为弥补传统监控视野范围固定缺陷,并实现运动人体实时监控,采用图像处理方法,设计了一个基于数字信号处理器件(Digital Signal Processor,DSP)车载摄像头运动人体小车跟随系统。采用帧间差分法检测识别运动人体,提出一个基于灰度直方图的连续自适应均值漂移运动人体跟随算法,通过计算运动人体反向投影图的0阶矩阵和1阶矩阵,求出运动人体的质心坐标和宽度,作为下一帧跟踪框位置和大小。实验结果表明:电荷耦合器件摄像头采集的运动人体图像在DSP进行检测识别和跟随处理,正确地获得左转、右转、前进和后退等四个驱动信号,驱动小车跟随运动人体。实地测试表明:该系统能够实现对运动人体目标左右前后跟随,并时刻与运动人体保持一定安全距离。
To make up for the defects of fixed visual range and monitor human moving in real time,by using of image processing method,a vehicle system of following moving human has been designed. The system is based on digital signal processor( DSP) and on-board camera. Frame difference method has been used to detect moving human,and a following algorithm based on continuously adaptive mean shift( Cam Shift) using gray histogram has been proposed. It calculates the zero and first order matrix of the moving human’s reverse projection and then determines its mass coordinate and width as the position and magnitude of tracking window in the next frame. Charge Coupled Device( CCD) camera captures the images of moving human and these images are processed in DSP later to detect and follow the target. By receiving signals of turning left,turning right,going forward and backward,a vehicle can be driven to follow moving human. The actual tests show that the system can achieve the goal of following human in left,right,front,rear direction as well as always keeping a safe distance with the moving human.