1 | % * This code was used in the following articles:
|
---|
2 | % * [1] Learning 3-D Scene Structure from a Single Still Image,
|
---|
3 | % * Ashutosh Saxena, Min Sun, Andrew Y. Ng,
|
---|
4 | % * In ICCV workshop on 3D Representation for Recognition (3dRR-07), 2007.
|
---|
5 | % * (best paper)
|
---|
6 | % * [2] 3-D Reconstruction from Sparse Views using Monocular Vision,
|
---|
7 | % * Ashutosh Saxena, Min Sun, Andrew Y. Ng,
|
---|
8 | % * In ICCV workshop on Virtual Representations and Modeling
|
---|
9 | % * of Large-scale environments (VRML), 2007.
|
---|
10 | % * [3] 3-D Depth Reconstruction from a Single Still Image,
|
---|
11 | % * Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng.
|
---|
12 | % * International Journal of Computer Vision (IJCV), Aug 2007.
|
---|
13 | % * [6] Learning Depth from Single Monocular Images,
|
---|
14 | % * Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng.
|
---|
15 | % * In Neural Information Processing Systems (NIPS) 18, 2005.
|
---|
16 | % *
|
---|
17 | % * These articles are available at:
|
---|
18 | % * http://make3d.stanford.edu/publications
|
---|
19 | % *
|
---|
20 | % * We request that you cite the papers [1], [3] and [6] in any of
|
---|
21 | % * your reports that uses this code.
|
---|
22 | % * Further, if you use the code in image3dstiching/ (multiple image version),
|
---|
23 | % * then please cite [2].
|
---|
24 | % *
|
---|
25 | % * If you use the code in third_party/, then PLEASE CITE and follow the
|
---|
26 | % * LICENSE OF THE CORRESPONDING THIRD PARTY CODE.
|
---|
27 | % *
|
---|
28 | % * Finally, this code is for non-commercial use only. For further
|
---|
29 | % * information and to obtain a copy of the license, see
|
---|
30 | % *
|
---|
31 | % * http://make3d.stanford.edu/publications/code
|
---|
32 | % *
|
---|
33 | % * Also, the software distributed under the License is distributed on an
|
---|
34 | % * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
|
---|
35 | % * express or implied. See the License for the specific language governing
|
---|
36 | % * permissions and limitations under the License.
|
---|
37 | % *
|
---|
38 | % */
|
---|
39 | function [Pair fail]=TestWholePostMatch(defaultPara, ImgInfo1, ImgInfo2, PriorPose) |
---|
40 | |
---|
41 | % This function load the Initial SurfMatches and the OccluSurfMatches |
---|
42 | % then, do RANSAC algorithm to prune the bad matches out |
---|
43 | % Calucluate New Camera Pose (R and T) |
---|
44 | |
---|
45 | % Default Parameters ----------------------------- |
---|
46 | fail = 0; |
---|
47 | % ------------------------------------------------ |
---|
48 | |
---|
49 | % Initialize Variables --------------------------- |
---|
50 | Img1 = ImgInfo1.ExifInfo.IDName; |
---|
51 | Img2 = ImgInfo2.ExifInfo.IDName; |
---|
52 | I1=imreadbw([defaultPara.Fdir '/pgm/' Img1 '.pgm']); % function from sift |
---|
53 | I2=imreadbw([defaultPara.Fdir '/pgm/' Img2 '.pgm']); % function from sift |
---|
54 | % ------------------------------------------------ |
---|
55 | |
---|
56 | % Combine Both Initial SurfMatches and the OccluSurfMatches ------------------------- |
---|
57 | % Also Assign Confidence of each Matches |
---|
58 | % (Used as RANSAC Sampling Distribution) |
