Ask Your Question

Detect moving objects on an image with an moving camera

asked 2016-09-07 04:01:12 -0500

svenbob gravatar image


I am currently working on a project using ROS (Robot Operating System) and OpenCV methods to implement a navigation algorithm to help blind people avoid obstacles. I am using a Kinect v2 to generate a 3d Map based on the SLAM algorithm. I therefor compute the visual odometry and map the color and depth data. Then i detect static obstacles on this map and plan paths around them.

Now i want to expand the project to detect dynamic obstacles to predict and avoid possible collisions. I want to detect this object on the image frame of the kinect. Is there a way to compensate the camera movement with the known visual odometry data? My problem is, that i can detect movement on the images but I am not able to segment objects from the image. Then I want to reconstruct the 3D object positions and transform them to the map coordinate frame.

It would be great, if anyone has any ideas on my thoughts or can help me out.

Thanks in advance and greetings,


edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2016-09-07 19:05:47 -0500

Tetragramm gravatar image

The simple answer is Optical Flow + Depth gives you 3d displacement vectors. Cluster those, and you have your background and your objects. Of course, doing that well is difficult.

Did you perchance try Google the question to see what has already been done? Here are results 2, 4, and 5 from the search for moving objects from RGBD data.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-09-07 04:01:12 -0500

Seen: 3,081 times

Last updated: Sep 07 '16