Aerial Imagery Resolution; "The Good, The Bad and Not So Ugly"
First, this article was written to help those who are just beginning to investigate the use of drones or small plane aerial imagery for mapping purposes. Recently, I attended a conference and was amazed at the number of people who desire to use this technology for their organizations. It is hoped that this article will help these people understand the basics of SfM. This article is not meant to be offer anything new or earth shattering for those practiced in this field.
Decision Making Aerial Imagery Products such as orthomosaics are produced through a photogrammetric process known as Structure from Motion (SfM). Different from the "classic" photogrammetry that required calibrated cameras and significant expertise, for relatively small areas, SfM can use low cost cameras and the software available today to leap the tall buildings that once required expensive cameras significant skill, knowledge and experience. For people at a knowledge level that find this article informative, I would keep project size to less than 40 acres.
So with images that are also called rasters, the pixel is the basic building block of data. Each pixel has a single value that for images represent a color value. The size of the pixel we refer to the resolution and specifically the size of the pixel on the surface of the earth is the Ground Sample Distance or GSD. The GSD is the distance between two consecutive pixel centers measured on the ground. The bigger the value of the image GSD, the lower the spatial resolution of the image and the less visible image detail. One might think that the smaller the GSD and higher the spatial resolution the better life is. And to a certain extent that is true. However, with everything there is a trade-off.
The good side of small GSD is the higher level of detail that can be seen by the person looking at the images. The Bad side of the small GSD and high level of resolution is that the cost of acquisition can be significantly higher because for a given aerial acquisition platform. Whether it is acquired from a drone or an airplane, the camera covers a smaller footprint on the ground when the GSD is smaller (the resolution is higher). So it takes a longer flight time for the airplane or drone to fly over a given area when the pixel size is smaller versus a larger pixel size. When a drone camera acquires at 50 meters elevation and produces a pixel of 1 cm, the area covered by the same camera at 100 meters which produce a pixel of 2 cm is roughly 4 times larger. Thus, when you increase the resolution by double, you typically increase the number of images by 4 which may increase your processing time significantly.
Structure from Motion (SfM) is a photogrammetric range imaging technique for estimating three-dimensional structures from two-dimensional image sequences. As a rule, the more images and the higher the camera sensor pixel count, the longer the processing time. SfM is generally broken into two categories, 2D Mapping and 3D Modeling. The 2D Mapping requires that images be acquired with the camera at a vertical or NADIR perspective with the camera looking straight down from the aircraft. The images need to have an overlap and side-lap generally around 70% however there are several variables that can affect the required overlap and side-lap that are not the topic of this article. Generally, 3D Modeling requires significantly more images for the same areas than the 2D Mapping product. The number of images is generally more than double and are taken from a combination of vertical and oblique angles, and depending on the desired detail and complexity of the surface can be more than five times the number of images. The more images the longer it takes for the computer to process the two-dimensional image sequences into a three-dimensional structure.
As stated above, the processing time varies based on the number and pixel size of the images, but the configuration of the computer makes a significant difference in how long it takes to process an SfM project. The configuration of the computers based on the software that you use to process the project. However, in general the computer for larger projects needs to be at least an Intel I7 or AMD Ryzen 7 with with 32 GB of RAM, a GPU such as the Nvidia GTX or RTX Series GPU graphics adapters, and finally large (1 or 2 tb) quality SATA solid state drives. Generally, you can purchase a good video "gaming" computer for about $3,000 and load the SfM software and begin processing. The one caveat is that SfM software generally doesn't need the latest and greatest GPU to process. I use the Nvidia GTX 1660 which is a fraction of the RTX 2080 on a couple computers and they do a decent job.
Finally, SfM is a powerful tool that has changed not only the aerial imaging industry but how we make our decisions. Once you make your first orthomosaic image you are just opening the possibilities and the adventure of Decision Making Aerial Imagery & GIS Products.