这篇写在两年前!!!
目标检测问题,对应英文:Target Detection
下面是解决这类问题的state-of-the-art方法的基本介绍:
https://www.zhihu.com/question/34223049
https://zhuanlan.zhihu.com/p/21533724
http://zhangliliang.com/2015/05/19/paper-note-object-proposal-review-pami15/
http://people.eecs.berkeley.edu/~rbg/index.html
https://zhuanlan.zhihu.com/p/21412911
http://blog.cvmarcher.com/posts/2015/05/17/multi-region-semantic-segmentation-aware-cnn/
最早的方法是:Appearance-based methods,这种方法是我们理解一个目标最直接的方式,比如RGB颜色分布,形状大小,位置方向等等;比如Halcon中的各种Region Selection方法。
传统的边缘匹配,滑窗,灰度匹配,梯度匹配,直方图及改进的直方图算法匹配,Large modelbases等。
模板匹配,我觉得模板匹配应该是基于Appearance-based methods的,具体我不熟:
http://blog.csdn.net/zouxy09/article/details/8549743
http://www.halcon.com/download/reference/create_template.html
然后便是Feature-based methods
对于传统的方法,其对应经典特征的基本介绍:
Hog:
http://blog.csdn.net/zouxy09/article/details/7929348
http://blog.csdn.net/abcjennifer/article/details/7365651
http://www.cnblogs.com/tornadomeet/archive/2012/08/15/2640754.html
https://github.com/DaHoC/trainHOG
https://github.com/ivanaslamov/hog_sse
Sift:
http://blog.csdn.net/abcjennifer/article/details/7639681
SURF:
https://zh.wikipedia.org/wiki/%E5%8A%A0%E9%80%9F%E7%A8%B3%E5%81%A5%E7%89%B9%E5%BE%81
这些特征OpenCV中都有:
https://www.zhihu.com/question/24038129
DPM算法:
https://www.zhihu.com/question/29300042
http://blog.csdn.net/ttransposition/article/details/12966521
http://blog.csdn.net/ttransposition/article/details/41806601
https://github.com/rbgirshick/voc-dpm
最后便是深度学习了
其他的一些背景资料:
https://en.wikipedia.org/wiki/Outline_of_object_recognition
https://zhuanlan.zhihu.com/p/21344595?refer=xlvector
人脸识别(人脸识别可以推广到各种物体识别),下面是传统的,OpenCV里有的方法:
http://blog.csdn.net/smartempire/article/details/21406005
http://blog.csdn.net/smartempire/article/details/23249517
http://blog.csdn.net/smartempire/article/details/23377385
未完待续...
原文地址:https://www.cnblogs.com/heubme/p/9349958.html