Hi,
I am researching about long-range stereo vision systems because I am planning to estimate the depth of an object that is +300m away.
I use the method proposed by Zhang in 2000 to calibrate the stereo system for close-range applications, but I got stuck here. For the described situation, I would need a large chessboard printing, which is almost impossible.
Therefore, I started to read some papers regarding self-calibration algorithms, but I didn't find anything applied to long-range situations like mine.
Firstly, do you think that this task is possible? What would be the limitations?
I am planning to work with a small baseline due to my hardware configuration.
I'll be glad if you provide/recommend any information source.
Thanks in advance.
I think you may be interested in the works of "Christopher Hahne" in this field; you may check the following links:
https://github.com/hahnec/plenopticam
https://github.com/hahnec/plenoptisign
https://link.springer.com/article/10.1007%2Fs11263-017-1036-4
What depth precision do you require? Attached, you will find an Excel sheet giving an estimate of the error you might get from a Point Gray BumbleBee X3 which has a 24cm baseline which shows big error at 50 meters.
https://www.flir.com/products/bumblebee-xb3-firewire/
The range you describe is mostly incompatible with a "small baseline" stereo...
Abdullatif Baba thanks for the recommendations, I'll check them.
Bruno Martin thanks as well. I am looking for something precise, let's say
Paula, bom dia! Não sei se entendi bem sua questão. Seroa sobre a metodologia na investigação a profundidade? Se for isso. peça ao sujeito que dimensione numa trena métrica qual a percepção dela (estimativa) da profundidade). Trata-se disso?
Hello Paulo,
You can have a look at the Zed camera. It has a small baseline but it's maximum range is about 20m-30m
https://www.stereolabs.com/
As the depth estimation error goes up with the square of the distance, I'm don't think that stereo is appropriate for the resolution you would need at the distance you want (up to 300m). That's typically a Lidar range...
https://robotics.stackexchange.com/questions/11749/estimate-disparity-error-for-depth-accuracy-estimation
Arthur Alves, eu preciso calibrar um sistema estéreo para medir um objeto com precisão que estará a mais de 300m de distância das câmeras. Não estou encontrando nada sobre aplicações análogas.
Bruno Martin, I was giving a try to stereo due the price of a lidar to cover such range. I didn't know about this relation: As the depth estimation error goes up with the square of the distance. Where I can find studies to show it? It would be good to convince my collegues about the drawbacks of the current strategy.
Hello Paulo Araujo , you can have a look at these:
http://www.cvlibs.net/projects/autonomous_vision_survey/literature/Pinggera2014ECCV.pdf
https://robotics.stackexchange.com/questions/11749/estimate-disparity-error-for-depth-accuracy-estimation
The idea is that very far objects will have very small disparity with your small baseline and that even a sub-pixel error on the disparity will translate in huge error in distance...
Olá, Paulo! veja este artigo? Article The role of perception and action on the use of allocentric ...
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps