Hi everyone,
The goal: i want to register crop point clouds and spectral information in a plot of the field.
The data detail:
1.The crop point clouds will be obtained by terrestrial Lidar (or UAV with rgb using SFM method) in a plot which is of high quality, you can see the accurate 3D structure of the crop.
2. The spectral information of crop will be obtained by UAV with Multispectral camera.
This is a field experiment, and i am a rookie in this field, I hope you can give me some suggestions, if it is some literature, it would be better.
Thank you !
Thank you sir Changhui Jiang .
I used Hyperspectral LiDAR indoor to scan maize working with Institute of remote sensing, Chinese Academy of Sciences. But it is too hard to use in the field.
Can you scan the crop in the field ? It would be great if it can be mount on a UAV.
The answer depends on the method of creating the point cloud (e.g., LiDaR, photogrammetry, or SfM) and the type of camera/sensor used to generate the multi-spectral images.
I've aligned multispectral image sets with a 3D point cloud using SfM software formerly called Photoscan (now Metashape). With SfM (e.g., MetaShape, Pix4D, and others), the images are automatically aligned with the point cloud because the point cloud is derived from matching features in the overlapping source images.
It requires all of the spectral image sets for each waveband to be registered with each other, i.e., all wavebands must be captured from precisely the same camera position with per-pixel registration. As a co-requirement, all the images would need to have the same lens calibration. An example of an SfM spectral imaging system are the cameras made by MicaSense (no endorsement implied), and there are other similar types of multispectral camera systems designed for use with UAS.
You would generally use one image set (such as RGB or single-waveband monochrome images) to generate the point cloud in SfM, then substitute the other spectral images to generate orthomosaics for each waveband. Once the dense point cloud is aligned and optimized for accurate lens calibration, you can produce an orthomosaic image for each waveband simply by substituting the registered images for each waveband.
You'll need to organize the images for each waveband into a separate folder and it will generally require renaming the image sets so that the set of images taken from the same location in each folder has the same name. Then, the software can make the substitution and generate the orthomosaic for each waveband or combination of wavebands (e.g., false color IR).
If the image sets for each waveband are not registered per-pixel, you can use various image alignment techniques (the contour cues paper above looks interesting), but that's beyond the scope of this response. Hyperspectral cameras with line-scanning or "pushbroom" sensors are different. Some systems combine hyperspectral cameras with LiDaR as others suggested--those are also outside my experience.
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy
- Crop Science
- Crop