top of page
Writer's pictureswarnagsk

Mohan’s Musings Drones – This & That




Mapping industry is so tiny that it hardly affords to invest in hardware research. Hence, it discreetly absorbs and embraces the technologies developed elsewhere - altogether for different purposes. Example: stereo-ready monitor, graphic card and the latest to join the club is drone.Seeing a huge potential in surveying, mapping and monitoring, drones are turned into flying data collection devices by equippingwith different types of sensors and cameras. This augmentation resulted in UAV system - a drone fitted with a whole set of hardware and software tools.

And that is the disruptive technology of the day whose merits and pitfalls we like to identify – albeit partially. Its closest sibling is aerial photogrammetry – obtain clearance, fly, download photos, and process – except for the major variance is in size and stability of the flying device. Some of the applications that it is effectively put to use are:

· Surveying

§ Terrestrial surveying

§ Asset management: Highways

· Mapping

§ Produce sub-metre contours, ortho photos, X-sections and L-sections

· Monitoring

§ Power: Inspection of transmission lines for condiuctor sag, interference of vegetation

§ Agriculture: Manage and monitor crop growth – through periodic NDVI data

§ Construction: Monitor the progress of highway building

§ Insurance: Validate a claim through rapid data acquisition

§ Mining: Produce DSMs of the open-cast mines; compute the volumes of over-burden, stock piles

Let us take a look at the products of a drone project. They are quite familiar – the subtle variance lies in the fact that we are looking at the terrain reconstruction at centimetre level details.Pixel size (ground sampling distance) as fine as 1cm is possible with drone surveys. And of course, they require matching GCPs of sub-centimetre accuracy.

Raw products:

Photographs with camera position and orientation stored in its header

Scans (LiDAR profiles)

Geospatial data products:

3D models and point clouds (sets of data points in object space, tagged with XYZ)

Surface model (DSM)

Orthophotos (geometrically corrected aerial images)

Value-added products

X-sections

L-sections

Contours

Photo-realistic terrain simulation

All is not well with drones. Once airborne, these tiny birds are fragile enough to get swayed away by winds resulting in irregular flight lines and overlaps. Let’s look at some of the issues:

· GPS errors: Let's bear in mind that the position always comes with an inconsistent error. This manifests inaccurate profiles and improper ortho mosaics.

· GCPs: Adequacy, distribution, and more importantly their stability through the data collection.

· Camera calibration: The interior orientation parameters may change from one job to another – as the cameras employed are non-metric

· Automation: Increased dependence on automated processing lands you in trouble at times. Make sub-blocks when the project area comprises geographical features such as rivers and streams.

Accuracy

We have discussed the terms and significance of accuracy – absolute and relative in the previous musing. Let us quantify them for drone surveys from literature. You must, however, remember that the errors mentioned below are only indicative - may be larger in complex terrains such as high hills, forest, desert, and water.

· One can expect a relative accuracy of 1-3 times the pixel size for a correctly reconstructed model, both horizontally and vertically.

· One can expect an absolute accuracy, of 1-2 times the pixel size horizontally and 1-3 times the pixel size vertically for the correctly reconstructed model.

Check the accuracy

Visual inspection of products (such as 3D models, orthoimages) could be quite misleading – as inaccurate and accurate products appear to be same. Hence, quantitative checks through the use of a set of check points is necassary (not used in the process yet). Measure them from the reconstructed model and compute the error at each point in plan and height separately to arrive at statistical measures such as RMSe, LE, CE.

Caution on Pixel size selection:One tends to choose the finest possible pixel size to achieve the best accuracy – but there is an associated baggage! A survey of 1 km2 area with 1 cm resolution generates nearly 200 GB of raw data while 2 cm resolution reduces data to ¼ th (i.e. 50 GB). And also processing times will be cut down in same ratio. Hence, reckon the application, desgn a project and deliver the commensurate to the enduse - rather than getting carried away by what the system can deliver!

Drone-based mapping is a reality and they appear to have become a fixture in photogrammetry. Listed above are some sample applications and we can expect a spur of a quite innovative ones!Enjoy this novel tool with a feel of I-have-done-it-myself-flying-to-processing!

Cost of large scale mapping ( spatial data capture ) by the traditional digital photogrammetry works out to less than Rs 200 per acre, Drone mapping is 40 % lesser. In fact this cost is insignificant compared to the applications based on this mapping. In conclusion, it can be said that Drone surveys have emerged as efficient latest technology for large scale mapping in terms of accuracy, time and cost.

13 views0 comments

Comments


bottom of page