Aerial Surveying

... issues and solutionsHome

A quadrat being used to assess mussel quality
fig1 - A 'Quadrat' being used to assess mussel quality
Eye level view of a crab tile field
fig2 - Eye level view of a Crab Tile field
Faced with the ever increasing cost and workload of surveying and mapping, it is reasonable that many project commissioners are considering Photogrammetric methods and are reaching for the nearest available Small Unmanned Aircraft (SUA/UAV/RPAS/Drone etc) to do the job for them. Figs 1&2 show typical estuarial scenarios where monitoring of live mussel density and crab tile populations are being carried out. See here for a project we carried out a few years ago on the EXE estuary.

There are of course very convincing economic and technical arguments for undertaking these surveys by deploying camera carrying SUAs. Operating below 400 feet, the upper limit set by the Civil Aviation Authority (CAA) they are capable of delivering high resolution geo-referenced imagery needed for the identification of ground features otherwise invisible to manned aircraft at higher altitudes, or missed altogether by on-the-ground surveyors.

All is not rosy in the aerial garden though! As with life, compromises have to be made which are not always obvious at the outset and a rush to the drone shop may well end in tears!

The purpose of these pages is to identify the key issues so that pitfalls may be avoided.
The issues It is probable that the purpose of a survey will be to identify specified ground features and to distribute and/or display relevant data in a Geographic Information System (GIS) such as for example MapInfo, QGIS or ArcGIS. Whilst implementing the survey it will be desirable to minimise the number of images acquired (for reasons of cost) whilst maintaining image resolution in order to discriminate the specified ground features. These two parameters directly compete with each other and since the ability to discriminate small ground objects is limited by the Ground Sampling Distance (GSD), this must be the initial determining factor. For all practical purpose only the camera altitude needs then to be specified. To demonstrate the effect of altitude on image quality and GSD, the selectively enlarged Images (fig3-5) below taken at altitudes of approx 100, 50 and 20M show footprints (approx 300mm long) on sand adjacent to a corner of a tarpaulin. The captions show in sequence: Altitude(M), GSD(cm) and Ground Area in square meters.

selective enlargement at 100M altitude
Fig3 - 100M/1.72cm/6000M2
selective enlargement at 50M altitude
Fig4 - 50M/0.88cm/1555M2
selective enlargement at 20M altitude
Fig5 - 20M/0.36cm/260M2
At an altitude of 100M and GSD of just under 2cm, fig3 shows that a single footprint (about 300mm) incorporates about 15 pixels, and would be a reasonable choice for identifying ground features of similar size. Coincidentally, a typical house tile used as cover by moulting crabs (see survey report on this subject) about the same size. The effective ground area covered by a single image at this altitude is about 0.6Ha. However, overlap will typically reduce the effective yield per image to between 0.1 and 0.2 Ha.
Having acquired the images, what to do with them?  One possibility, well established in the public domain is 'Photo-Stitching'. Readers may be aware that this is a process whereby several overlapping images can be automatically combined into a single 'panorama' either 'in-camera' or through computer based software.

The images below show an example of photo-stitching on a small scale with a single image (fig6) and 4-image composite (fig7)  produced by 'Hugin'. A detail common to both images is outlined in a blue rectangle. 

A single aerial image of a mussel bed
fig6 - A single aerial image of a mussel bed
This is an effective, rapid and cost-effective process where an 'overview' image is sufficient. However, there are limitations to this process which will cause it to fail (see TECHNOTE 1), neither can it produce images which are simultaneously high resolution, of manageable size and georeferenced. (see TECHNOTE 2) Also, typical public domain photo-stitching software will not generate a georeferenced result which cannot therefore be a candidate for GIS inclusion.
A composite of 4 images stitched with Hugin
fig7 - A composite of 4 images 'stitched' with Hugin

TECHNOTE 1: Fundamental to the stitching process is the availability of common fixed ground features in adjacent overlapping images but, as is obvious from these examples there are both large featureless areas (water/sand) and moveable items such as buoys, water craft and birds.  Under these circumstances no-match = no-stitch!   Georeferencing data is ignored by Hugin so the composite image cannot be used as a legitimate source for a GIS. For a good (reasonably non-technical) overview of the subject and its limitations look here
TECHNOTE 2: Whilst it IS feasible to photo-stitch georeferenced images, the highly specialized software to do so is expensive to acquire, requires considerable technical user interaction and may still not achieve the required ground resolution. In any event it will be subject to the same intrinsic limitations as all stitching processes which rely on feature matching. See our work on 3D modelling here.
For all practical purposes therefore, photo-stitching cannot be a realistic option where high ground resolution is combined with a large survey area It is inevitable therefore that analysis will need to proceed on an image by image basis, an onerous task for a significant survey area!

The analytical process following image acquisition will involve the validation of the data in terms of their Georeferenced parameters (Lat/Long, altitude and camera orientation), by visual inspection to identify, enumerate and document the required ground features (whilst rejecting redundant or duplicated data created through image overlap) and the compilation of GIS layer files.

Unless applied to a trivially small survey, this would be a highly labour intensive process with many opportunities for error and may put at risk the realisation of the anticipated economic benefits of the aerial surveying process itself!

In addressing these issues we have used our extensive background and expertise in Information and Database Management Systems to design 'IMAGIS', a unique and user friendly suite of software tools which significantly improves the accuracy, quality and efficiency of the analytical process.

We believe that the significance and importance of this phase of the survey cannot be overstated so we include these software tools at no extra cost to the client!  Further information about APCO services and 'IMAGIS' software tools can be found here.

In conclusion and for those perhaps not particularly familiar with the subject of SUAs generally whether camera carrying or not, it is worth reflecting on some of the legal issues surrounding their operation (see TECHNOTE 3)
TECHNOTE 3: The overiding consideration is Public Safety. This requires compliance with the relevant articles of the Civil Aviation Air Navigation Order (ANO) and corresponding guidance notes CAP722. (165 pages if you have the time). Where 'Aerial Work' is being carried out (i.e. commercial operations as opposed to sport or recreation) then operators and pilots must be in posession of a 'Permission for Aerial Work' granted only to those who have undergone appropriate training, have passed both ground tests and flight assessments and carry mandatory Public Liability Insurance. These permissions must be renewed annually and pilots must demonstrate 'currency' by meeting minimum flying hours criteria. Where flying platforms are camera-carrying then 'Data Protection' requirements must also be met, to be found in guidelines issued by the Information Commissioners Office (ICO). (only 44 pages)