-
Notifications
You must be signed in to change notification settings - Fork 13
/
Copy pathmain.m
277 lines (243 loc) · 13 KB
/
main.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
%% parameters of user setting %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% opts.path: directory of saved mat files
% opts.load_all(0/1):
% 0: only load the files specified in mat_files structure
% 1: load all the files from the opts.path
% opts.datatype: 0: experiment data
% 1: simulation data
% opts.show_results(0/1): plot the calibrated results on figure
clear all, clc
%
% mat_path = 'F:\Calibration\matlab_utils\general';
% addpath("F:\Calibration\convex_relaxation_sim3");
% opts.path = "F:\Calibration\intrinsic_lidar_calibration\intrinsic_calibration_mat\robotics_building\tetrahedron4\";
% v_path = "F:\Calibration\intrinsic_lidar_calibration\intrinsic_calibration_mat\robotics_building\validation\t4\";
% opts.save_path = ".\results\tetrahedron4\Lie\";
mat_path = '/home/brucebot/workspace/matlab_utils/general';
addpath(genpath("/home/brucebot/workspace/test/CVPR17"));
opts.path = "/home/brucebot/workspace/lc-calibration/LiDAR_intrinsic_calibration/tetrehedron_data/tetrahedron4/";
v_path = "/home/brucebot/workspace/lc-calibration/LiDAR_intrinsic_calibration/tetrehedron_data/validation/t4/";
opts.save_path = "./paper_results/tetrahedron4/Lie/";
addpath(mat_path);
%%
opts.load_all = 1;
opt_datatype = ["Experiment", "Simulation"];
opts.datatype = 1;
opts.show_results = 1;
opts.estimatePlaneMean = 0;
opts.calibrateMean = 0;
opts.fixXY = 0;
opt_noise_type = ["3_param", "6_param", "sim3"];
opts.noise_type = 3;
opts.noise_level = 0;
opt_formulation = ["Lie","BaseLine1","BaseLine2","BaseLine3"]; % Lie group, 3 params, 6 params, 8 params
opts.method = 1; % "Lie","BaseLine1","BaseLine2", "BaseLine3"
opts.iterative = 0;
opts.num_beams = 32;
opts.num_scans = 1;
opts.num_iters = 200; % user defined iterations
opts.threshold = 0;
%% Load datasets
clc; % DO NOT add 'clear' here
if exist('pc','var') && exist('data', 'var') && exist('mat_files', 'var')
% if exist('pc','var')
disp("Data have been loaded, it may take a while to reload them.")
prompt = "Are you sure you want to reload them again? [Y] [N]";
urs_ans = input(prompt, 's');
if contains(urs_ans, 'N','IgnoreCase', true)
disp('Reload datasets cancelled')
return
else
clear
disp("ALL varialbes cleared! ")
disp("Keep reloading datasets...")
end
end
% Single variable called point_cloud
% path = "/home/chenxif/Documents/me590/Calibration/IntrinsicCalibration/extracted_tags/";
disp("Loading names of data sets...")
if ~opts.load_all
mat_files = struct('file_name', {opts.path+'velodyne_points-Intrinsic-LargeTag--2019-11-21-22-04.mat';...
opts.path+'velodyne_points-Intrinsic-SmallTag--2019-11-21-22-00.mat';...
opts.path+'velodyne_points-Intrinsic4-SmallTag--2019-11-22-22-54.mat';...
opts.path+'velodyne_points-Intrinsic5-LargeTag--2019-11-22-23-02.mat';...
opts.path+'velodyne_points-Intrinsic5-SmallTag--2019-11-22-23-00.mat';...
opts.path+'velodyne_points-Intrinsic-further-LargeTag--2019-11-22-23-05.mat';...
opts.path+'velodyne_points-Intrinsic-further-SmallTag--2019-11-22-23-09.mat';...
opts.path+'velodyne_points-Intrinsic-further2-LargeTag--2019-11-22-23-15.mat';...
opts.path+'velodyne_points-Intrinsic-further2-SmallTag--2019-11-22-23-17.mat';...
opts.path+'velodyne_points-upper1-SmallTag--2019-12-05-20-13.mat';...
opts.path+'velodyne_points-upper2-SmallTag--2019-12-05-20-16.mat';...
opts.path+'velodyne_points-upper3-SmallTag--2019-12-05-20-19.mat';...
opts.path+'velodyne_points-upper4-SmallTag--2019-12-05-20-22.mat';...
