Support Website Contact Support Blog

GoPro on a pole?

I’m looking for advice on aerial scene mapping using a GoPro Hero 6 on a pole instead of a drone. Currently I have a 10’ pole extended with an 18" Manfrotto magic arm extending out from the top of the pole with the GoPro mounted on that tilted slightly up resembling the 70° position in which the DJI Phantom 4 would capture the scene for 3D mapping. Using .5 sec time lapse I walk in a grid pattern the same as the drone would getting easily 80% overlap. When processing the iimages in PIX4D Mapper Pro I keep getting the errors “no blocks computed” and e0046 “no calibrated cameras” even though it recognized the GoPro Hero 6 in setup. Any advice?

You need more overlap. You shoul have a lot of overlap and with a pole in your hand the photos will be not stable as you think. It will move a lot.

So maybe you should walk slowly for the overlap can reach at least 90%. And you should make a timlapse for 2 seconds. If you see you got too much data (photos), then you can select the photos to use.

That’s my point of view but I never used a camera on a pole before.

1 Like

Hi Jamey. I was able to respond via a support ticket but if anyone else is interested in a similar workflow, here is my response. 

I would agree with Jorge that the likely reason for non-calibrated cameras is lack of overlap. 

Step one is not completing because the cameras are uncalibrated and as such Pix4D cannot begin matching. Similarly, I would suggest that you should try moving slower or increasing the capture interval. It could also be that the images are blurred or overexposed. Check the images to ensure that they are clear and properly exposed.

Alternatively, you could also try capturing video as this would allow you to fine-tune the overlap by selecting the image select interval when importing the video. For more information about processing, video see the link HERE. Note that images extracted from video frames will not be geolocated so the resulting project would need additional steps to scale and geolocate. 

If anyone else has thoughts or experience with this sort of workflow feel free to provide additional advice.

Thank you. Yes, I did walk slower to get more overlap and I did get better results. Thank you!

Excellent! good to hear.

1 Like

Hello 

 

with goPro hero 5

I take picture every 0,5 second

FOV large

12 MP

walk slowly

and i have this erreur “no blocks computed” and e0046 “no calibrated cameras”  too

the pictures looks ok

link for picture https://we.tl/t-ejyr2CqYDS until  17 september

dany.lothaire@province.namur.be

 

Thank you for help

Hi Nicolas,

I was able to get the images to calibrate using the following Step 1 processing options. 

_ Keypoint image scale = Full_

_ Matching Image Pairs = Free flight or terrestrial_

_ Calibration Method = Alternative_

_ Internal parameters optimization = All Prior_

 Despite being able to calibrate there is a substantial amount of noise generated by the sky. You can use the carve annotation tool to help reduce the noise before running step 2. Also, I would recommend creating a few MTPs as that also helped improve reconstruction. Below is an image of what I was able to generate. You may be able to improve results by increasing camera height and pointing towards the ground to reduce sky. Multiple passes at different heights would also likely improve reconstruction. 

Hi Holden,

When i change option processing, it’s ok.

Thanks you for informations.

Best,

Hi Holden,

I have proccess, i have picture.

But very far to

https://medium.com/the-science-of-drone-mapping/no-drone-gopro-ahead-6e434546aacc

You have more informations about this project ?

Thanks

Best,

I think the main differences are the number of images and overlap. Also, there is some image redundancy in the project from the Medium article which can help also. Consistent and high overlap with minimal sky will produce the best results. Also ensuring that there is a continuous path of images as they are taken. No large gaps will help the images orient relative to each other. 

Thank for answer,

I continue my test.

Good Luck!