从公元2000年左右Viola和Jones提出级联检测器框架并成功应用于人脸检测后,实时目标检测才真正商业化。之后许多研究者对其改进,以下罗列了一些比较有代表性的论文和技术报告,并简洁地总结了他们的思想。如有谬误,欢迎拍砖。
Rapid Object Detection using a Boosted Cascade of Simple Features. Paul Viola, Michael J. Jones. CVPR 2001.
Robust Real-Time Face Detection. Paul Viola, Michael J. Jones.IJCV, 2004.(经典的VJ-detector)
对VJ-detector的经验分析:
Empirical Analysis of Detection Cascades of Boosted Classifiers for Rapid Object Detection. Rainer Lienhart, Alexander Kuranov, Vadim Pisarevsky.
An Analysis of Viola-Jones Face Detection Algorithm. Yi-Qing Wang.IPOL. (非常详细地解析了VJ-detector, 比网上几乎任何资料都详细)
以下列出几篇基于VJ思想的级联检测器论文:
1. Learning Object Detection from a Small Number of Examples the Importance of Good Features. Kobi Levi, Yair Weiss. CVPR 2004.
级联Dominant Orientation Feature
2. Joint Haar-like Features for Face Detection. Takeshi Mita, Toshimitsu Kaneko, Osamu Hori, ICCV, 2005.
利用了特征的共生特性
3. Fast Human Detection using A Cascade of Histograms of Oriented Gradients. Qiang Zhu, Shai Avidan, Mei-Chen Yeh, Kwang-Ting Cheng. CVPR 2006.
级联HoG特征(局部图像块的HoG)
4. Face Detection Based on Multi-Block LBP Representation. Lun Zhang, Rufeng Chu, Shiming Xiang, Shengcai Liao, Stan Z. Li.ICB 2007.
级联块状LBP特征, OpenCV2.x以后已经实现了上述MB-LBP cascade detector.
5. Fast Training and Selection of Haar Features using Statistics in Boosting-based Face Detection. Minh-Tri Pham Tat-Jen Cham. ICCV 2007.
使用统计量快速训练方法, 训练速度提高10X, 缺点是特征池规模相同情况下收敛速度下降。
6. Fast Asymmetric Learning for Cascade Face Detection. Jianxin Wu, S. Charles Brubaker, Matthew D. Mullin, James M. Rehg.TPAMI 2008.
分析了检测与分类问题的区别, 使用非对称方法改进弱分类器。该文基于Viola早前的论文:
Fast and Robust Classification using Asymmetric Adaboost and a Detector Cascade. NIPS, 2001.
7. Learning SURF Cascade for Fast and Accurate Object Detection. Jianguo Li, Yimin Zhang. CVPR 2013.
级联SURF特征, 模型较小, 出自Intel lab。
从VJ之后的论文来看,主要从两个方面来改进:1. 改进弱分类器; 2. 找更具有鉴别能力的局部的、能快速计算的特征; 3. 使用更好的boosting算法。