A number of factors should be considered for collecting UAV data to produce 3D point cloud and orthophoto. I would like to list a few key points in this article.
High overlap is required to produce good quality 3D models. A minimum of 80% forward overlap and 60% sidelap should be maintained. These numbers should be increased with the presence and abundance of complex elevated objects in the study site.
Spatial resolution/ground sampling distance:
The general rule of thumb for determining a pixel size in remote sensing data collection is that ‘the smallest resolvable object should be at least two times larger than the minimum pixel size’. However, my experience is that this rule does not quite work for UAV data due to uncertainty in photo quality, undesired changes in altitude and flight path and many other factors. My suggestion, in this case, is that the minimum pixel size should be at least one-fourth of the smallest resolvable object.
On the other hand, flying at a lower altitude to obtain high-resolution photo might result in higher perspective distortion close to the edges of the photos. This becomes especially problematic when we have many elevated objects (e.g., trees, buildings) in the study area.
Shadow minimization/photo quality:
Ideally, we do not want any shadow in our images. The experiment by Kenn McLean well explains the impact of shadow on point cloud accuracy.
However, when trying to collect shadow free imagery, we also need to make sure that there is enough stable and diffuse light so that the objects are clearly visible in the image.
Nevertheless, it is not always possible to collect shadow free images with sufficient diffused light, especially when data is collected over a large area. We are currently working on an algorithm that is supposed to help to produce good quality orthophoto and point cloud from images with hard shadow. Please keep an eye on this blog to know more about the algorithm.
Another thing to consider is the frequency of capturing photo (intervalometer settings) vs the data rate (the time it takes to write a photo to the memory). We need to make sure that there is enough time to write one image to the memory before shooting the next.
Shutter speed/exposure time:
The exposure time should be less than the time that the carrier (UAV) takes to travel half a pixel distance.
Photo capture mode:
Photos should be captured in RAW mode to maximize the quality. Later, if required, they can always be compressed to a different format.
Other Camera Settings:
These are a quick summary of what I have learned in the last 3-4 years. Please feel free to add to the list (by commenting). Besides, you are welcome to open a discussion if you have a different opinion on any of these.