人体特征检测算法的设计与实现

摘要

Human feature detection is the basis of feature recognition technology which has been applied to many areas,from the face of criminal investigation to fingerprint unlock in the field of mobile application is involved. In order to achieve the analysis and study of hu-man characteristics detection algorithm,in platform on field-programmable gate array,a body feature detection algorithm based on SAda-boost is presented. The algorithm combines the advantages of SVM and Adaboost classifier algorithm to reduce the dimension of human characteristics and to achieve the detection of human characteristic elements by image classification processing. Entire design is based on ZedBoard,which has a strong ability to reconfigurable and parallel processing,completing the design and implementation of human fea-tures image detection algorithm,realization of detection for human body elements,including face,eye,and human framework. Through comparative analysis of the experimental results,the validity of the algorithm is verified.%人体特征检测是特征识别技术的基础，特征检测技术运用到诸多领域，从刑侦领域的人脸检测到手机应用领域的指纹解锁都有涉及。为了实现人体特征检测算法的分析与研究，在现场可编程逻辑门阵列平台中，提出一种基于SAda-boost的人体特征检测算法。该算法结合了支持向量机和Adaboost分类器算法的优点，实现人体特征图像的降维处理，并对图像进行分类处理，实现人体特征元素的检测。整个设计基于ZedBoard硬件平台，该平台具有较强的可重构性和并行处理的能力，完成了人体特征图像检测算法的设计与实现，实现了对人体特征图像中的人脸、人眼、人体框架3种人体元素的检测。通过对比分析实验结果，验证了该算法的有效性。