This is the source code for the UT Austin FRI project by Victor Yap, Eysa Lee, and Daniel Cheng, in the BWI research stream.

This project involves the detection and tracking of moving obstacles at varying depths using a Kinect sensor.

Dependencies: OpenNI, OpenCV

This node reads data from '/camera/depth/image' and publishes to '/obsnav/depth_blobs/threshold' and '/obsnav/depth_blobs/blobs'.

To start this node, run

rosrun obstacle_navigation depth_blobs

The topic '/obsnav/depth_blobs/threshold' displays a depth threshold image. By default, it shows objects between 0 and 1 meters of the Kinect sensor. As of now, this range can only be changed by modifying the 'depth_blobs.cpp' source code and recompiling. The variables 'slice_count' and 'slice_size' in lines 16 and 17 determine the number of depth slices and their precision. The slice that is displayed to '/obsnav/depth_blobs/threshold' is chosen in line 182.

Information about the largest blob in each slice can be found in the topic '/obsnav/depth_blobs/blobs', which uses a custom blobs message.

Possible future improvements include:

• Tracking moving obstacles
• Using motion information at each depth to update the cost map used for autonomous navigation

Additional information about this project can be found on the wiki page.