Best way to detect different objects on image

asked 2020-01-02 08:33:31 -0500


I'm trying to build a Raspberry Pi car, that detect objects in its way while driving on a white floor and then tries to avoid this object. It is also equipped with a distance sensor, so I know from the value of the distance sensor when there is a obstacle.

It has to be able to detect any object that is in front of it. So I can't use methods which will need a training set of possible objects.

I tried thresholding the white floor, so that what's left should be the object. This works sometimes but it was very depending on the lighting. It would only detect the object under certain lighting. Maybe there's a solution to fixing this or there's a completely different approach to this.

Also it has to be possible, that the object can partially leave the camera and still be detected.

I hope someone can help me out with this one, thanks a lot. :D

edit retag flag offensive close merge delete



It has to be able to detect any object that is in front of it.

unreasonable wish

I tried thresholding the white floor,

please try to read up on "object detection" and cnns (like SSD,RCNN,YOLO), before trying to hack at it

berak gravatar imageberak ( 2020-01-02 09:29:30 -0500 )edit

I did research towards this. But as I understand I need to have some training data to make those algorithms work, which would mean that I could only detect the objects, which I "gave" the algorithm beforehand or am I missing something?

itslindi gravatar imageitslindi ( 2020-01-02 10:12:24 -0500 )edit