Make fragments

The first step of the scene reconstruction system is to create fragments from short RGBD sequences.

Input arguments

The script runs with python run_system.py [config] --make. In [config], ["path_dataset"] should have subfolders image and depth to store the color images and depth images respectively. We assume the color images and the depth images are synchronized and registered. In [config], the optional argument ["path_intrinsic"] specifies the path to a json file that stores the camera intrinsic matrix (See Read camera intrinsic for details). If it is not given, the PrimeSense factory setting is used instead.

Register RGBD image pairs

33# examples/Python/ReconstructionSystem/make_fragments.py
34def register_one_rgbd_pair(s, t, color_files, depth_files, intrinsic,
35                           with_opencv, config):
36    source_rgbd_image = read_rgbd_image(color_files[s], depth_files[s], True,
37                                        config)
38    target_rgbd_image = read_rgbd_image(color_files[t], depth_files[t], True,
39                                        config)
40
41    option = o3d.odometry.OdometryOption()
42    option.max_depth_diff = config["max_depth_diff"]
43    if abs(s - t) is not 1:
44        if with_opencv:
45            success_5pt, odo_init = pose_estimation(source_rgbd_image,
46                                                    target_rgbd_image,
47                                                    intrinsic, False)
48            if success_5pt:
49                [success, trans, info] = o3d.odometry.compute_rgbd_odometry(
50                    source_rgbd_image, target_rgbd_image, intrinsic, odo_init,
51                    o3d.odometry.RGBDOdometryJacobianFromHybridTerm(), option)
52                return [success, trans, info]
53        return [False, np.identity(4), np.identity(6)]
54    else:
55        odo_init = np.identity(4)
56        [success, trans, info] = o3d.odometry.compute_rgbd_odometry(
57            source_rgbd_image, target_rgbd_image, intrinsic, odo_init,
58            o3d.odometry.RGBDOdometryJacobianFromHybridTerm(), option)
59        return [success, trans, info]

The function reads a pair of RGBD images and registers the source_rgbd_image to the target_rgbd_image. Open3D function compute_rgbd_odometry is called to align the RGBD images. For adjacent RGBD images, an identity matrix is used as initialization. For non-adjacent RGBD images, wide baseline matching is used as an initialization. In particular, function pose_estimation computes OpenCV ORB feature to match sparse features over wide baseline images, then performs 5-point RANSAC to estimate a rough alignment, which is used as the initialization of compute_rgbd_odometry.

Multiway registration

 61# examples/Python/ReconstructionSystem/make_fragments.py
 62def make_posegraph_for_fragment(path_dataset, sid, eid, color_files,
 63                                depth_files, fragment_id, n_fragments,
 64                                intrinsic, with_opencv, config):
 65    o3d.utility.set_verbosity_level(o3d.utility.VerbosityLevel.Error)
 66    pose_graph = o3d.registration.PoseGraph()
 67    trans_odometry = np.identity(4)
 68    pose_graph.nodes.append(o3d.registration.PoseGraphNode(trans_odometry))
 69    for s in range(sid, eid):
 70        for t in range(s + 1, eid):
 71            # odometry
 72            if t == s + 1:
 73                print(
 74                    "Fragment %03d / %03d :: RGBD matching between frame : %d and %d"
 75                    % (fragment_id, n_fragments - 1, s, t))
 76                [success, trans,
 77                 info] = register_one_rgbd_pair(s, t, color_files, depth_files,
 78                                                intrinsic, with_opencv, config)
 79                trans_odometry = np.dot(trans, trans_odometry)
 80                trans_odometry_inv = np.linalg.inv(trans_odometry)
 81                pose_graph.nodes.append(
 82                    o3d.registration.PoseGraphNode(trans_odometry_inv))
 83                pose_graph.edges.append(
 84                    o3d.registration.PoseGraphEdge(s - sid,
 85                                                   t - sid,
 86                                                   trans,
 87                                                   info,
 88                                                   uncertain=False))
 89
 90            # keyframe loop closure
 91            if s % config['n_keyframes_per_n_frame'] == 0 \
 92                    and t % config['n_keyframes_per_n_frame'] == 0:
 93                print(
 94                    "Fragment %03d / %03d :: RGBD matching between frame : %d and %d"
 95                    % (fragment_id, n_fragments - 1, s, t))
 96                [success, trans,
 97                 info] = register_one_rgbd_pair(s, t, color_files, depth_files,
 98                                                intrinsic, with_opencv, config)
 99                if success:
100                    pose_graph.edges.append(
101                        o3d.registration.PoseGraphEdge(s - sid,
102                                                       t - sid,
103                                                       trans,
104                                                       info,
105                                                       uncertain=True))
106    o3d.io.write_pose_graph(
107        join(path_dataset, config["template_fragment_posegraph"] % fragment_id),
108        pose_graph)