---|
59 | % 0) load cleaned Fisrt set of matches |
---|
60 | %defaultPara.MaxUniqueRatio |
---|
61 | load([defaultPara.Fdir '/data/' Img1 '_' Img2 '_PoseMatch.mat']); |
---|
62 | defaultPara.MaxUniqueRatio = 100; |
---|
63 | % 1) load Initial SurfMatches |
---|
64 | [f1, f2, matches] = readSurfMatches(Img1, Img2, defaultPara.Fdir, ... |
---|
65 | [ defaultPara.Type 'Dense_' num2str(defaultPara.AbsThre) '_' num2str(defaultPara.RatioThre)], 1, 1); |
---|
66 | if isempty(f1) |
---|
67 | [f1, f2, matches] = readSurfMatches(Img1, Img2, defaultPara.Fdir, ... |
---|
68 | ['Dense_' num2str(defaultPara.AbsThre) '_' num2str(defaultPara.RatioThre)], 1, 1); |
---|
69 | end |
---|
70 | |
---|
71 | % 2) load OccluSurfMatches |
---|
72 | [f1, f2, OccluedMatches] = readSurfMatches(Img1, Img2, defaultPara.Fdir, [ defaultPara.Type 'OccluDense'], 1, 1, 3);% need UniqueRatio//Min jobs |
---|
73 | UniqueRatio = OccluedMatches(3,:); |
---|
74 | %UniqueRatio = min(UniqueRatio,defaultPara.MaxUniqueRatio); |
---|
75 | Ptr = UniqueRatio > defaultPara.MaxUniqueRatio; |
---|
76 | UniqueRatio(Ptr) = Inf;%defaultPara.MaxUniqueRatio; |
---|
77 | |
---|
78 | OccluedMatches = OccluedMatches(1:2,:); |
---|
79 | if false |
---|
80 | disp('use CleanMatches'); |
---|
81 | matches = Matches; |
---|
82 | % removing initial matches that inconsistent with OccluedMatches |
---|
83 | [c i] = intersect( matches(1,:), OccluedMatches(1,:)); |
---|
84 | matches(:,i) = []; |
---|
85 | [c i] = intersect( matches(2,:), OccluedMatches(2,:)); |
---|
86 | matches(:,i) = []; |
---|
87 | matches = [matches OccluedMatches]; |
---|
88 | |
---|
89 | else |
---|
90 | disp('Number of Occlumatches used') |
---|
91 | size( OccluedMatches,2) |
---|
92 | |
---|
93 | % removing initial matches that inconsistent with OccluedMatches |
---|
94 | [c i] = intersect( matches(1,:), OccluedMatches(1,:)); |
---|
95 | matches(:,i) = []; |
---|
96 | [c i] = intersect( matches(2,:), OccluedMatches(2,:)); |
---|
97 | matches(:,i) = []; |
---|
98 | |
---|
99 | NumInitialMatches = size(matches, 2); |
---|
100 | matches = [matches OccluedMatches]; |
---|
101 | if isempty(matches) |
---|
102 | disp('Zeros Surf matches'); |
---|
103 | fail = 1; |
---|
104 | return; |
---|
105 | end |
---|
106 | |
---|
107 | % 3) Construct Prior Dist for All matches |
---|
108 | NMatches = size(matches, 2); |
---|
109 | PriorDist = ones(1, NMatches); |
---|
110 | PriorDist( (NumInitialMatches+1):end) = UniqueRatio; |
---|
111 | % 4) RANSAC |
---|
112 | disp('Number of matches used') |
---|
113 | |
---|
114 | if defaultPara.Flag.StorageDataBeforeRansac |
---|
115 | disp('Storage Data Before Ransac'); |
---|
116 | save([defaultPara.Fdir '/data/DataBeforeRansac.mat']); |
---|
117 | % return; |
---|
118 | end |
---|
119 | |
---|
120 | % Ensemble method to determine confidence of inliers |
---|
121 | fittingfn = @fundmatrix; |
---|
122 | distfnEnsmble = @fundistEnsmble; |
---|
123 | degenfn = @isdegenerate; |
---|
124 | x = [[f1(:, matches(1,:)); ones(1, NMatches)]; [f2(:, matches(2,:)); ones(1, NMatches)]]; |
---|
125 | [ SampsonDist ] = EnsembleRansac(defaultPara, x, fittingfn, distfnEnsmble, degenfn, 8, PriorDist', min(NMatches*10, defaultPara.MAXEnsembleSamples), 0); |
