Foreground_Detection-master
所属分类：其他
开发工具：matlab
文件大小：17843KB
下载次数：13
上传日期：2017-09-17 21:04:27
上 传 者：
马德里审判长
说明： 对于视频中的前景目标提取，并转为灰度图片
(For the foreground target in the video, extract it and turn it into a grayscale image)
文件列表:
Manifold_Ranking (0, 2016-03-23)
Manifold_Ranking\AdjcProcloop.m (972, 2016-03-23)
Manifold_Ranking\EMR.m (3486, 2016-03-23)
Manifold_Ranking\MR_demo.m (5452, 2016-03-23)
Manifold_Ranking\ReadDAT.m (424, 2016-03-23)
Manifold_Ranking\adjacency.m (2101, 2016-03-23)
Manifold_Ranking\colorspace.m (14019, 2016-03-23)
Manifold_Ranking\makeweights.m (215, 2016-03-23)
Manifold_Ranking\normalize.m (2365, 2016-03-23)
Manifold_Ranking\removeframe.m (1190, 2016-03-23)
Preprocess (0, 2016-03-23)
Preprocess\import_data.m (262, 2016-03-23)
Preprocess\parse_segtrackv1.m (1436, 2016-03-23)
Preprocess\parse_segtrackv2.m (2410, 2016-03-23)
Preprocess\removeBoundary.m (318, 2016-03-23)
SLICSuperpixelSegmentation.exe (192512, 2016-03-23)
Stage1 (0, 2016-03-23)
Stage1\diff_between_images.m (709, 2016-03-23)
Stage1\fill_fg_with_prop.m (806, 2016-03-23)
Stage1\get_init_mask.m (3012, 2016-03-23)
Stage1\get_init_mask_1.m (1447, 2016-03-23)
Stage1\get_init_mask_2.m (2047, 2016-03-23)
Stage1\get_init_mask_3.m (1465, 2016-03-23)
Stage1\get_init_mask_4.m (1353, 2016-03-23)
Stage1\get_init_mask_5.m (2926, 2016-03-23)
Stage1\get_mask_hist.m (792, 2016-03-23)
Stage1\match_proposals.m (3984, 2016-03-23)
Stage1\merge_hist_info.m (471, 2016-03-23)
Stage1\refine_with_adjacent_frame.m (4223, 2016-03-23)
Stage1\select_proposal.m (4743, 2016-03-23)
Stage2 (0, 2016-03-23)
Stage2\MR_image.m (3204, 2016-03-23)
Stage2\get_final_mask.m (3471, 2016-03-23)
detectGMM.m (1908, 2016-03-23)
get_saliency.m (4666, 2016-03-23)
gop (0, 2016-03-23)
gop\000019.jpg (58495, 2016-03-23)
... ...
# 动态背景下的前景提取
> 本项目为本科毕业设计的题目。
本实验旨在提出一种动态背景下的快速前景算法。首先,利用帧差、GMM[1]等方法计算出图像中等运动区域,然后结合GOP算法[2]得到的object proposal选取初步前景区域,并对用前后帧的proposal进行二分匹配,修正前景结果。接着使用对图像的超像素点[3]进行流形排序[4],以初步结果中的超像素为对探测点,构建像素点间的图模型。在排序后的基础上利用大津算法[5]实施二值化分割,得到最终的前景结果。
### 创新点
- 将流形排序应用于前景提取领域
- 对前后帧的前景结果使用二分匹配进行修复
- 在不使用光流信息的前提下实现对动态背景下的前景提取
### 数据集
[SegTrack数据集](http://cpl.cc.gatech.edu/projects/SegTrack/)
### 运行环境
* 操作系统:Windows 10 专业版***位
* 编程语言:Matlab 2015b
* 第三方库:[vlfeat](http://www.vlfeat.org/)
### 运行方法
1. 下载源码至本地`git clone https://github.com/chenzeyuczy/Foreground_Detection.git`。
2. 打开项目目录`cd Foreground`。
3. 在“Preprocess/import_data.m”文件中设置数据集路径。
3. 运行主程序`main()`。
### 发布协议
[MIT](https://opensource.org/licenses/mit-license.php)
### 引用
- [1]. LI, Hao; ACHIM, Alin; BULL, David R. GMM-based efficient foreground detection with adaptive region update. In: Image Processing (ICIP), 2009 16th IEEE International Conference on. IEEE, 2009. p. 3181-3184.
- [2]. P. Krahenbuhl and V. Koltun. Geodesic Object Proposals. ECCV. 2014. 725-739.
- [3]. R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua and S. Susstrunk. SLIC Superpixels Compared to State-of-the-Art Superpixel Methods. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2012. 2274-2282.
- [4]. Zhou D, Weston J, Gretton A. Ranking on Data Manifolds[J]. Neural Information Processing Systems, 2004
- [5]. Nobuyuki Otsu. A Threshold Selection Method from Gray-Level Histograms. IEEE Transactions on Systems, Man and Cybernetics. 1979. 9 (1): 62–66.
近期下载者：
相关文件：
收藏者：