This script uses the technique demonstrated in Multiway registration. Function make_posegraph_for_fragment builds a pose graph for multiway registration of all RGBD images in this sequence. Each graph node represents an RGBD image and its pose which transforms the geometry to the global fragment space. For efficiency, only key frames are used.

Once a pose graph is created, multiway registration is performed by calling function optimize_posegraph_for_fragment.

12# examples/Python/ReconstructionSystem/optimize_posegraph.py
13def run_posegraph_optimization(pose_graph_name, pose_graph_optimized_name,
14                               max_correspondence_distance,
15                               preference_loop_closure):
16    # to display messages from o3d.registration.global_optimization
17    o3d.utility.set_verbosity_level(o3d.utility.VerbosityLevel.Debug)
18    method = o3d.registration.GlobalOptimizationLevenbergMarquardt()
19    criteria = o3d.registration.GlobalOptimizationConvergenceCriteria()
20    option = o3d.registration.GlobalOptimizationOption(
21        max_correspondence_distance=max_correspondence_distance,
22        edge_prune_threshold=0.25,
23        preference_loop_closure=preference_loop_closure,
24        reference_node=0)
25    pose_graph = o3d.io.read_pose_graph(pose_graph_name)
26    o3d.registration.global_optimization(pose_graph, method, criteria, option)
27    o3d.io.write_pose_graph(pose_graph_optimized_name, pose_graph)
28    o3d.utility.set_verbosity_level(o3d.utility.VerbosityLevel.Error)
29
30
31def optimize_posegraph_for_fragment(path_dataset, fragment_id, config):
32    pose_graph_name = join(path_dataset,
33                           config["template_fragment_posegraph"] % fragment_id)
34    pose_graph_optimized_name = join(
35        path_dataset,
36        config["template_fragment_posegraph_optimized"] % fragment_id)
37    run_posegraph_optimization(pose_graph_name, pose_graph_optimized_name,
38            max_correspondence_distance = config["max_depth_diff"],
39            preference_loop_closure = \
40            config["preference_loop_closure_odometry"])

This function calls global_optimization to estimate poses of the RGBD images.

Make a fragment

110# examples/Python/ReconstructionSystem/make_fragments.py
111def integrate_rgb_frames_for_fragment(color_files, depth_files, fragment_id,
112                                      n_fragments, pose_graph_name, intrinsic,
113                                      config):
114    pose_graph = o3d.io.read_pose_graph(pose_graph_name)
115    volume = o3d.integration.ScalableTSDFVolume(
116        voxel_length=config["tsdf_cubic_size"] / 512.0,
117        sdf_trunc=0.04,
118        color_type=o3d.integration.TSDFVolumeColorType.RGB8)
119    for i in range(len(pose_graph.nodes)):
120        i_abs = fragment_id * config['n_frames_per_fragment'] + i
121        print(
122            "Fragment %03d / %03d :: integrate rgbd frame %d (%d of %d)." %
123            (fragment_id, n_fragments - 1, i_abs, i + 1, len(pose_graph.nodes)))
124        rgbd = read_rgbd_image(color_files[i_abs], depth_files[i_abs], False,
125                               config)
126        pose = pose_graph.nodes[i].pose
127        volume.integrate(rgbd, intrinsic, np.linalg.inv(pose))
128    mesh = volume.extract_triangle_mesh()
129    mesh.compute_vertex_normals()
130    return mesh
131
132
133def make_pointcloud_for_fragment(path_dataset, color_files, depth_files,
134                                 fragment_id, n_fragments, intrinsic, config):
135    mesh = integrate_rgb_frames_for_fragment(
136        color_files, depth_files, fragment_id, n_fragments,
137        join(path_dataset,
138             config["template_fragment_posegraph_optimized"] % fragment_id),
139        intrinsic, config)
140    pcd = o3d.geometry.PointCloud()
141    pcd.points = mesh.vertices
142    pcd.colors = mesh.vertex_colors
143    pcd_name = join(path_dataset,
144                    config["template_fragment_pointcloud"] % fragment_id)
145    o3d.io.write_point_cloud(pcd_name, pcd, False, True)

Once the poses are estimates, RGBD integration is used to reconstruct a colored fragment from each RGBD sequence.

