Calibration of images that contain large portions of water bodies is challenging. It is recommended to increase the overlap and making sure that some land is visible in the images.
Images that capture only water cannot be calibrated (optimize the internal and external orientation of the camera). The reason is that it is very hard to find keypoints on water surfaces as the water is very uniform and reflective. Even if keypoints are found, it is very hard to match them between images, as these points are not static (water moves).
Since the images that contain only water do not get calibrated, they do not contribute at all to the processing and can not be used to give color to the orthomosaic. Therefore, there might be holes (lack of information) in the orthomosaic on water surface areas.
It is suggested to adjust the flight plan and:
- Increase the flight height to cover a larger area with a single image.
- Increase the overlap.
- Images should contain at least 30% of land.
Example: See the Water rescues thanks to drone mapping and Out in the blue: mapping along the Indian Ocean blog post where the orthomosaic got generated due to the surrounding land area.
In this case, the images could get calibrated and contribute to the orthomosaic generation. For more information about how to map water surfaces: Step 1. Before Starting a Project > 1. Designing the Image Acquisition Plan > a. Selecting the Image Acquisition Plan Type.
- If the images get calibrated but a lot of artifacts and noise are generated in the water area:
- (Pix4Dmapper) Use Point Cloud Editor to disable the noise before processing step 3. DSM, Orthomosaic and Index
- (Pix4Dmapper) Draw surfaces in the point cloud to add a constraint for processing. This way the area will be flatted out
- Try processing with Pix4Dreact and Pix4Dfields. The artifacts are still expected in water areas but the chance of generating an orthomosaic without holes is higher.
if we have multiple images taken simultaneously, thus removing the static part of the problem, would it be possible to carry out image correction through automatic/manual picking of keypoints?
Hi Raul. This would likely still not work very well depending on the texture of the body of water. Some water conditions are better for calibration than others. Also, manual marking is very tedious and would likely take a very long time to get the images properly calibrated. Even then the ATP generation would be very low. It is hard to say as the conditions are important.