摘要
研究人体运动状态下目标特征采集准确度优化问题。传统的视觉检测方法在运动视觉差的干扰下,对运动人脸检测存在较大缺陷,结合人类的视觉特性,提出一种通过实时视频中的图像快速检测出运动物体的类型及其特征的算法。利用基于边界追踪和多角度成像的方法对数码摄像机定位,然后判断实时采集图像的类型,如果图像为人物,则用DWT算法进行人脸检测与特征定位。通过仿真可以表明,改进的算法具有计算量小,效率高的特点,适用于实际生活。
The accuracy optimization of target feature acquisition in body motion was researched. Combined with human visual characteristics, an improved detection and feature extraction algorithm was proposed. The classification and feature of moving object can be detected in real time video. The digital camera was located based on boundary tracing and multi angle imaging method. The type of real-time image acquisition was judged. If the image is human, the DWT algorithm is used for face detection and facial feature location. Simulation results show that the improved al- gorithm has the advantages such as less computing cost and high efficiency. It can be applied in the real life.
出处
《计算机仿真》
CSCD
北大核心
2014年第1期434-437,共4页
Computer Simulation
关键词
边界追踪
旋转矩阵
运动目标检测
小波变换
人脸特征定位
Boundary tracing
Transformation matrix
Moving target detection
DWT
Facial feature localization