Honda Research Institute Europe Gmbh

Edgar Körner

Andreas Richter

Bernhard Sendhoff

Christian Goerick

Heiko Wersing

Bernard Sendhoff

Marc-Oliver Gewaltig

Ursula Körner

Iterative (nondeterministic) optimization of aerodynamic and hydrodynamic surface structures can be accomplished with a computer software program and a system using a combination of a variable encoding length optimization algorithm based on an evolution strategy and an experimental hardware set-up that allows to automatically change the surface properties of the applied material, starting with the overall shape and proceeding via more detailed modifications in local surface areas. The optimization of surface structures may be done with a computing device for calculating optimized parameters of at least one (virtual) surface structure, an experimental hardware set-up for measuring dynamic properties of a specific surface structure, and an interface for feeding calculated parameters from the computing device to the experimental set-up and for feeding measured results back to the computing device as quality values for the next cycle of the optimizing step.

Autonomous experimental design optimization

One embodiment of the present invention provides a method for gathering information from an environment. In a first step, visual information is gathered from the environment. In a second step, information actively transmitted by objects in the environment is received. According to one embodiment, the information actively transmitted by objects in the environment is received wirelessly. In a third step, the visual information is combined with the received information in order to recognize objects.

Integrating visual and object information in a pervasive computing environment

For object recognition, an image is segmented into areas of similar homogeneity at a coarse scale, which are then interpreted as surfaces. Information from different spatial scales and different image features is simultaneously evaluated by exploiting statistical dependencies on their joint appearance. Thereby, the local standard deviation of specific gray levels in the close environment of an observed pixel serves as a measure for local image homogeneity that is used to get an estimate of dominant global object contours. This information is then used to mask the original image. Thus, a fine-detailed edge detection is only applied to those parts of an image where global contours exist. After that, said edges are subject to an orientation detection. Moreover, noise and small details can be suppressed, thereby contributing to the robustness of object recognition.

Preparation of a digital image with subsequent edge detection

Within the frameworks of hierarchical neural feed-forward architectures for performing real-world 3D invariant object recognition a technique is proposed that shares components like weight-sharing (), and pooling stages () with earlier approaches, but focuses on new methods for determining optimal feature-detecting units in intermediate stages () of the hierarchical network. A new approach for training the hierarchical network is proposed which uses statistical means for (incrementally) learning new feature detection stages and significantly reduces the training effort for complex pattern recognition tasks, compared to the prior art. The incremental learning is based on detecting increasingly statistically independent features in higher stages of the processing hierarchy. Since this learning is unsupervised, no teacher signal is necessary and the recognition architecture can be pre-structured for a certain recognition scenario. Only a final classification step must be trained with supervised learning, which reduces significantly the effort for adaptation to a recognition task.

Growing community of inventors

Seligenstadt, Germany

Edgar Körner