基于高性能数字多媒体处理芯片TITMS320DM6446的类Haar特征人眼检测算法.通过将人眼检测算法封装为算法包并进行优化,实现基于DM6446平台的实时人眼检测系统.封装后的算法包置于DSP端供远程调用.ARM端的应用程序负责从摄像头采集人脸图像,并将结果输出到显示器.人眼检测算法包也被应用于人脸识别系统中.该算法具有搜索速度快、检测率高等优点,适合在便携式平台上实现.实验结果表明,该实时人眼检测系统能在占用较少资源的情况下能有效检测并定位眼睛.
A Haar-like features based eye detection algorithm was introduced in this paper. With its fast searching speed and high detection rate, the algorithm fits to be implemented on portable platform. TI TMS320DM6446 is a high pefformance digital multimedia processing chip provided by Texas Instruments Corporation. With the eye detection algorithm packed into eodec, an optimized real time eye detection system was implemented on DM6446 platform. The detection algorithm codee was loaded into the DSP-side for remote call. Application on the ARM-side captured face images from camera and output detection results to computer monitor. The implemented eye detection codec was also employed in a face recognition system. Experimental results demonstrate that the real time eye detection system on DM6446 can search and locate eyes effectively with limited chip resources occupied.