Figure 1. Aerial video dataset and distribution of ground control points(GCPs) in regular scene.
We collect an aerial dataset which consists of two aerial video sequences captured using the DJI M300 RTK 231 drone with the DJI P1 camera. Figure 1 illustrates the two sequences: one with a regular 232 strip configuration and the other with an irregular configuration. These videos were 233 recorded at the Informatics Department of Wuhan University at an altitude of 200 m. The 234 recording frequency was set at 60 frames per second (fps), with a resolution of 3860x2160 235 pixels. The average ground resolution achieved was 0.03 m.
videos-and-keyframes file link includes collected videos and corresponding keyframes. The structure is as follows:
--irregular:
---- video
---- keyframes:
------ 1.jpg
------ 2.jpg
...
------ 368.jpg
--regular:
---- video
---- keyframes:
------ 1.jpg
------ 2.jpg
...
------ 914.jpg
camera-calibration file link includes a video of apriltag and the camera intrinsics of raw video and 4x downsampled.
Figure 1. Aerial video dataset and distribution of ground control points(GCPs) in regular scene.
Figure 2. Aerial video dataset and distribution of ground control points(GCPs) in irregular scene.
We also collected 16 GCPs that are evenly distributed throughout the dataset 243 area. Some of these GCPs were utilized to compute a high-accuracy trajectory, while the 244 remainder served as checkpoints to assess the accuracy of the trajectory. The GCPs were 245 measured using a high-accuracy GPS receiver and processed to achieve a localization 246 accuracy of 9.0 mm.
ground_truth file link includes the observations of GCPs from keyframes, groundtruth poses of both scenes and the coordinates of GCPs. The structure is as follows:
--irregular:
---- keyframe_GCP_obsevations.txt
---- ground_truth.txt
--regular:
---- keyframe_GCP_obsevations.txt
---- ground_truth.txt
--points.txt
points.txt stores the coordinates of GCPs, which is organized as follows:
GCP_id X Y Z
ground_truth.txt stores the extrinsics of keyframes, which is organized as follows:
timestamp tx ty tz qx qy qz qw
The i-th row represents the extrinsics of the keyframe i.jpg.
keyframe_GCP_obsevations.txt stores the observations of GCPs from keyframes, which is organized as follows:
videos-and-keyframes
--observation_1
---- keyframe_path
---- projection:
------ projection_x
------ projection_y
--observation_i
---- keyframe_path
---- projection:
------ projection_x
------ projection_y
--observation_n