can i set Image ROI in sensor so that i can get more then 30 FPS . as i need to have more frame rate . does this sensor support custom ROI setting in sensor.
You just need to set the resolution to less than or equal to 320x240 for more than 30 FPS. As for a custom ROI, please use the set_windowing() method.
I am working on a split time project. And, is planning to use several openMV(s) to capture some particular lines such as finish line.
By applying image.difference(image) continuously to check sudden change to the image that aims to detect any object cross the line.
My question is if set_windowing() method is really helpful to the reduce the image size and so to the computational cost? and if any suggestion to motion detection?
I would like to constraint the ROI (say 158, 70, 4, 100) to achieve high frame rate (say 200fps or even 500fps) for triggering third party system.
It is possible to do so?
The camera data stream is about 90 FPS at max. At best you’ll hit in the 40+ FPS area in a real situation. If you want 200 FPS to 500 FPS you’re looking at buying an exotic system that’s going to costs tens of thousands of dollars.
Anyway, attached is the latest OpenMV Cam firmware with a new method that lets you do in memory frame differencing. You can hit 40 FPS while actually looking for changes between the difference image and it requires no SD card.
firmware.zip (2.02 MB)
My openMV is shipping to me at next year. I don’t have the hardware and try it right now.
As mentioned, the new method is about memory frame differencing.
It looks a low level manipulation and find the difference with a whole image.
Is it possible to further increase the FPS by limiting the ROI?
Or, is there any method to handle those memory at relatively low level?
The maximum frame rate out of the camera is about 90 FPS. If you limit the amount of pixels you look at to something like 40x30 then you can likely hit 80+ FPS from the camera data stream. But, that’s about as high as you’ll get. For 200 and 500 FPS operation you need more specialized hardware.