# real-time stereo vision system for long range (up to 100m) depth estimation

I want to implement a FPGA-based real-time stereo vision system for long range (up to 100m) depth estimation. Also, I want to use two ip cameras in the system. I have calculated depth error using the equation below with these parameteres:

baseline = 1m (at the expense of increased frontal blind zone area), z = 100m, f = 4mm, pixel_size = 4um, disparity_accuracy = 0.25px,

depth_error = dz = ((z^2)/b.f)*dp = 2.5m

My questions are: Are these calculations and equation above valid for long range stereo vision?

Are there any other considerations for designing this long range stereo vision system which may not be important in typical (short range) stereo vision systems?

I will be grateful for any information you can provide.

• Are you sure about your focal length? 4mm sounds too short if you want to see something in this distance. Mar 6 '17 at 18:17
• @FooBar : I am a little confused. Focal length has nothing to do with subject distance and just determine field of view. Am I right? Mar 6 '17 at 22:15
• exactly. and if your field of view is too large, objects in a large distance will only have some pixels on them so your matching will be difficult. Mar 7 '17 at 7:47
• Thank you @FooBar . So, using a camera with larger focal length have better results in larger distances. Is there any side effect for using larger focal length in long range? I have asked this question because it seems everything becomes better for long range if we use a camera with larger focal length? Mar 7 '17 at 8:10