Batch processing

167# examples/Python/ReconstructionSystem/make_fragments.py
168def run(config):
169    print("making fragments from RGBD sequence.")
170    make_clean_folder(join(config["path_dataset"], config["folder_fragment"]))
171    [color_files, depth_files] = get_rgbd_file_lists(config["path_dataset"])
172    n_files = len(color_files)
173    n_fragments = int(math.ceil(float(n_files) / \
174            config['n_frames_per_fragment']))
175
176    if config["python_multi_threading"]:
177        from joblib import Parallel, delayed
178        import multiprocessing
179        import subprocess
180        MAX_THREAD = min(multiprocessing.cpu_count(), n_fragments)
181        Parallel(n_jobs=MAX_THREAD)(delayed(process_single_fragment)(
182            fragment_id, color_files, depth_files, n_files, n_fragments, config)
183                                    for fragment_id in range(n_fragments))
184    else:
185        for fragment_id in range(n_fragments):
186            process_single_fragment(fragment_id, color_files, depth_files,
187                                    n_files, n_fragments, config)

The main function calls each individual function explained above.

Results

Fragment 000 / 013 :: RGBD matching between frame : 0 and 1
Fragment 000 / 013 :: RGBD matching between frame : 0 and 5
Fragment 000 / 013 :: RGBD matching between frame : 0 and 10
Fragment 000 / 013 :: RGBD matching between frame : 0 and 15
Fragment 000 / 013 :: RGBD matching between frame : 0 and 20
:
Fragment 000 / 013 :: RGBD matching between frame : 95 and 96
Fragment 000 / 013 :: RGBD matching between frame : 96 and 97
Fragment 000 / 013 :: RGBD matching between frame : 97 and 98
Fragment 000 / 013 :: RGBD matching between frame : 98 and 99

The following is a log from optimize_posegraph_for_fragment.

[GlobalOptimizationLM] Optimizing PoseGraph having 100 nodes and 195 edges.
Line process weight : 389.309502
[Initial     ] residual : 3.223357e+05, lambda : 1.771814e+02
[Iteration 00] residual : 1.721845e+04, valid edges : 157, time : 0.022 sec.
[Iteration 01] residual : 1.350251e+04, valid edges : 168, time : 0.017 sec.
:
[Iteration 32] residual : 9.779118e+03, valid edges : 179, time : 0.013 sec.
Current_residual - new_residual < 1.000000e-06 * current_residual
[GlobalOptimizationLM] total time : 0.519 sec.
[GlobalOptimizationLM] Optimizing PoseGraph having 100 nodes and 179 edges.
Line process weight : 398.292104
[Initial     ] residual : 5.120047e+03, lambda : 2.565362e+02
[Iteration 00] residual : 5.064539e+03, valid edges : 179, time : 0.014 sec.
[Iteration 01] residual : 5.037665e+03, valid edges : 178, time : 0.015 sec.
:
[Iteration 11] residual : 5.017307e+03, valid edges : 177, time : 0.013 sec.
Current_residual - new_residual < 1.000000e-06 * current_residual
[GlobalOptimizationLM] total time : 0.197 sec.
CompensateReferencePoseGraphNode : reference : 0

The following is a log from integrate_rgb_frames_for_fragment.

Fragment 000 / 013 :: integrate rgbd frame 0 (1 of 100).
Fragment 000 / 013 :: integrate rgbd frame 1 (2 of 100).
Fragment 000 / 013 :: integrate rgbd frame 2 (3 of 100).
:
Fragment 000 / 013 :: integrate rgbd frame 97 (98 of 100).
Fragment 000 / 013 :: integrate rgbd frame 98 (99 of 100).
Fragment 000 / 013 :: integrate rgbd frame 99 (100 of 100).

The following images show some of the fragments made by this script.

../../_images/fragment_0.png ../../_images/fragment_1.png ../../_images/fragment_2.png ../../_images/fragment_3.png