智能监控系统中自适应人脸检测跟踪算法改进  

Improvement of Adaptive Face Detecting and Tracking Algorithm in Intelligent Monitoring System

作　　者：何登平[1,2] 黄凌云[1] 

机构地区：[1]重庆邮电大学通信新技术应用研究中心,重庆400065 [2]重庆信科设计有限公司

出　　处：《单片机与嵌入式系统应用》2016年第5期11-14,共4页Microcontrollers & Embedded Systems

摘　　要：针对传统人脸检测跟踪算法在复杂环境中准确率不高,以及在跟踪过程中易受到周围相似色物体干扰、遮挡丢失等问题,提出了一种改进型自适应人脸检测跟踪算法。该算法通过人脸检测(Adaboost)与主动形状建模(ASM)算法相结合,降低了复杂环境中的人脸误检率;通过对运动目标跟踪(Camshift)算法提取H-S二维颜色概率直方图,并结合Kalman滤波器有效解决了相似色干扰及遮挡丢失问题。实验证明,改进型算法不仅在复杂环境中人脸检测率高、抗干扰能力强,且满足实时性的需求,是一种建立实时智能监控系统的实用方法。Aiming at the problems of the traditional face detection and tracking algorithm in complex environment such as the face detec- tion accuracy is not high, it is easily affected by the surrounding similar color object interference in the process of tracking and shielding loss etc, so an improved adaptive face detecting and tracking algorithm is proposed. The algorithm combines the face detection and the ac- tive shape model algorithm, so the false detection rate of faces in a complex environment is reduced. The H-S two-dimensional color prob- ability histogram is extracted based on the moving target tracking algorithm,which combines the Kalman filter to solve the similar color interference and shielding loss. The experiment results show that the improved algorithm not only has the high rate of face detection, strong anti-interference ability in the complex environment, but also meets the real-time requirement, so it is a kind of practical method of the real-time intelligent monitoring system.

关 键 词：人脸检测跟踪算法 二维颜色概率直方图 KALMAN滤波器 TMS320DM3730 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]