The first thing that you need to take note of, in any mapping project, is the list of deliverables and the spatial frame they are defined in. Next point to ponder about is if the inherent accuracy of the raw images will suffice the accuracy requirement. Every image these days – obtained through aircraft, satellite or drone- carries geotags as part of its header. However, these geotags suffer from errors – such as bias in satellites (besides drift in drones and aircraft) – calling for additional inputs to improvise the absolute location accuracy of the end products. However, if one is interested in relative accuracy through measures like length of a pipe line, perhaps one can proceed right away. This is where we consider employing a few accurately measured, sharp and static marks – popularly called Ground Control Points, or popularly GCPs.
GCP acquired for Cartosat may not be valid for Geoeye.
Sources of GroundCoordinates
A GCP is a feature that you can clearly identify in the raw image for which you have a known ground coordinate. Ground coordinates can come from a variety of sources such as GPS, conventional ground surveys, orthophotos, vectors, GIS layers, legacy topographic maps, GCP libraries,..
A GCP establishes the relationship between the raw image and the ground by associating the image space coordinates of the imaged point (line, pixel) to the object space coordinates (XYZ or LLH) on the ground. The relationship can be established through simple math models like polynomials or rigorous sensor models.
It is always desired to go for minimum number of GCPs as their procurement is expensive.
What it means to be a good GCP?
The quality of GCPs directly affects the accuracy of your math model, and that in turn, determines the end product quality. Note the following important qualifiers:
• Identify the features in the raw image that you want as ground control before collecting GCP coordinates in the field using a GPS or ground survey. The GCP acquired for Cartosat may not be valid for Geoeye.
• GCP distribution is important. The horizontal spread of the points in the project area is often stressed upon. It is equally important to collect GCPs from a variety of height across the project elevation range. Public domain DEMs such as of SRTM can be highly helpful for planning.
• Plan the GCPs in such a way that each of them is spotted in as many images. In a project of high overlaps, it is not uncommon that a GCP can be measured in 8 images (or in otherwords, 8 rays per point). This produces a more accurate model as one GCP is stitching multiple images together.
• Select features that are close to (or on) the ground. Features having relief, such as buildings, may appear to lean in the image. When a point located on a boundary wall is to be measured, you need to offset for the wall height if GPS is kept on ground.
• Avoid using shadows as GCPs. Although shadows may be easy to spot in the image, they can move from one image to another due to time interval in image acquisition.
• Avoid selecting commonly occuring or repetitive features such as street furniture, parking areas. Point transfer can be ambiguous.
How many GCPs I need?
It depends. The stability of on-board sensor, terrain ruggedness, the math model employed to process the images are some of the crucial factors. And of course, the fraction of points that you like to earmark for evaluating the deliverables – called Independent Check Points (ICPs).
If the terrain is flat, and that you choose to rectify the images with first-degree polynomial, go for 4 points (3 GCPs + 1 ICP) per image.If you choose to employ rigorous RPC model, for stable satellites such as Cartosat, GeoEye, it is okay with one GCP and one ICP per image. Please note these are bare minimum configurations. It is always desired to acquire a bit more points as deploying field teams is expensive – both in time and money.
You do not need GCPs here!
You are lucky you do not need any GCPs or ICPs if only relative accuracy – such as:
• Requirements such as measuring length, area, and volume of features
• Construction Site Management and Oversight
• Crop monitoring
• Accident site reconstruction
• Creating promotional material
Processing is fast that you don’t plan and wait for GCP surveys. Collect the tie points; perform the block adjustment. And that’s it!
Applications such as Cadastral surveys, as-built surveys, pipeline alighnment do require GCPs and 30% of the points are jack-knifed for checking purpose.
And finally about EPSG
Ensure the spatial frame of the GCPs iscompatible with end-product specification. If the points are in different frame, do necessary transformations. Two instances to illustrate:
• The deliverable is Indian Everest and the GCPs are acquired with GPS, carry out WGS84 to Everest transformation.
• If contours are to be delivered in MSL frame, convert the ellipsoidal heights of GPS survey using the EGM model or through conventional levelling.
Brace yourself for surprises
Plan the marking and measuring of GCPs appropriatelyand quickly soon after acquiring the raw images. GCP survey is a field activity and hence be prepared for unforeseen uncertainties. In a heli-borne LiDAR project, the GCP team assured me they would complete the task by dawn so that the copter could take off in the ensuing morning. The team went incommunicado till noon. Along chase ensued to locate the team – forget the GCPs. The team gave a sombre narrative of the ordeals they went through– the primary being their inability to reach the planned location due to darkness. A nightmarish learning – it gets dark in the nights!So, factor in for the known and unknown unknowns.
Comments