# Topographical survey using multi-rotor and camera/LRF

A while back I was asked if there was a way to work out the volume of compost stacks at a local company that manufactures compost.

So you've got 20-30 +-5m high piles of compost. The way they currently do the volume calculations is a GPS is "walked" over each stack in multiple directions and the volume is calculated using the gps co-ordinates and the altitude. This is simple but cant be that accurate, with regards to the altitude data from the GPS.

I had the idea of flying a quad at a fixed altitude (say 10m) above the stacks, with a laser rangefinder on a gimbal pointing directly down. This would give your distance to ground.

Then you plan a search pattern like mission over the area you want surveyed.

The data you would get would be gps information as well as distance to ground/objects directly below the quad.

With this information it would be fairly simple to build up a 3D topographical map of the area you've flown in. Then some of that wonderful calculus stuff in matlab and you would have the volume of each stack of compost.

My question is:

1. Im pretty sure this method will work (especially with the new 10cm accurate baro on APM2)

What do you guys think?

2. Can you get a 3D topographical map using digital photos alone or will this method not be very accurate?

Thanks!

G:)

Tags: aerial, muti, rotor, survey, topographical

Views: 1994

### Replies to This Discussion

This depends on a few things, for sure. If someone were to do volume calculations using GPS technology, it would be survey-grade equipment, meaning accurate to within 5cm (see RTK GPS). If they were using sport-grade handheld units, then throwing rocks at the stack would give better results.

Though your idea is very sound in the theoretical workings, I don't imagine it would work as well as you're hoping. First, there is the issue of accurate elevation above the ground. I don't know what the margin of error is in the 10cm barometer, but I tend to think it might be a bit rough. In addition, the GPS for horizontal position could be off by a matter of 1-5 meters. Which brings us to the next point of errors in the rangefinder. The laser is susceptible to its own distance error, as well as angular error while flying (imagine pointing a laser at a distant object and trying to keep it in one place). Again, if they were using surveying GPS units, the results would be hard to beat. If they were using handhelds, then your idea may be worth a try.

Since helis are already popular with photographers, why not use that? You can get good images of the compost stacks (stereo pairs) and run the images through any photogrammetric software (Photomodeler, PhotoScan, etc.) and get survey-quality results in the form of 3D points, or a 3D surface suitable for volumetric computations. Food for thought, but my trex 600 is built just for that purpose: close-range aerial photogrammetry!

Cheers!

Eric

This can certainly be done. We do exactly this with our aircraft ( look at GUPPY in the forums under airframes – search NAMPILOT-) We overfly the area in a grid, taking downwards photos along  each grid line. The photos must overlap at least 60% along track, and at least 40% between tracks, so that will set the spacing of your grid, ie, work out what the ground coverage of the photo will be from your height AGL and the camera field of view. Then make your grid spacing so that you overlap each adjacent image at least 40%. Then the aircraft ground speed will determine the rate  at which you take photo’s since you need to get a 60% forward overlap.

Place a number of GPS located markers in the area of interest - try to get 4 to 6 markers within a 1000 square meter area.

Then, what we use is a software package by the Australian Co- Adamtech –

The software then takes the sets of images, with the GPS markers, and generates a 2D composite image of the entire image set, and then a full 3D image of the area – the overlapping images are used to do this in the same way the ‘old’ stereoscopic image were created using 2 cameras with known spacing in between.

This image, since it is geo referenced, has attributes that can be measured, ie, volume, area, etc . We use this to measure how much ore has been mined out of a quarry from month to month

Downside is that the Adamtech SW is very expensive!

Cheers

Joe

@Eric: Thanks for the reply, makes a lot of sense. I suppose the guys use pro stuff, but how accurate are the altitudes in those RTK gps units?

@Joe: Do you use a fixed wing or multi-rotor? I want to use a multi rotor.

Also, what camera do you use? And whats the name of the program from Adamtech?

Without getting into crazy theory and whatnot, it's reasonable to say that the elevation recorded with RTK can be within the 5-8cm range. It is elevation (height above a datum, as referenced by the 'base' GPS unit) vs. altitude which is typically a height above mean sea level, using barometric pressure.

Even with the most pro equipment, trying to climb up a stack of exceptionally loose compost while holding a gps on a pole can be difficult and as you climb, you changing the shape of the stack.