---|
126 | kurtosisValue =kurtosis(SampsonDist'); |
---|
127 | |
---|
128 | [F0, inliers, NewDist, fail, ind]=GeneralRansac(defaultPara, f1, f2, matches, [], [], kurtosisValue', 8); |
---|
129 | figure(100); plotmatches(I1,I2,f1, f2,matches(:,inliers), 'Stacking', 'h', 'Interactive', 3); |
---|
130 | matches = matches(:,inliers); % accept the pruning result |
---|
131 | if isempty(matches) |
---|
132 | disp('Zeros After Ransac matches'); |
---|
133 | fail = 2; |
---|
134 | return; |
---|
135 | end |
---|
136 | end |
---|
137 | % ------------------------------------------------------------------------------------ |
---|
138 | |
---|
139 | |
---|
140 | % Apply Bundle Adjustment Refinment Algorithm to Prune the Matches further ----------- |
---|
141 | % And Estimated the Pose |
---|
142 | % 1) initialize the 3D position of the matches given Prior Pose and Depths |
---|
143 | x_calib = [ inv(defaultPara.InrinsicK1)*[ f1(:,matches(1,:)); ones(1,size(matches,2))];... |
---|
144 | inv(defaultPara.InrinsicK2)*[ f2(:,matches(2,:)); ones(1,size(matches,2))]]; |
---|
145 | |
---|
146 | % Estimate F using NonLine LS on every inlier |
---|
147 | MatchDensityWeights1 = CalMatchDensityWeights(f1(:,matches(1,:)), max(size(I1))/defaultPara.radius2imageSizeRatio); |
---|
148 | MatchDensityWeights2 = CalMatchDensityWeights(f2(:,matches(2,:)), max(size(I2))/defaultPara.radius2imageSizeRatio); |
---|
149 | MatchDensityWeights =mean([MatchDensityWeights1; MatchDensityWeights2], 1); |
---|
150 | F = getFnpt( F0, f1(:, matches(1,:))', f2(:, matches(2,:))', MatchDensityWeights); |
---|
151 | E = defaultPara.InrinsicK2'*F*defaultPara.InrinsicK1; % Camera essential Matrix |
---|
152 | if ~isempty(PriorPose) |
---|
153 | [ R0, T0, lamda1, lamda2, inlier, Error] = EstPose( defaultPara, E, x_calib, [], PriorPose.R(1:3,:)); |
---|
154 | else |
---|
155 | [ R0, T0, lamda1, lamda2, inlier, Error] = EstPose( defaultPara, E, x_calib, [], []); |
---|
156 | end |
---|
157 | T0 = [T0; - R0'*T0]; |
---|
158 | R0 = [R0; R0']; |
---|
159 | matches = matches(:,inlier); % delet matches give negative depths |
---|
160 | x_calib = x_calib(:,inlier); |
---|
161 | lamda1 = lamda1(inlier); |
---|
162 | lamda2 = lamda2(inlier); |
---|
163 | |
---|
164 | X_obj_1 = x_calib(1:3,:).*repmat(lamda1, 3, 1); |
---|
165 | X_obj_2 = R0(4:6,:)*(x_calib(4:6,:).*repmat(lamda2, 3, 1)) + repmat(T0(4:6), 1, size(matches,2)); |
---|
166 | X_obj = (X_obj_1+X_obj_2)/2; |
---|
167 | |
---|
168 | % 2) |
---|
169 | [R T X_obj_BA X_im_BA dist1_BA dist2_BA]=SparseBAWraper(defaultPara, R0(1:3,:), T0(1:3), [f1(:,matches(1,:)); f2(:,matches(2,:))], X_obj, [ ImgInfo1 ImgInfo2], 1); |
---|
170 | if false % Min Modified for not pruning using BundleAdjustment |
---|
171 | if all(isnan( dist1_BA)) || isempty(R) || any(isnan(R(:))) |
---|
172 | disp('BA failed'); |
---|
173 | fail = 3; |
---|
174 | return; |
---|
175 | end |
---|
176 | while length(X_im_BA) >= defaultPara.MinimumNumMatches |
---|
177 | disp('BundleAdjClean') |
---|
178 | outlier_thre1 = prctile(dist1_BA,90); |
---|
179 | outlier_thre2 = prctile(dist2_BA,90); |
---|
180 | if outlier_thre1 >= defaultPara.ReProjErrorThre || outlier_thre2 >= defaultPara.ReProjErrorThre |
---|
