摘要
随着社会的不断发展,实时人脸检测与跟踪已在安全监控、人机交互等领域广泛应用。Adaboost算法利用haar特征以及其检测速度快被广泛使用,但对侧面人脸、扭曲变形人脸检测效果不理想。针对这一问题,提出将椭圆肤色检测引入人脸检测系统中。针对Camshift算法需要手动选择跟踪对象,跟踪检测出的人脸,实现实时、自动人脸跟踪。同时,在跟踪过程中引入距离约束条件,使跟踪结果更加稳定。实验结果表明,在opencv的基础上,采用肤色检测、Adaboost算法以及Camshift算法相结合的方法,实现了快速、自动和准确的人脸检测和跟踪。
With the continuous social development,the real-time face detection and tracking is widely used in the security monitoring, human-computer interaction and other fields. Adaboost algorithm, for Haar feature and fast detection speed, is widely applied, but is not so ideal for side-face and distorted-face detection. To solve this problem, elliptic skin color detection is introduced into face detection system. Camshift algorithm requires manual selection of the tracking object, thus to track the detected face and realize real-time and automatic face tracking. Meanwhile, distance constraint is introduced into the tracking process, thus to make the tracking results more stable. The experimental results indicate that, based on the OpenCV and in combination with skin color detection, Adaboost algorithm and Camshift algorithm can realize fast, automatic and accurate face detection and tracking.
出处
《通信技术》
2017年第7期1412-1416,共5页
Communications Technology
