Kohji Kamejima

Image based operation: a human-robot interaction architecture for intelligent manufacturing

A method for interactive robot operation, based on human mental images, is presented. By introducing a multidimensional information structure into a coherent intelligence process architecture, consistency between the system model and the real world is successfully maintained. A mental image of the task is evoked by combination of a visual work space image and action knowledge. The visual image is modeled by transition of a frame system, called the work space model, using a dynamic model matching process. Action knowledge is modeled by a transition network, called the action network, and a schematic description, called the motion schema. The proposed method was verified through a prototype and experiments.<<ETX>>

Generic Representation of Self-Similarity via Structure Sensitive Sampling of Noisy Imagery

An adaptive sampling scheme is presented for discrete representation of complex patterns in noisy imagery. In this paper, patterns to be observed are assumed to be generated as fractal attractors associated with a fixed set of unknown contraction mappings. To maintain geometric complexity, the brightness distribution of self-similar patterns are counted on 2D array of Gaussian probability density functions. By solving a diffusion equation on the Gaussian array, capturing probability of unknown fractal attractor is generated as a multi-scale image. The totality of local maxima of the capturing probability, then, yields a pattern sensitive sampling of fractal attractors. For eliminating background noise in this sampling process, two filters are introduced: input filter based on local structure analysis on the Gaussian array, and, output filter based on probabilistic complexity analysis at feature points. The sampled image through these filters are structure sensitive so that extracted feature pattrers support invariant subset with respect to mapping sets associated with observed patterns. As the main result, a generic model is established for unknown self-similar patterns in background noise. The detectability of the generic model has been verified through simulation studies.

From self-navigation to driver's associate: an application of mobile robot vision to a vehicle information system

The concept of a vehicle information system, called the driver’s associate, is presented for identifying a driving environment in cooperation with human drivers. Sharing the image feature extracted and segmented in observed imagery, a set of frustration resolution schemes for top-down processing and 2-D syntax analysis for bottom-up processing are integrated into a dynamic perception mechanism that is implemented by a set of LSIs (Large Scale Integration chips). The schematics of the software are verified through simulation studies to demonstrate the basic operation scenario of the driver’s associate.

Dynamic work space model matching for interactive robot operation

A dynamic workspace model-matching method for interactive robot operation is proposed. The matching mechanism is realized by combining a low-level real-time pattern-matching process with the workspace model described in a frame system paradigm. In this method, the matching process is dynamically controlled by an interactive motion generation process. The workspace model gives an effective prediction, called workspace context, to the matching process. This prediction reduces three-dimensional location matching to two-dimensional pattern matching. The low-level pattern-matching process is implemented in a recursive algorithm using a pattern kinetics matching method and combined with the interactive robot operation system. The method was verified through implementation and experiments.<<ETX>>

Image based operation for human-robot interaction

A prototype system for human-robot interaction based on the mental image of the task is described. The mental image is a visual model that reflects a real-world situation, such as object placement in the work space, and simulates the effects of robot action in this situation. A human operator estimates and plans the task using this mental image. The system architecture used to produce the mental image for the operator is presented.<<ETX>>

Knowledge representation for image based robot operation

A knowledge-representation method is proposed for understanding operator messages to generate motion in a robot mechanism. With this method, a generation process which activates a set of rewrite rules, translates the operators message to a sequence of symbolic motion modules called motion schema. Further, an instantiation process of the motion schema, which binds schemata with geometric data stored in the work-space model, generates instant motion. Through an experimental study using a prototype system, certain aspects of the method were verified.<<ETX>>

Multiparticipant Geographics Annotation for Interactive Rendezvous and Cooperative Monitoring

A new framework is presented for multiparticipant coordination of over-the-horizon maneuvering processes. In this framework, geographical information is decentralizedly augmented via the multitude of annotation processes: landmark localization by map builder, connection generation by planner, and GPS tracking by probe vehicles. By integrating the augmentation process on a common satellite image, the subscriber participants reuse the geographics within specific maneuvering context. Based on graph theoretic representation of the multiparticipant augment process, an interactive geographics annotation system was developed and verified within the context of interactive rendezvous and cooperative monitoring.

Image-Based Satellite-Roadway-Vehicle Integration for Informatic Vicinity Generation

A new framework is presented for integrating satellite/avionics sensors with onboard vision for expanding the horizon of perception. Real time bindings of the birds eye image and the drivers view via GPS provides informatic vicinity for supporting in situ decision processes. As the invariant feature for multi-viewpoint association, chromatic complexity of roadway texture is extracted from frontal scene and adapted to the birds eye imagery. The feasibility of the framework with requirements for vision system is discussed through simulation and experimental studies

Recycling on network: an information-control architecture for ecologically-conscious industry

An inter-site communication framework is presented for the computer integration of product manufacturing and resource recovery operations. Product knowledge is transferred via the inter-site network to induce the incentives for recycling.<<ETX>>

Saliency-based boundary object detection in naturally complex scenes

A stochastic scheme is presented for cooperative detection of landmark objects distributed in roadway boundaries. By indexing chromatic diversity within a locally Gaussian color space, saliency patterns are extracted with respect to the as-is primary system. Through saccadic scan of the saliency patterns, boundary objects are successively articulated into a system of fractal attractors consistent with the ground-object structure. As the result, the fractal model is indicated within the perspective of the naturally complex scenes.