matlab求指教

PSYZ1666

基于稀疏低秩分解的杂草种子配准那篇文章有人知道么，里面有个rasl算法，可以人脸识别配准，它人脸圈划出红绿两种颜色的框，原理是什么，为什么我更改识别区域的话红绿框的大小就不一样了，求大神解答，我是用这个算法配准图像识别猫眼目标的。
这是图片

                               
登录/注册后可看大图

这是部分代码

                               
登录/注册后可看大图

又是那隻貓

图片挂了

PSYZ1666

额。。。图片怎么挂了。。。

PSYZ1666

1. % Yigang Peng, Arvind Ganesh, November 2009.
   
2. % Questions? abalasu2@illinois.edu
   
3. %
   
4. % Copyright: Perception and Decision Laboratory, University of Illinois, Urbana-Champaign
   
5. %            Microsoft Research Asia, Beijing
   
6. %
   
7. % Reference: RASL: Robust Alignment by Sparse and Low-rank Decomposition for Linearly Correlated Images  
   
8. %            Yigang Peng, Arvind Ganesh, John Wright, Wenli Xu, and Yi Ma. Proc. of CVPR, 2010.
   
9. %
   
10. 
    
11. % robust batch image alignment example
    
12. 
    
13. % clear
    
14. clc ;
    
15. %clear all;
    
16. close all ;
    
17. 
    
18. % addpath
    
19. addpath RASL_toolbox ;
    
20. addpath data ;
    
21. addpath results ;
    
22. 
    
23. %% define images' path
    
24. 
    
25. currentPath = cd;
    
26. 
    
27. % input path
    
28. imagePath = fullfile(currentPath,'data','LFW') ;
    
29. pointPath = fullfile(currentPath,'data','LFW') ; % path to files containing initial feature(eye) coordinates
    
30. userName = 'Jacques_Chirac' ;
    
31. % one of
    
32. % 'Gloria_Macapagal_Arroyo', 'Jennifer_Capriati', 'Laura_Bush',  
    
33. % 'Serena_Williams', 'Barack_Obama', 'Ariel_Sharon',
    
34. % 'Arnold_Schwarzenegger', 'Colin_Powell', 'Donald_Rumsfeld',
    
35. % 'George_W_Bush', 'Gerhard_Schroeder', 'Hugo_Chavez',
    
36. % 'Jacques_Chirac', 'Jean_Chretien', 'John_Ashcroft',
    
37. % 'Junichiro_Koizumi', 'Lleyton_Hewitt', 'Luiz_Inacio_Lula_da_Silva',
    
38. % 'Tony_Blair', 'Vladimir_Putin'
    
39. 
    
40. % output path
    
41. destRoot = fullfile(currentPath,'results') ;
    
42. destDir = fullfile(destRoot,userName) ;
    
43. if ~exist(destDir,'dir')
    
44.     mkdir(destRoot,userName) ;
    
45. end
    
46. 
    
47. %% define parameters
    
48. 
    
49. % dispaly flag
    
50. raslpara.DISPLAY = 1 ;
    
51. 
    
52. % save flag
    
53. raslpara.saveStart = 1 ;
    
54. raslpara.saveEnd = 1 ;
    
55. raslpara.saveIntermedia = 0 ;
    
56. 
    
57. 
    
58. % for face images
    
59. raslpara.canonicalImageSize = [ 120 100  ];
    
60. raslpara.canonicalEyeCoords = [ 5  25 ; ...
    
61.                                 5 5  ];
    
62.                            
    
63. % parametric tranformation model
    
64. raslpara.transformType = 'SIMILARITY';
    
65. % one of 'TRANSLATION', 'SIMILARITY', 'AFFINE','HOMOGRAPHY'
    
66. 
    
67. raslpara.numScales = 1 ; % if numScales > 1, we use multiscales
    
68. 
    
69. % main loop
    
70. raslpara.stoppingDelta = .000001; % stopping condition of main loop
    
71. raslpara.maxIter = 250; % maximum iteration number of main loops
    
72. 
    
73. % inner loop
    
74. raslpara.inner_tol = 1e-6 ;
    
75. raslpara.inner_maxIter = 1000 ;
    
76. raslpara.continuationFlag = 1 ;
    
77. raslpara.mu = 1e-3 ;
    
78. raslpara.lambdac = 3 ; % lambda = lambdac/sqrt(m)
    
79. 
    
80. 
    
81. %% Get training images
    
82. 
    
83. % get initial transformation
    
84. transformationInit = 'SIMILARITY';
    
85. [fileNames, transformations, numImages] = get_training_images( imagePath, pointPath, userName, raslpara.canonicalEyeCoords, transformationInit) ;
    
86. 
    
87. 
    
88. %% RASL main loop: do robust batch image alignment
    
89. 
    
90. [D, Do, A, E, xi, numIterOuter, numIterInner ] = rasl_main(fileNames, transformations, numImages, raslpara, destDir);
    
91. 
    
92. %% plot the results
    
93. 
    
94. layout.xI = 5 ;
    
95. layout.yI = 7 ;
    
96. layout.gap = 2 ;
    
97. layout.gap2 = 1 ;
    
98. rasl_plot(destDir, numImages, raslpara.canonicalImageSize, layout)
    

复制代码

这是主要的代码部分，其中有
% for face images
raslpara.canonicalImageSize = [ 120 100  ];
raslpara.canonicalEyeCoords = [ 5  25 ; ...
                                5 5  ];
我感觉是识别区域的，求大神讲解

PSYZ1666

又是那隻貓 发表于 2015-5-29 20:27
图片挂了

大神我问个愚蠢的问题，怎么正确的发图片啊。。。

PSYZ1666

求大神啊！！！