Research On PTU Camera Calibration Technology Based On GPS Positioning

In the field of computer vision, studies on object tracking are most visual tracking methods based on 2D image; visual tracking can provide a lot of specific information about moving targets but no the coordinates in real 3D space. It is unique and convenience when expresses position of one target in the form of 3D spatial coordinates, such as GPS coordinates,which is of great significance. Camera calibration technology is the bridge of connectting 2D images and 3D world,and it is the prerequisite for machine vision research,and gradually developed camera calibration technology has been widely used in vision measurement, target identification and tracking and so on.In order to expand the scope of locating and tracking, we use PTU camera to locate a target,however,the flexibility of PTU camera makes it difficult to be calibrated,and the existing video surveillance systems do not provide information about the coordinates of targets in the FOV, and most measuring systems based on calibration pattern are not suitable for large field environment. This paper presents an improved method with geometric constraints for camera calibration technique based on GPS positioning,in order to solve the locating and tracking problem in large field of view situation, in which GPS positioning coordinates are used to establish a unified global world coordinates in the process of calibrating a camera,and this method is fast,convenient,and can be use in large FOV.Firstly, we select the appropriate camera calibration method and determine the parameters of the model to be calibrated, according to the application characteristics of this article. To ensure the of the camera calibration in the case of the limited of GPS positioning accuracy, we use a method based on improved circular geometric constraints to optimize the GPS positioning coordinates, and extend it to the calibration of PTU camera. We also studied on one kind of target detection algorithm based on color histogram of target is presented in this paper in order to obtain image coordinates of the calibration target automatically and precisely.Secondly, as the GPS locating signals are susceptible to interference and the coordinates do not satisfies Gauss distribution, GPS coordinates with large errors are optimized away with RANSAC algorithm based on the geometry constraints to GPS antenna. Experiments show that the optimization to observed data is useful to improve the accuracy of the calibration,even when the localization accuracy of GPS is low.Thirdly,a dynamatic model due to the variation of attitude angles is used to update the external parameter matrix of calibrated cameras when the Pan-Tilt attitude angles are changed,in order to locate and track the target in world coordinate system (GPS latitude and longitude coordinates system). The trajectory in image coordinate system of moving target is obtained with target tracking algorithm based on correlation filter using kernel functions in order to achieve the 2D image coordinates of a certain target.Finally, target detecting and tracking experiments and experiments about measuring the Euclidean distance between selected points are designed to show the feasibility of the proposed method in this paper.