LiDAR Quality Control

The quality of LiDAR point clouds is tied to laser range measurement accuracy, system calibration, and trajectory.

• Laser range measurement accuracy affects the noise level within the point cloud.

• System calibration and trajectory impact the alignment of the point clouds form different flight lines.

Noise Level

The noise level within the point cloud becomes evident when examining the point cloud over a flat surface. The thickness or spread of the point cloud indicates its noise level.

Figure 1 illustrates the point cloud captured over a horizontal surface by a LiDAR with a range accuracy of a) ±3 mm and b) ±3 cm. The spread of the point cloud in the vertical direction indicates the noise level. As depicted in the figure, the point cloud generated by the LiDAR with superior range accuracy yields a more defined plane, indicating reduced noise within the data.

Figure 1. Point cloud over a horizontal surface from LiDAR with a range accuracy of a) ±3 mm and b) ±3 cm.

Point Cloud Alignment

The critical task for LiDAR quality control is to examine the alignment between point clouds reconstructed from different flight lines, a combined effect of system calibration and trajectory. Alignment issues are typically noticeable when examining solid surfaces (e.g., ground and buildings) or linear features (e.g., light poles and traffic signs). Therefore, an effective method for assessing point cloud quality would include:

• Extracting various well-distributed profiles or objects from the point cloud

• Evaluating the alignment quality by inspecting the side view of the profiles or objects

Example

An example involves extracting thin N-S and E-W profiles and visually inspecting their side view is described below. CloudCompare is used for point cloud manipulation.

Figure 2 displays a sample point cloud (colored by height) in an agricultural field. P1 and P2 are two profiles in N-S and E-W direction, respectively.

Figure 2. Sample LiDAR point cloud (colored by height) where P1 and P2 are profile locations.

Figure 3 shows the side view of P1 with zoom-ins on the ground (colored by GPS time) in two scenarios, showcasing a) good alignment and b) poor alignment. In Figure 3a, the point clouds from different flight lines exhibit good agreement. In this case, the spread of the point cloud along the vertical direction would indicate the noise level. Conversely, the discrepancies in ground representation shown in Figure 3b underscore the poor alignment quality among point clouds from different flight lines. Similar pattern can be observed for P2, as depicted in Figure 4.

Figure 3. P1 side view in two scenarios: a) good alignment and b) poor alignment.

Figure 4. P2 side view in two scenarios: a) good alignment and b) poor alignment.

Interpretation of Quality Control Results

• High noise level within point cloud

  • Potential cause: Suboptimal laser range measurement accuracy

• Alignment issues detected across all datasets

  • Potential cause: System calibration issue

  • Explanation: Error in mounting parameters would affect all datasets.

• Alignment issues detected in a single dataset

  • Potential cause: Trajectory issue

  • Explanation: Trajectory is independent for each flight mission.