Flying over it and taking pictures, then using a program like 3DM analyst or photomodeler to generate DTM seems alot easier and more accurate. Your thoughts?

Yes, for things like that, a photogrammetric approach can be very useful. I've used this approach for doing stockpile volume computations, and it provides very good results without the need to run around in the field too much.

As I indicated, have a look at my forum post on GUPPY – it is a fixed wing 35kg A/C.

Regardless, any aircraft will do, you just have to get the 60%/40% overlap, or thereabouts. The greater the overlap the better the result. Altitude control is not that critical, but better than 2 or 3 meters is highly desirable. To improve accuracy even further you should calibrate the camera lens for distortion corrections, but if your overlap is large, and you take only downwards looking photos, then the central section of the images are used anyway where the lens distortion is minimum. The Adamtech SW perspective corrects all the images based on the GPS ‘targets’ placed in the images, and on the fact that there are many mages that cover the same point or sets of points, therby also relaxing the altitude accuracy requirments. Photoscan works similarly, but lacks many of the features of Adamtech, and the results are not, in my opinion, as accurate.

I have many pics but they are Huge – a small quarry of 200m X 150m close by us gives a 3D mesh image of 1.8GB……

The camera I used for this quarry is a Canon G7, now obsolete, which I fitted with a stabilized gimbal.

Adamtech SW for the full Monty is around \$18k to \$26K US.

Joe

Also, whats the price on the adamtech software?

I think you should stick to digital cameras, there are people that fly LiDAR or laser scanners on multirotor-copters (e.g Geodetic Institute Helsinki, Finland; University of Tasmania, Australia) they both use the "ibeo LUX" sensor which costs about 13,900.00 EUR. The problem is that the IMU accuracy is very important since you need to know the exact position and orientation of the sensor in order to calculate the position of the generated return pulse from the LiDAR.

I would go for digital photogrammetry and a good digital camera. As others have pointed out here, there are a few people already doing that kind of work. I just want to add a few software options:

• Agisoft Photoscan Pro, about \$5000. possibly one of the best photogrammetric software for small UAS and consumer cameras.
• MosaicMill, EnsoMosaic, Finland. from 5000 EUR (2009, I don't have an uppdated price). Has been around for awhile.
• Photogrammetry Software Suits, Leica, Inpho, Trimble, Erdas, professional but very expensive and made for large mapping cameras (but work for consumer cameras as well)
• Pix4D, Swiss company,  Web service! No software. You upload your geotaged images and they process it to 3d data. Cost unknown.
• Pieneering, Finland. Web service? and Software?

I'm planning to do much the same as you and I think I'll start with sending images to Pix4D for processing and then when I have earned enough money and are more in to the market buy the AgiSoft software.

Cheers,

Jonas

I agree, for a small vehicle, photogrammetry is probably the best option. It doesn't require any more equipment than the vehicle, camera mount, and camera.

I've used Photoscan and performed an analysis of the results compared to a LiDAR dataset of the same subject which I acquired at the same time. The Photoscan model fit with the LiDAR model to within a couple centimeters (many of the sampled points within 5-10mm). For the Professional version of Photoscan, it is only \$3500! Certainly something I will continue to be using in the future for jobs like this.

This is the kind of work we are looking at getting into, there is a commercially available product from sensefly that uses a small foam aircraft and digital camera, I think it's marketed by Korec in the UK but costs around £10k, seems pricey for what you get but it proves the concept, I imagine people are doing much better with their own designs though.

Greetings,

I've not flown a SenseFly but I've looked at it pretty closely. My impression is that it is quite well constructed, the systems are well integrated, and the technical specs are quite good. One nice touch is that the camera lens retracts into the belly to protect it on landing. Yes, we could do that on our own, but it'd require quite a bit of work. Their mission planner also runs on Windows and Mac and appears to be well designed. And the entire plane fits into a single case, a very desirable feature if you're working a lot of sites. They also seem to have good customer support.

-David

1

2

3

4

5

6

7

8

9

10

## Groups

93 members

1796 members

26 members

72 members

• ### Spanish arducopter Users

107 members

Season Two of the Trust Time Trial (T3) Contest has now begun. The fifth round, 3D model building requiring contestants to survey and reconstruct a building or natural scene from aerial images.

A list of all T3 contests is here.