Hi all i'm using raspberry pi 2 with raspbian , i'm working on face and eyes detection using opencv and c++ , i want to expand the field of view for the pi official camera module any suggestions will be appreciated :)
1 Answer
if you expand the field distortion will occur and it will make the process of pattern recognition much harder. I think you should stick to the original.
But if you have already decided to use a lens, https://www.youtube.com/watch?v=IPTU8NrVqmQ in the video you can see a view with a fisheye lens (which is for iphone), and also you can see the distortion. You can also check these http://www.raspberrypi-spy.co.uk/2013/06/cheap-interchangeable-lenses-for-the-raspberry-pi-camera-module/ , but they have distortion as well which is inappropriate for recognition applications.
-
OK thanks you i haven't think this way , so there is no lens that don't produce distorsion ?? pss : i'm soorry i can't vote up now i'm still having less than 15 reputation i will do it when i reached it Commented Jul 21, 2015 at 15:54
-
Of course there are so said "distortionless" lenses, but as you can see in the link aliexpress.com/item/… they are many times more expensive and if you go to bottom of that page, you can see that there is still distortion, and also there are softwares for distortion recovery but it will be too much work, cost, and the software will be too much for Pi, and also I think camera module's field of view isn't that bad, its okay I hope it helped you. Commented Jul 21, 2015 at 16:06
-
You can also check this link for a further understanding of distortion photographylife.com/what-is-distortion Commented Jul 21, 2015 at 16:07
-
you helped me a lot thank you ... you're right if the image is distorted i can't detect faces so i think i mustn't use a lens to expend the view angle i will stay with the default angle pf view ... however can i use some ZOOM lens without having a low quality ?? Commented Jul 21, 2015 at 19:05
-
1i got it now so i will use a zoom lens because i need to detect faces that are far for about >= 2 m Commented Jul 21, 2015 at 19:46