opts.path+'velodyne_points-upper5-SmallTag--2019-12-05-20-23.mat';...
opts.path+'velodyne_points-upper6-SmallTag--2019-12-05-20-26.mat';...
opts.path+'velodyne_points-upper7-SmallTag--2019-12-05-20-29.mat';...
opts.path+'velodyne_points-upper8-SmallTag--2019-12-05-20-29.mat'});
for file = 1:length(mat_files)
mat_files(file).tag_size = identifyTagSizeFromAName(mat_files(file).file_name);
end
else
mat_files = loadFilesFromAFolder(opts.path, '*.mat');
end
num_targets = length(mat_files);
disp("Loading point cloud from .mat files")
pc = struct('point_cloud', cell(1,num_targets));
for t = 1:num_targets
pc(t).point_cloud = loadPointCloud(mat_files(t).file_name);
end
disp("Pre-processing payload points...")
data = struct('point_cloud', cell(1,num_targets), 'tag_size', cell(1,num_targets));% XYZIR
for t = 1:num_targets
for iter = 1: opts.num_scans
data(t).scan(iter).payload_points = getPayload(pc(t).point_cloud, iter);
data(t).tag_size = mat_files(t).tag_size;
end
end
disp("Done loading data!")
%% post process data, remove outliers with prior knowledge
for iter = 1: opts.num_scans
%filter out outliers in tetrahedron1
% data(1).scan(iter).payload_points = data(1).scan(iter).payload_points(:, ~(data(1).scan(iter).payload_points(5,:)== 6 & data(1).scan(iter).payload_points(1,:) > 4.22));
% for n = 7:8
% data(1).scan(iter).payload_points = data(1).scan(iter).payload_points(:,~(data(1).scan(iter).payload_points(5,:)== n & data(1).scan(iter).payload_points(1,:)> 4.3));
% end
% for n = 12
% data(1).scan(iter).payload_points = data(1).scan(iter).payload_points(:,~(data(1).scan(iter).payload_points(5,:)== n & data(1).scan(iter).payload_points(1,:)> 4.4));
% end
% for n = 9:11
% data(1).scan(iter).payload_points = data(1).scan(iter).payload_points(:,~(data(1).scan(iter).payload_points(5,:)== n & data(1).scan(iter).payload_points(1,:)> 4.38));
% end
% data(2).scan(iter).payload_points = data(2).scan(iter).payload_points(:,~(data(2).scan(iter).payload_points(5,:)== 7 & data(2).scan(iter).payload_points(1,:)> 4.1));
% data(2).scan(iter).payload_points = data(2).scan(iter).payload_points(:,~(data(2).scan(iter).payload_points(5,:)== 8 & data(2).scan(iter).payload_points(1,:)> 4));
% for n = 6:7
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 1.14));
% end
% for n = 8:10
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 1.11));
% end
% for n = 11:14
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 1.067));
% end
% for n = 15:17
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 1.02));
% end
% for n = 18:20
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 0.99));
% end
% for n = 21:22
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 0.96));
% end
% for n = 23:25
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 0.93));
% end
% for n = 26:28
% data(4).scan(iter).payload_points = data(4).scan(iter).payload_points(:,~(data(4).scan(iter).payload_points(5,:)== n & data(4).scan(iter).payload_points(2,:)> 0.87));
% end
%filter out outliers in tetrahedron2
% for i = 1:4
% data(i).scan(iter).payload_points = data(i).scan(iter).payload_points(:,~(data(i).scan(iter).payload_points(5,:)== 0));
% end
%filter by intensity
% for i = 1:1
% data(i).scan(iter).payload_points = data(i).scan(iter).payload_points(:,~(data(i).scan(iter).payload_points(4,:) > 50));
% end
end
%% Optimize intrinsic parameters
clc
%target 16 and 25 have some problems
% pick_targets = [ 26 28 32 36];
% normal CN 3.0612
% base CN 5.8373
% num_targets = length(pick_targets);
% pick_data = data(pick_targets);
[calib_param, plane_original, plane, distance_original, distance, ...
original_points, updated_points] = intrinsicCalibration(...
opts, data, num_targets);
if ~exist(opts.save_path, 'dir')
mkdir(opts.save_path)
end
final_parameters = computeCalibParameters(calib_param, opts.num_beams, opts.method);
save(opts.save_path + "realExpDelta.mat", 'calib_param');
%% Show graphical results
% if opts.show_results
% disp("Now plotting....")
plotCalibratedResults(num_targets, opts.num_scans, ...
plane_original, original_points);
plotCalibratedResults(num_targets, opts.num_scans, ...
plane, updated_points);
% disp("Done plotting!")
