Drone topographical survey for agriculture in Africa - What have we done wrong? Business opportunity

We (www.gmxconsulting.co.uk) are developing a large-scale agriculture project in Nigeria (5,000ha) and started to use drones to survey and map topography.

The intention was then to use the Digital Elevation Model for irrigation planning and design. However, when we post processed the images we got very large x y z errors - please see the photos below.

1. Could anyone tell us where we have done wrong? Can we correct NOW (after the aerial survey)?

2. Also we look to outsource the drone surveying and post processing for our projects in Africa - Anyone interested please let us know.

Here are some more details:

Platform: E384

AGL: 120 - 180m 

Overlap: Front 60% Side 60%

GCPs: No

Flight time: <60 mins

Cameras: Canon S110 RGB and NGB

Number of photos: 300-400

Photo taken out before processing: None

Software: Photoscan

Many thanks in advance!

Views: 3374

Replies to This Discussion

Hi Quan Le,

First of all, it looks like worse precision areas are located in the border of your mission region, probably where the plane turns and where you are taking (and afterwards processing) no nadir images (ie, not zenital, with large camera location error). In those cases you can exclude these images from your analysis in Photoscan. That should improve overall precision of your models.

If that region is important for you, you must extend your missión farther next time to keep that area in level flight (ie, make the turns outside there). You can also increase image overlapping to 70 or 80% in at least one side. If your camera is fast enought that should not increase your flight time very much (you can also flight slightly slowly, if possible).

Another way to correct that is using some GCPs at easily recognized items of your images. Even with a conventional GPS you will improve your models (at least, its geographical position).

Hope that helps you.

Regards

Thank you Eladio for your advice. We thought lacking GCPs could be the issue. As this was a remote area we didn't have a recognised datum so could never quite confirm the exact geographical position reference point. We will next time use DGPS first to work out "where we are in the world" then put down the GCPs as recommended. We will also try to delete the photos around the edges. All the images looked quite sharp though.

We were reluctant to increase the overlapping because of flight time constraints. We did see that people use 80% for both. That would increase the processing time significantly though.

When you process the image with Photoscan, does quality option 'medium' or 'high' affect precision? Again we were limited by computing power - just a Quadcore i7 16GB Ram laptop -  so couldn't go for the 'high' option. Any suggestion on building a cost-effective high-power processing hardware would be appreciated. Or can we lease computing power?

Hi Quan,

I really love business opportunities and 

like your

Drone topographical survey for agriculture in Africa

challenges 

since as Climate Change Consultant with UNFCCC I join projects by UN agencies,

World Bank, UN Foundation and other public agencies.

I need to study a number of your images, geolocated as Google Earth/ Maps overlay first.

I need geolocations for your regions to study with available Google Map, Earth, OpenMap mapping systems and more.

Could you provide me with technical specification of your drones to let me check if they are fit to accomplish your task.

Let me know your deadlines now and size of your imaginery database in MB, GB

and if you can upload images over the Internet if necessary.

I am sure, we can get a lot of invaluable materials from IRENA in support of your irrigation planning and design. 

pls email me

darius

CChC

manta103g@gmail.com 

Thank you Darius for your comments. I will try to send a link to dropbox folder containing the raw images and processed data from Photoscan.

You will see a google kmz file with location on Google Earth.

For drone we used E384 https://www.event38.com/Articles.asp?ID=289

Thank you Quan,

pls email me directly.

darius
manta103g@gmail.com
Thank you Quan,

pls email me directly.

darius
manta103g@gmail.com

Hi Quan, I am looking to buy E384 and would love to learn your experience with them so far. I am also available at ruchit@harvesting.co

Hi Quan Le,

We are a Survey Company located at Paarl in South Africa and have been using UAV's for the last 3 years to do Aerial Topographical Surveys throughout Southern Africa. It has been an expensive "learning curve" through which we had to go to iron out mistakes and get a working "Work Flow" in place.

At this stage GCP are necessary to have an accurate survey of the area that has been flown, but we are looking at possible solutions where GCP are not necessary. The method entails having a Base Station on the ground and then a Rover on your Plane.

You could correct the Survey by visiting the area again and fixing (X, Y & Z) features over the area that are visible on your Photos.

We would be very interested in doing the Surveys for you and we can be contacted at info@aerialpix.co.za.

Regards

Michael

@Quan,

files received, 2 files parsing problems.

email me back

manta103g@gmail.com

Hi Quan Le

Like many people have mentioned, work in Africa can be exciting.

Also, like many people have mentioned Ground Control Points are necessary for accurate results.  A recent project we complete was to generate a flood line assessment for a mine in South Africa.  Through GCP's we were able to improve our results down to 1.5cm accuracy (x,y & z).

Saying this, the density of the vegetation on the ground will also influence the accuracy of your results, as you mentioned you need to plan for irrigation.  At the end of the day the combination between using a UAS/RPAS and an accurate survey grade GPS will give you the results you want.

Give me a shout if you have anymore questions.

Regards,

Gareth 

Hi Quan Le,

I agree with the above comments. You need accurate ground control points if you are looking to produce an accurate model. The number and positioning and accuracy of the GCPs is also important and depends on the topography. This is something that comes with experience. Another aspect that hasn't been mentioned is image overlap. For consistent results, I would recommend a forward overlap of at least 80% and side lap of 75%. With a simple terrain you could probably get away with less, but rather safe than sorry. Their are other factors such as weather conditions, flying speed, camera settings that all come with experience.

Having said that, you could probably get reasonable results from your images just by adding 5 or 6 well positioned ground control points.

Let me know if you need any more info. You're welcome to contact me any time at ian@aerovision-sa.com.
We are an aerial surveying and mapping company based in South Africa.

Regards,
Ian Freemantle

There is a way for you to improve your UAV produced photomap precision significantly (cm accuracy level) without the need to use any GCP.  Check this link out. Or, you may try this for RTK GPS on your plane.

RSS

Groups

Season Two of the Trust Time Trial (T3) Contest 
A list of all T3 contests is here. The current round, the Vertical Horizontal one, is here

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service