|
|
Jens KÖRTEUniversity of Kaiserslautern Informatics
Faculty
|
Project Thesis: "Sensor based Collission Detection" (Presentation slides) (Part of SIMERO Project)
How about using cameras to detect whether there will be a collison or not. By using pictures which show the actual scene
(including robot and obstacles), pictures showing the robot only and pictures that show the robot in the future (i.e. the
position the robot will have after moving) you can decide whether there will be a collision or not.
The only problem to solve was to detect obstacles in front or behind the robot. I used to handle this with the fundamental matrix
which provides information about 2 cameras (i.e. the transformation from one into the other camera). By using the 8-point algorithm
I used to get the fundamental matrix.
Some tests have been done. First to know how much cameras we will have to use to get the best result and second where to put these cameras
to obtain the best result.
Last update: