基于聚类算法的统计学习侧面人脸检测

提出了一种新型的基于聚类算法的统计学习侧面人脸检测算法。通过对侧面人脸的各个视角建立分类器，使用新的AdaBoost训练策略，然后应用该聚类算法进行检测结果融合，从而有效检测出人脸。并使用该算法成功解决耳朵检测问题，取得了较好的实验结果。实验表明，该检测算法可以有效地检测出侧面人脸和耳朵，是一种普遍有效的目标检测算法；对282幅侧面人脸图像进行人脸检测，检测率在漏检19个时为93.26%，；对1000包含耳朵的图像进行检测，耳朵检测率在误检个数为61时为91.9%。

Profile Face Detection Based on Statistical Learning

Propose a new profile face detection algorithm using clustering algorithm.First,we decompose the face into several planes and learn classifiers respectively.Second,we employ a new AdaBoost training procedure,and then use the clustering algorithm based on sample density to combine face detection results successfully.This algorithm we proposed can also detect ears with high detection rate.Experiments show that this algorithm is a general object detection algorithm,it can achieve 93.26% detection rate(false negative rate is 6.74%) on 282 profile face images,and it can reach the 91.9% detection rate(false positive rate is 6.1%) on 1 000 ear images.