algorithm to computer vision obstacle detection ??
My graduation project is indoor navigation system for blind people so I need to implement obstacle detection using android and open CV, I followed these steps but accuracy is not good : 1- convert to grayscale 2-canny function to detect edges 3- dilate the detected edges by 500% 4- blur the image using a Gaussian blur 5- use findContours function to detect closed shapes and draw bounding circles around the shapes any help with paper or algorithm I should follow??
Only one camera is used? If a stereo camera setup can be considered, I would use something like this article (Barry15a): as stereo matching should be costly, you don't need to compute the disparity for every distance and you can ignore obstacles farther than a threshold and just consider a limited range for collision detection.
Yes I use only my mobile camera
You are aware that with only one camera, you will not be able to get the depth easily (one obstacle in front of you can be near or extremely far)? Cannot help you with monocular obstacle detection, you should try with google scholar.
With a single camera you won't get information on depth AT ALL. Seems like your strategy is gonna have a hard time if you find a small obstacle very close and a bigger one lets say a few meters away...how are you gonna face this situation? I'd rather move to a stereo setup for your purpose...
I was thinking about structure from motion, that's why I said "not easily". Depth can be reconstructed up to a scale with structure from motion but coupled with other sensors or with an initialization step, the full depth can be estimated.
My objective is to make a simple blind assistant I am using "indoor atlas" for indoor navigation (mobile application) and I am trying to implement obstacle detection while positioning and tracking so the camera will be with some angle in front of him