Ground control points, or GCPs, are an important ingredient in photogrammetry to obtain accurate outputs. These are typically clearly visible physical markers on the ground, along with accurate positional information. The accurate physical coordinates, and the pixel coordinates of the GCPs in the images are processed together to improve the positional accuracy of the outputs.
The Vimana platform fully integrates the usage of GCPs in photogrammetry, and provides tools to assist in placing, verifying and processing photogrammetry inputs with GCPs. In this article, we focus on the placement of GCPs in a site.
A general rule of thumb while placing GCPs is to uniformly cover the area of interest (AOI) with GCPs. To make this notion precise, we introduce the notion of coverage radius.
The coverage radius of a specific placement of GCPs is the maximum distance of any point in the AOI from the closest GCP. For example, a coverage of 250m means that every point in the AOI is within 250m from the GCP that is closest to that point.
Coverage radius provides us a handle on the accuracy of the outputs of photogrammetry. Placing GCPs with similar coverage radius should yield outputs of similar accuracy, assuming other parameters (such as image quality, overlap, accuracy of GCP readings) remain the same. However, note that the number of GCPs required to achieve a given coverage radius increases linearly with the area of the AOI, and thus it may not be feasible to implement a small coverage radius for large AOIs.
While coverage is the prominent factor to influence GCP placement, a few additional concerns are also important depending on the site realities.
Un-even elevations. On sites with large change in elevation across it, GCPs must be placed at all ranges of elevation. Even though the elevation variance is typically unavailable during GCP planning, distributing GCPs uniformly should approximately cover the full range of elevations. However, small hill-like structures may be missed in the planning; in such cases, the closest GCP may be moved to around the top of the hill.
Inaccessible locations. Sites with ponds, and other inaccessible locations may inhibit the placement of specific GCPs as per the plan. In such cases, the GCP may be placed at the closest accessible location. Typically, the GCPs may be freely moved upto 10% of the coverage radius, from its specified location (say about 20-25m for a coverage of 250m.
GCP location. The specific location to place a GCP must be reasonably flat. Avoid placing GCPs on small rocks, or on top of narrow structures such as walls, or gates.
While GCPs help with improving accuracy, for verifying the accuracy of the outputs we use independent check points (ICPs). These are usually placed at random throughout the site, and their positions are measured similarly to GCPs. ICPs are not used for referencing the outputs in the photogrammetry process, and only used to verify the accuracy of the outputs. Using more ICPs increases the confidence in the accuracy of the outputs (eg. 10 ICPs yields a 90% confidence estimate on the RMS positional error). The ASPRS PAS – 2014 standards for digital geo-spatial data mandate a minimum of 20 ICPs for generating certified maps.
The notion of coverage radius also allows us to automate the placement of GCPs. The vimana platform provides tools for both: placement of GCPs in a given AOI with a specified coverage radius; and to calculate the coverage radius of a given placement of GCPs. These tools allow us to easily verify the feasibility of placement before stepping on the site!
Technically, we formulate the GCPs placement as a Geometric Set Cover problem, and employ a randomized approximation algorithm to solve it. This provides us a placement procedure that is economical in the number of GCPs to be placed. Refer our whitepaper for more details on the algorithm.