- For accurate mapping it is not advised to use videos: In most cases, the quality of the results will be inferior than those generated using still imagery.
- 4K videos from cameras such as GoPro 4 and DJI provide reasonable results.
- Full HD videos are usually not sufficient to get reasonable results.
Pix4Dmapper can process videos.
These points should be considered when using a video for processing:
1. It is suggested to use videos with high resolution, because even 4K video resolution is only 8 MP, the results may be of lower quality than using images.
2. It is possible to use videos with lower resolution, but it is recommended to acquire the video using 4K resolution. Videos with lower resolution can create a model with very low resolution or fail processing.
3. When adding the video, it is possible to use the .avi and .mp4 formats.
4. Create a project as described in Step 2. Creating a Project but on the wizard New Project, on the Select Images window click Add Video...
5. On the pop-up Select Video, navigate to the video in .avi or .mp4 format and click Open.
6. The pop-up Video Extraction Options appears:
7. (optional) Select the value that defines the interval at which frames are extracted. By default, the value is set to 1. Below the box, the Number of extracted frames for the selected value is displayed.
8. Click OK.
9. On the wizard New Project, on the window Select Images, click Next.
10. (recommended) Select the camera model. When adding a video for processing, Pix4Dmapper will obtain frames but the EXIF ID cannot be fully extracted. Therefore the camera model cannot be automatically selected and some default parameters are loaded.
10.1 If the camera model is in the database: select it from the list.
10.2 If the camera model is not in the database: edit the camera model manually. For more information: How to use the Editing Camera Model Options.
11. Continue the project creation as in Step 2. Creating a Project.
I've been processing from video for over a year now, using this method. I am using the Yuneec Typhoon H with the CGO3 camera. When I first processed a video, I discovered that while turning ( in the case of rotating a multirotor) it causes a stretching or ghosting effect in the output product. After multiple attempts of flying and manipulting the recording on and off, and without a way to establish a programmed flight pattern for overlap, I would fly "straight" lines while recording. Stop recording while turning to align for the next parallel run. In the end, I would look at the number of frames, the length of video ( in time) and calculate X number of frames per second. I would do that to reduce the number of frames to process, and still have a decent overlap... attempting to get 1-3 frames per second. That has worked very well for me, from a hobbyist perspective.
Hi Aron,
Thank you for sharing your experience and expertise :-) This is certainly helpful.
hello,
i am trying to process videos from hero5black.
first i tried by extracting frames with every 1 frame and i got error with geo-location and also camera model can't be set
then i tried by assigning geolocation from file which is extracted from video in gpx format although i am unable to set camera model and with processing default camera model i got unsatisfactory result.
Hi Anil,
As far as geolocation error.. I get the same. I've recalibrated the camera gimbal several times, which hasn't helped.
What are you trying to use it for? I was doing it to see IF I could develop a 3D image as far as personal interest, not for any sort of mapping or commercial project. It has been quite a while since I've processed any video.
Are you using the GoPro on a drone ? I've tried using my cell phone images also, with limited success. I have yet to try using video from cell phone.
hi Aron,
thanks for reply, actually i m trying process GoPro video for 3D modelling , same kind of work i have done with satisfactory result from Dji p4p video.
You're welcome. I've never used GoPro products. All the GoPro viseo's I've seen, attached to planes , drones, skydivers, etc...seem to be warped..or fish eye. I'm wondering if that may be the case with yours? Even if it is miniscule...or barely perceptable. Perhaps Pix4D can chime in on this aspect.
If I resolve my geo-spatial issue with the CGO3 on the Typhoon H platform, I'll post.
Pix4D Mapper can handle fisheye cameras as long as it is properly declared in the camera model. There are two options when selecting a custom camera model (perspective and Fisheye) so if you haven't selected Fisheye then give that a try. Also, the camera model parameters are likely only there for still image resolution. You may need to generate a camera model specifically for your GoPro model. You can read more about this process below. It basically consists of capturing a high quality, easy to calibrate dataset and use the resulting optimized internal paramters to save a custom model you can then use on later projects.
https://support.pix4d.com/hc/en-us/articles/202557009-How-to-calibrate-a-Fisheye-Lens-Camera