2005 DARPA Grand Challenge Source Code

The 2005 DARPA Grand Challenge is a 132 mile race through the desert with autonomous robotic vehicles. Lasers mounted on the car roof provide a map of the road up to 20 meters ahead of the car but the car needs to see further in order to go fast enough to win the race. Computer vision can extend that map of the road ahead but desert road is notoriously similar to the surrounding desert. Various machine learning algorithm (Classification and Regression Trees) provided a machine learning boost to find road while at the same time measuring when that road could not be distinguished from surrounding desert.

Source code, videos, and grundtruth data of

Bob Davies and Rainer Lienhart. Using CART to Segment Road Images. SPIE Multimedia Content Analysis, Management, and Retrieval 2006, 15-19 Jan. 2006, San Jose, 2006.

can be download from here:

• Source code
• Groundtruth data and videos

Sub-pixel feature detection via openCV

With openCV it is - besides many other functions - possible to detect features in correspondent images with a sub-pixel accuracy. The optical flow in this function (CalcOpticalFlow) is limited to simple translation. We have done some work to extend the function to affine transformations and extended the algorithm by adding illumination normalization. The new function (CalcAffineOpticalFlow) was already part of openCV, but not activated due to some bugs, which are fixed in our version. Some additional hints how to modify your openCV code can be found in following page: cvlkpyramid.cpp-Notes. The source-code [1] is availiable online, too.