| The traditional computer vision technology is mainly based on the color or gray level information captured by the color camera,which is greatly affected by the change of the ambient light and the color or texture of the object.The contour of the object is unstable in the image frame sequence,which is also a problem that computer vision technology needs to focus on.The appearance of ToF camera has a revolutionary influence on computer vision.ToF camera illuminates the scene with near-infrared light source and measures the phase shift between the illumination and reflection light to caculate the depth map of the target area.Under ToF camera,the contour of the object is very stable and the precision is high.The depth map is hardly affected by ambient light,color or texture.Target detection using depth map can avoid many problems in traditional algorithm.The accuracy of target detection will be improved if combined with color camera.In this paper,the people counting algorithms is divided into two parts: target detection and target tracking.Combined with the advantages of common algorithms,a people counting algorithm that meets the characteristics of security scene is proposed.Firstly,we studied and compared the advantages and disadvantages of the common target detection and target tracking techniques.Secondly,in the aspect of target detection,a new algorithm based on edge information is proposed,which combines the background subtraction and three-frame subtraction.The principle and the key problems of the algorithm are discussed in detail.In the aspect of target tracking,an algorithm of Meanshift and Kalman Filter combined is proposed,solving the problem of occlusion.Finally,we use ToF camera in the Microsoft Kinect 2.0 sensor to capture depth information,and OpenCV library to implement the algorithm. |
