Fabian Wiesel

The color and the shape: automatic on-line color calibration for autonomous robots

This paper presents a method for automatic on-line color calibration of soccer-playing robots. Our method requires a geometrical model of the field-lines in world coordinates, and one of the ball in image coordinates. No specific assumptions are made about the color of the field, ball, or goals except that they are of roughly homogeneous distinct colors, and that the field-lines are bright relative to the field. The classification works by localizing the robot(without using color information), then growing homogeneously colored regions and matching their size and shape with those of the expected regions. If a region matches the expected one, its color is added to the respective color class. This method can be run in a background thread thus enabling the robot to quickly recalibrate in response to changes in illumination.

Reinforcing the Driving Quality of Soccer Playing Robots by Anticipation Verbesserung der Fahreigenschaften von fußballspielenden Robotern durch Antizipation

Summary This paper shows how an omnidirectional robot can learn to correct inaccuracies when driving, or even learn to use corrective motor commands when a motor fails, whether partially or completely. Driving inaccuracies are unavoidable, since not all wheels have the same grip on the surface, or not all motors can provide exactly the same power. When a robot starts driving, the real system response differs from the ideal behavior assumed by the control software. Also, malfunctioning motors are a fact of life that we have to take into account. Our approach is to let the control software learn how the robot reacts to instructions sent from the control computer. We use a neural network, or a linear model for learning the robots response to the commands. The model can be used to predict deviations from the desired path, and take corrective action in advance, thus increasing the driving accuracy of the robot. The model can also be used to monitor the robot and assess if it is performing according to its learned response function. If it is not, the new response function of the malfunctioning robot can be learned and updated. We show, that even if a robot loses power from a motor, the system can re-learn to drive the robot in a straight path, even if the robot is a black-box and we are not aware of how the commands are applied internally.

Optimized route network graph as map reference for autonomous cars operating on German autobahn

This paper describes several optimization techniques used to create an adequate route network graph for autonomous cars as a map reference for driving on German autobahn or similar highway tracks. We have taken the Route Network Definition File Format (RNDF) specified by DARPA and identified multiple flaws of the RNDF for creating digital maps for autonomous vehicles. Thus, we introduce various enhancements to it to form a digital map graph called RND-FGraph, which is well suited to map almost any urban transportation infrastructure. We will also outline and show results of fast optimizations to reduce the graph size. The RNDFGraph has been used for path-planning and trajectory evaluation by the behavior module of our two autonomous cars “Spirit of Berlin” and “MadeInGermany”. We have especially tuned the graph to map structured high speed environments such as autobahns where we have tested autonomously hundreds of kilometers under real traffic conditions.

OBJECT CUT AND PASTE IN IMAGES AND VIDEOS

Most image and video editing applications implement only a set of low-level operations, such as linear and non-linear filters, scaling, and simple rectangular cropping. A skilled user is required to combine these traditional tools creatively to perform a useful task. This article presents an example of an approach that takes into account the semantics of the image and is therefore more powerful than traditional tools. Using a model of the foreground and the background of a given image, the SIOX tool (Simple Interactive Object Extraction) allows a user to cut out a certain object from an image or video with very little user interaction. The core of the algorithm is a color clustering approach that has been derived from image retrieval techniques and evaluated against a benchmark proposed by Microsoft Research. The method allows for sub-pixel accuracy, allows video segmentation in realtime (640 × 480 × 25), and is noise robust. SIOX has recently been integrated into the development versions of several open source image and video editing applications such as GIMP, Blender, and Inkscape.

Multi-Exposure Document Fusion Based on Edge-Intensities

This paper presents a new algorithm for fusioning images of text-documents taken with different exposures.It is compared to several standard block oriented exposure- and focus-blending-algorithms.The recognition rate of a publicly available OCR-engine is used as a benchmark to quantify the results.Experiments show in average an improvement in the recognition rate from 0.46 to 0.64 by employing exposure blending as preprocessing step to an OCR. The presented algorithm of blending high-pass filtered images instead of original images further increases the recognition rate to 0.95.