%
% end
%%
for i = 1:4
object = i;
checkRingOrderWithOriginalPerTarget(original_points, updated_points, object, opts.num_beams, opts.num_scans, 'display');
end
%% show numerical results
disp("Showing numerical results...")
disp("Showing current estimate")
results = struct('ring', {distance(end).ring(:).ring}, ...
'num_points', {distance(end).ring(:).num_points}, ...
'mean_original', {distance_original.ring(:).mean}, ...
'mean_calibrated', {distance(end).ring(:).mean}, ...
'mean_percentage', num2cell((abs([distance_original.ring(:).mean]) - abs([distance(end).ring(:).mean])) ./ abs([distance_original.ring(:).mean])), ...
'mean_diff', num2cell([distance_original.ring(:).mean] - [distance(end).ring(:).mean]), ...
'mean_diff_in_mm', num2cell(([distance_original.ring(:).mean] - [distance(end).ring(:).mean]) * 1e3), ...
'std_original', {distance_original.ring(:).std}, ...
'std_calibrated', {distance(end).ring(:).std}, ...
'std_diff', num2cell([distance_original.ring(:).std] - [distance(end).ring(:).std]), ...
'std_diff_in_mm', num2cell(([distance_original.ring(:).std] - [distance(end).ring(:).std])* 1e3));
% results = struct('mean_original', {distance_original.ring(:).mean});
struct2table(distance(end).ring(:))
disp("Showing comparison")
struct2table(results)
% check if ring mis-ordered
disp("If the rings are mis-ordered...")
[order mis_ordered_list] = checkRingOrder(updated_points, ...
final_parameters, num_targets, opts.num_beams, opts);
save(opts.save_path + "training_results.mat", "results");
%% Validation
validation_mat_files = loadFilesFromAFolder(v_path, '*.mat');
v_num_targets = length(validation_mat_files);
v_pc = struct('point_cloud', cell(1,v_num_targets));
for t = 1:v_num_targets
v_pc(t).point_cloud = loadPointCloud(validation_mat_files(t).file_name);
end
v_data = struct('point_cloud', cell(1,v_num_targets), ...
'tag_size', cell(1,v_num_targets));% XYZIR
for t = 1:v_num_targets
for iter = 1: opts.num_scans
v_data(t).scan(iter).payload_points = getPayload(v_pc(t).point_cloud, iter);
v_data(t).tag_size = validation_mat_files(t).tag_size;
% v_data(t).scan(iter).payload_points = v_data(t).scan(iter).payload_points(:,~(v_data(t).scan(iter).payload_points(4,:)> 50));
% v_data(t).scan(iter).payload_points = v_data(t).scan(iter).payload_points(:,~(v_data(t).scan(iter).payload_points(5,:)== 0));
end
end
% validation_data = parseValidationbag(v_path, "*.bag", 1, opts.num_scans);
[validation_results] = validation(v_data, final_parameters, opts);
summary1 = struct( 'mean_original', mean([validation_results.mean_original]), ...
'mean_calibrated', mean([validation_results.mean_calibrated]), ...
'mean_percentage', (mean([validation_results.mean_original])- mean([validation_results.mean_calibrated]))/mean([validation_results.mean_original]), ...
'std_original', mean([validation_results.std_original]), ...
'std_calibrated', mean([validation_results.std_calibrated]),...
'thickness_original',mean([validation_results.thickness_original]),...
'thickness_calibrated',mean([validation_results.thickness_calibrated]));
struct2table(summary1)
std_mean = mean([validation_results.std_original]);
for k = 1:3
threshold = k*std_mean;
index = [validation_results.mean_original] > threshold;
mean_original = [validation_results.mean_original];
mean_calibrated = [validation_results.mean_calibrated];
std_original = [validation_results.std_original];
std_calibrated = [validation_results.std_calibrated];
summary2 = struct( 'mean_original', mean(mean_original(index)), ...
'mean_calibrated', mean(mean_calibrated(index)), ...
'mean_percentage', (mean(mean_original(index))- mean(mean_calibrated(index)))/mean(mean_original(index)), ...
'std_original', mean(std_original(index)), ...
'std_calibrated', mean(std_calibrated(index)));
struct2table(summary2)
end
disp("All processes has finished")