Forest orthophoto showing only partial results

Hi guys, I have a problem with orthophoto processing. I did full coverage of 100ha (1km²) area with default overlap, ~ 8cm/px GDM. Result looks like this:

 

link:

https://cloud.pix4d.com/pro/project/229094/map?shareToken=dcf741f82129480ebc3f6081f943157f

I need full area orthophoto. Is it possible to improve this reults? How?
This is critical for me, any help would be great.

 

Forest zones trie to increase the overlap until 85-90 and the hight of the flight at least 30m over the tallest tree

My altitude was 267m (230m above trees), side/front overlap 65%/75% , speed 15 m/s.

I will retry with more overlap. Key problem here is that I have really large area to cover (100 km²) in relatively short time (month), low budget. I’m looking for parameters to give me successfully stitched maps done with 1-2 batteries max per km².

I’m working non-profit for nature conservation organization, we are gathering data about illegal forest logging done by our own government (i know it sounds crazy, but this is reality in Poland, you can read about it here https://www.theguardian.com/environment/2017/nov/21/poland-faces-100000-a-day-fines-over-illegal-logging-in-biaowieza-forest).

Hey Alx,

I am not very knowledgeable about the Pix4D cloud. What processing options are you using? And are you able to change the processing options? If you use Pix4D Mapper, I know there are few different parameters that you can tweak in order to get a better stitched orthomosaic. 

1 Like

Hi Alx, thanks for sharing the project information and the background story.

The problem for reconstructing such areas is to match keypoints between images. The geometry of the forest is very complex and generally creates many keypoints in an image, but it will be difficult to find the same keypoint in the next image. Increasing the overlap can help for this, for such areas a 85% frontal and 70% side overlap is recommended. Another way to improve is to reduce the detail in each pixel, i.e. increase the GSD either during the acquisition or in the processing options.

If the goal is to identify areas of illegal forest logging, I would argue that a less detailed GSD would still work for the purpose. For example 10cm/pixel or 15cm/pixel should still enable you to identify those areas. Hence, if you use the same drone and camera I would fly even higher than you do (if that is possible with the regulations of your country). Otherwise, you could use a different camera, so that you can fly lower but keep a similar GSD. 

I tried to use 1/2 and 1/4 for the Keypoints Image Scale parameter in the processing options of step 1, but it did not significantly improve the reconstruction. In my opinion this shows that a higher overlap would help. There are more tips for modifying the processing options here: https://support.pix4d.com/hc/en-us/articles/202560159

@Selim: You can change the processing options on the desktop and then upload the project to the Cloud. This is how you can modify the options in the Cloud processing. More about this here: https://support.pix4d.com/hc/en-us/articles/115004575526#label3

Information on image acquisition for forest and dense vegetation: https://support.pix4d.com/hc/en-us/articles/202557459#label2

Good luck for your project. Please let us know if there are more questions and keep us posted about how this turns out.

1 Like

Dear Pierangelo,

Thanks for your reply. I have never used the Pix4D Cloud but I am definitively going to look into links that you have provided.  

2 Likes

@ Pierangelo Rothenbühler Thank you very, very much for your answer. I used pix4d desktop today and i think it helped me to identify the issue here. Processing report is a great feature in desktop version!

I captured this dataset using Drone Deploy which does really terrible job and as it turns out in pixd4d desktop report, I have a lot of “unconnected” photos without any overlap at all in this dataset. I think this a obvious reason results are so poor. It would be impossible to spot this by eye, without pix4d desktop.

If the weather will be good I’m going to shoot new dataset tomorrow, using pix4d app and settings that you suggested.

off-topic:

Is it possible to apply for any discount if we are going to use your software in non-commercial work? We are all volunteers, working pro bono and using own equipment, private time and resources, trying to stop logging in the europe’s last primeval forest. We are working against our own government, including police and against time.

I have tried bunch of different softwares ranging from Drone Deploy, Agisoft PhotoScan, DroneMapper to OpenDroneMap, but pix4d seems to beat them all in terms of ease of use and quality, and support. Unfortunately price is very high for us.

1 Like

You are welcome Alx and Selim! 

I would like to point out that you can download the quality report in your cloud project as well by clicking “Download” in the top right corner and the selecting “quality report”: 

As for the non-commercial use of the software, there is such a version. Please contact sales directly with your request. You can link to this community post in your email to provide more background for my colleagues in Sales.