181 | Outlier = dist1_BA > outlier_thre1 | dist2_BA > outlier_thre2; |
---|
182 | matches(:,Outlier) = []; |
---|
183 | if isempty(matches) |
---|
184 | disp('Zeros After BA matches'); |
---|
185 | fail = 4; |
---|
186 | return; |
---|
187 | end |
---|
188 | lamda1(Outlier) = []; |
---|
189 | lamda2(Outlier) = []; |
---|
190 | X_obj_BA(:,Outlier) = []; |
---|
191 | x_calib(:,Outlier) = []; |
---|
192 | [R T X_obj_BA X_im_BA dist1_BA dist2_BA]=SparseBAWraper(defaultPara, R, T, [f1(:,matches(1,:)); f2(:,matches(2,:))], X_obj_BA, [ ImgInfo1 ImgInfo2], 1); |
---|
193 | if all(isnan( dist1_BA)) || isempty(R) || any(isnan(R(:))) |
---|
194 | disp('BA failed'); |
---|
195 | fail = 5; |
---|
196 | return; |
---|
197 | end |
---|
198 | else |
---|
199 | break; |
---|
200 | end |
---|
201 | end |
---|
202 | end |
---|
203 | % ------------------------------------------------------------------------------------ |
---|
204 | |
---|
205 | |
---|
206 | % Triangulation ---------------------------------------------------------------------- |
---|
207 | % modified the x_calib |
---|
208 | tempf1 = X_im_BA(1:2,:); |
---|
209 | tempf2 = X_im_BA(3:4,:); |
---|
210 | x_calib = [ inv(defaultPara.InrinsicK1)*[ tempf1; ones(1,length(tempf1))];... |
---|
211 | inv(defaultPara.InrinsicK2)*[ tempf2; ones(1,length(tempf2))]]; |
---|
212 | [ lamda1 lamda2 Error] = triangulation( defaultPara, R, T, x_calib); |
---|
213 | % Clean outlier triangulated depth |
---|
214 | %LamdaOutlier = lamda1 > 1000 | lamda1 <1; |
---|
215 | %matches(:,LamdaOutlier) = []; |
---|
216 | %if isempty(matches) |
---|
217 | % disp('Zeros After Tri matches'); |
---|
218 | % fail = 6; |
---|
219 | % return; |
---|
220 | %end |
---|
221 | |
---|
222 | %lamda1(LamdaOutlier) = []; |
---|
223 | %lamda2(LamdaOutlier) = []; |
---|
224 | %x_calib(:,LamdaOutlier) = []; |
---|
225 | |
---|
226 | % Pair Image Depth Scale |
---|
227 | [D1 IND1] = PorjPosi2Depth(size(I1), size(ImgInfo1.Model.Depth.FitDepth), f1(:,matches(1,:)), ImgInfo1.Model.Depth.FitDepth); |
---|
228 | [D2 IND2] = PorjPosi2Depth(size(I2), size(ImgInfo2.Model.Depth.FitDepth), f2(:,matches(2,:)), ImgInfo2.Model.Depth.FitDepth); |
---|
229 | Depth1ProjDepthRatio = sqrt(sum(x_calib(1:3,:).^2, 1)); |
---|
230 | Depth2ProjDepthRatio = sqrt(sum(x_calib(4:6,:).^2, 1)); |
---|
231 | DProj1 = D1./Depth1ProjDepthRatio; |
---|
232 | DProj2 = D2./Depth2ProjDepthRatio; |
---|
233 | [DepthScale1] = UniformDepthScale( defaultPara, DProj1, lamda1, ones(1,length(lamda1))); |
---|
234 | [DepthScale2] = UniformDepthScale( defaultPara, DProj2, lamda2, ones(1,length(lamda2)) ); |
---|
235 | %if DepthScale1 > 20 | DepthScale1 <0.05 | DepthScale2 > 20 | DepthScale2 <0.05 %//Min used to use 10 and 0.2 |
---|
236 | % disp('Unrealistic in Rescaleing the depth, Check matchings'); |
---|
237 | % fail = 6; |
---|
238 | % return; |
---|
239 | %end |
---|
240 | |
---|
241 | Pair.lamda = [lamda1; lamda2];%//Min add for debug |
---|
242 | Pair.DepthScale = [DepthScale1; DepthScale2]; |
---|
243 | Pair.R = R; |
---|
244 | Pair.T = T; |
---|
245 | Pair.Xim = [f1(:,matches(1,:)); f2(:,matches(2,:))]; |
---|
246 | % ----------------------------------------------------------------------------------- |
---|