The first thing to establish is what exactly are the required outputs for and how will they be used? Whilst very accurate results can be achieved using photogrammetry (1/2mm) this in many cases is not required. There are companies who can provide these levels of accuracy and to do so the right hardware has be to employed together with a robust methodology. The majority of our clients are in the construction, quarrying and landfill sectors where the ouptuts don't need to have mm accuracy and in many cases 40/50mm is suitable. It must be noted that we do not provide photogrammetry derived surveys for setting out or detailed construction work as the outputs are not sufficiently accurate.
Much of our work is in providing surveys for uses such as initial site planning/Master Planning, monitoring site development, ecological monitoring, mining subsidence, flood risk and volumetrics. In all these cases and with any job the oupts and accuracies required are agreed as part of the fee proposal.
How though are accuracies acheived? The key is in the planning and and capture of ground controls. I personally prefer to use ground controls as opposed to emplying RTK/PPK enabled drones. Professional topograhpical surveyors who I work with have persuaded me that the use of ground controls is the best option for the desired outputs. They take longer to employ but results have been consistent. I accept that over time we may change our thinking on RTK/PPK.
During flight planning it is important to distrubute the ground controls over the area to be surveyed. We place a number around the edges of the survey in (inset by say 50m) and in places where they will be captured by adjoining flights. This assumes that some large surveys require multiple flights. In addition one of more need to be placed within the survey area. If they are large variances in height I will try and place GCP's close to high and low points as well.
When you place your GCP's on site take your time. Make sure the GCP is pegged and secure, make sure you GNSS rover/survey pole is placed in the centre of the GCP, ensure it is vertical. When capturing the location set the station up so it averages 5 readings.
When you fly the site ensure that the image is in focus and you have set the correct exposure.
We flew seven sites for a client last week and these have all now been processed. When matching GCP's we will match at least 10 images with the GCP. Some guidance suggests that 3 will suffice. Our thinking is that it you have them use them. The results shown left show the RMSE errors for one of (and the smallest area surveyed). You will see that they range from 0.0025m to 0.0061m ie 2.5 to 6.1mm. This is useful information but not a realistic figure to be quoted. For this survey the ground sampling distance (GSD) was 20mm and the general rule is that X/Y accuracy will be at best 2x GSD and in Z 3x GSD. These are the parameters we were working to and the accuracy the client required.
Four our client they have had seven areas surveyed and results delivered within a week (orthomosaic/DSM/point cloud/video). The alternative was a traditional survey team who would have required 6+ days on site with the client having to provide H&S backup which involved a number of personnel. Without going into details this site comprised of various "high risk areas" to be surveyed and if you wanted to go into these areas H&S and rescue personnel had to be in attendance. Thereafter the data would take a few days to process.