基于自举算法和支持向量机的人脸检测系统

Face detecting system based on AdaBoost algorithm and Support Vector Machine

针对复杂背景下的灰度图像人脸检测存在计算量大且负检率高等问题,提出了一种有较好可用性的层级递进的人脸检测系统。系统第一部分采用扩展的Haar型特征并结合自举算法,使其分类性能要优于原始的Haar型特征。在系统的第二部分,采用从粗到细的视觉处理逻辑对图像采样,并提出了正面直立人脸的像素值的置信度的概念,且以支持向量机作为学习算法,使系统具有良好的检测性能。该系统在实际应用图像的测试中取得良好效果,具有可用性。

For there were some tough problems in face detection under various background in gray level images,like too time-consuming and high false positive rate,this paper developed an applicable cascade face detecting system.First part of the system adopted extended Haar-like features combined with AdaBoost algorithm,which made performance of classifying better than using original Haar-like features.Second part of the system adopted vision logic,a strategy from general to detail to sample images,and introduced the conception of front-right face image pixel confidence,and took Support Vector Machine as learnnig algorithm.This part ensured the good accuracy of the detection.This system perfomed well on real image test.This was an applicable face detecting system.