How can the issue be resolved if 2D reconstruction results show obvious stitching traces?
Issue Description
After 2D reconstruction, there are obvious traces of stitching on the water surface or crop surface in the result files (DOM), as shown below.
Cause of Issue
As the water surface is not diffusely reflective, the water surface of the same area will differ greatly from different drone positions. In addition, the photo parameters shot by the same camera will also differ, resulting in a large difference in photo brightness. As such, obvious traces of stitching may occur. At the same time, large areas of crops will produce a certain degree of non-diffuse reflection, so there will be relatively obvious stitching traces in the 2D reconstruction result of field scenarios.
Diffuse reflection is the phenomenon where light beams striking the rough surface of an object are reflected in an unordered manner in all directions. When a beam of parallel light strikes such a rough surface, the unevenness of the surface means that the normal direction at each point is different, causing the reflected light to scatter in various directions. Even though the incident light is parallel, the reflected light will scatter irregularly. This type of reflection is referred to as diffuse reflection.
The water surface is like a mirror with smooth surface, and when parallel light beams strike it, most of the reflected light will also be parallel, so the reflection on the water surface belongs to non-diffuse reflection.
Solutions
1. Modify the camera metering mode as average metering. This operation will solve this issue to some extent.
Note: Zenmuse P1 and Zenmuse L1 use global metering by default.
2. Avoid data collection when the sunlight is strong (e.g., at noon).
3. The algorithm will be optimized in later updates.
4. In DJI Terra v3.4.0 or above, enable Light Uniformity. This operation will solve this issue to some extent.Version V5.0.0 algorithm optimization, the reconstruction effect is not much different from when the function is turned on, so the function is cancelled.
How can the issue be resolved if there are white spots, missing parts, or black parts in 2D reconstruction results?
Issue Description
There are white spots, missing parts, or large empty areas in the reconstruction results.
Cause of Issue
Insufficient photo overlap or object features
1. The photos have insufficient overlap:
For example, the flight altitude is insufficient when shooting tall buildings. This could result in insufficient side features in the photos. Or with highly undulating terrain, Terrain Awareness Mode is not enabled.
2. The object features are not clear:
For example, in white walls or large fields, areas with distinct textures can be successfully reconstructed. However, in areas without distinct textures, DJI Terra is not able to identify the feature points of that part in the photo, leading to unsuccessful reconstruction and large empty parts in the output.
3. The scene involves roads with many moving vehicles or water surface with many moving ships.
4. When capturing a solitary object such as a billboard, if the object is too thin, and photos are insufficient, the generated point clouds are sparse on the billboard, resulting in instability of generating grids, ultimately causing empty areas when fitting the texture to the model.
Third-Party Software Compatibility
The data results of 2D reconstruction display normally in DJI Terra, but when opened in some third-party software, areas that were originally pure black will turn into white spots. This issue is cause by the settings and display of the third-party software.
Solutions
1. If the original photos are not sufficient, it is advisable to retake photos to increase the overlap rate or create the flight route and perform the mission again.
2. Currently it is not able to avoid empty areas in water surface reconstruction. In DJI Terra v3.6.0 or above, it is recommended to enable “Refine Water Surface”.
If the issue cannot be solved after going through all instructions above, consult a third-party modeling software such as TJH DP-Modeler or DASPATIAL ModelFun to check if the model can be repaired.
When capturing a highly repeated scene, such as a white wall, large areas of forests or fields, etc., it is unavoidable that the objects themselves come with a texture without distinctive features. It is recommended to make some special markings on the solid color parts of the objects, such as adding feature targets to easily recognize features in DJI Terra.
3. If there are constantly moving objects in the scene, it is recommended to wait until the movement stops before collecting data for reconstruction.
4. For capturing a solitary object such as a billboard, it is recommended to take photos of the object from different angles after performing auto flight route, and it is better to take photos of the front side of the object and all surrounding objects. Then import all photos into DJI Terra for reconstruction.
5. If the reconstruction results can be displayed normally in DJI Terra but cannot be in third-party software, it is recommended to view them in DJI Terra, or refer to the help document of the third-party software.
How can the issue be resolved if there are missing parts in LiDAR point cloud reconstruction results?
Issue Description
There are missing parts in LiDAR point cloud reconstruction results in DJI Terra, but if the same photos are used for visible light reconstruction missions separately, the photos are continuous and the results are complete.
Solutions
1. Please perform reconstruction missions with DJI Terra v3.8.0 or above. Click here to download the latest version of DJI Terra.
2. Please ensure that calibration flights are performed during the mission with fixed RTK, which can be verified by photo parameters; if PPK is involved, please ensure the base station data can cover a complete flight route.
3. Please check if the missing parts are not water areas, excluding any missing caused by the object features itself.
4. Please check if the missing parts include no terrain with significant elevation changes, excluding any missing caused by Zenmuse L1 or Zenmuse L2 in terms of insufficient range.
5. Please make sure that regular calibration has been performed during the flight to avoid removed LiDAR point clouds because of poor attitudes.
6. Please check if the microSD card has been used for a long period or its speed is too low. Please replace the microSD card to check whether the issue is resolved.
If the issue cannot be resolved after going through the above steps, there could be some recording errors with the LiDAR module of Zenmuse L1 or Zenmuse L2. Please send your device back to DJI for testing, and click here to submit an online repair request.
How can the issue be resolved if there are white parts or missing parts in the RGB page of LiDAR point cloud results?
Issue Descriptions
1. If RGB coloring is not enabled when using the Matrice 300 RTK with Zenmuse L1, or no photos are added during reconstruction, it is normal for the point cloud result shown on the RGB page to be white in DJI Terra, as shown below:
2. When Matrice 300 RTK with Zenmuse L1 performs a multiple-route mission, there are missing parts in the RBG page of LiDAR point cloud results in DJI Terra, as shown below:
Cause of Issue
1. Please check if RGB coloring is enabled for all missions. If it is not enabled for some flight routes, it is normal for the results to be missing.
2. Please check if visible light data and LiDAR data are stored separately, as shown below:
Solutions
1. If any phenomenon above exists, it is recommended to store visible light data and LiDAR data together and import them again into DJI Terra for reconstruction.
2. The following microSD cards are recommended. Please check if replacing the microSD card helps.
SanDisk Extreme 128GB UHS-I Speed Grade 3
SanDisk Extreme 64GB UHS-I Speed Grade 3
SanDisk Extreme 32GB UHS-I Speed Grade 3
SanDisk Extreme 16GB UHS-I Speed Grade 3
Lexar 1066x 128GB U3
Samsung EVO Plus 128GB
If the issue cannot be resolved after going through the above steps, please provide the following information and contact DJI Support for assistance.
1. Gimbal log
2. Original data
3. Flight controller data
How can the issue be resolved if some edges of the buildings in the reconstructed photos is jagged?
Issue Descriptions
After a model is reconstructed, some jagged building edges, which do not exist in reality, may appear in the reconstructed photo, which is caused by photo stitching.
Solutions
These jagged edges are residual traces during the photo stitching process. Currently, it is unable for DJI Terra to eliminate these jagged traces, but our engineers will continue to optimize the algorithm.
