KINECT SENSOR CALIBRATION

Language: C++

Environment: Visual Studio

Duration: < 1 year

Team Size: Independent project

I had to develop a program, using openCV and C++, that could calibrate the depth and color sensors of a Kinect camera. Then we combined those outputs into an RGB-D data file.

First part of the program involved finding the 3D coordinates of the corners of the checkerboard. We could then detect the board corners, and thus the board pose. After many more transformations, and many more PnPRansac functions, we could compute the external camera parameters, and then paint the depth image onto the external images.

One of the main challenges with this project was the artifacts in the depth images. Many times corners could not be detected with the standard procedure due to missing corners, incorrect depth images, and so on. To fix this I added heuristics that would search for depth not just at the edge of the checkerboard but inside as well. By interpolating with corners that were not missing, I reconstructed the missing ones.

Overall, this was a fun project, and I look forward to experimenting more with the Kinect.