is a primary and vital topic in intelligent visual surveillance.Combined with the charateristics of intelligent surverllance, this paper proposes a fast, flexible and effective camera-field calibration approach on the basis of the existing calibration algorithm. The main job is as follows: we introduce Zhang's 2D planar object approach, 1D object based approach and scene's structure information based approach in detail, as well as present advantages and disadvantages of these methods. However, in absence of calibration objects or scene's structure information in camera field, these methods can not calibrate camera but play a simulating role. So this paper employs a new calibration method by using vertical ground objects.In vertical ground objects based approach, vertical vanishing point and the plane vanishing line can be obtained by top-points and bottom-points of vertical objects(always, human). Then, according to vanishing point and the plane line being function of camera's intrinsic parameters we can solve all the parameters of uncalibrated camera, including focal length, principal points, rotation metrix and translation metrix. Considering the importance of vanishing point and vanishing line to the whole algorithm and sensitivities of these two to noise, an optimal approach is proposed to increase the robustness and precision of the algorithm. Different from traditional cross-ratio method, this paper develops an uniform algebraic approach to gain vertical ground objects and vertical space objects. What is worth metioning is that these two applications are independent of camera parameters.At last, computer simulation shows us the optimal number and position of the man in camera field-calibration. It points out that the more sampling points can not better experimental result, indicates that the same number of sampling points must distribute in camera field as far as possible to obtain more precise results; Both computer simulation and real data validate the robustness and precision of proposed method for camera's parameters and measurement to the height of vertical ground objects and vertical space objects.