提出一种基于TMS320DM642DSP平台的实时人眼检测系统设计.所用算法结合了人眼在主动照明下的亮瞳效应,使用包括Adaboost,SVM的“人脸-人眼”二级检测算法搜索并确定人眼位置,同时针对DSP平台特点进行相应优化.经实验表明,该DSP人眼检测系统对于640×480分辨率的连续视频能达到实时性要求,并有着良好的准确性.%A design of real-time eye detection system based on TI DM642 DSP is proposed. The detecting algorithm combines eye's bright pupil effect under active illumination. "Face-Eye" two level structure algorithm with Adaboost and SVM is adopted to search and locate user's eyes. Meanwhile, the algorithm is optimized according to DSP's characters. The experiment results indicate that this DSP eye detection system can reach the requirement of real-time processing for 640 * 480 pixels continuous video, and the accuracy and stability are satisfactory.
展开▼
