The structure of the internal representation of surrounding space, the so-called cognitive map, has long been debated. A Euclidean metric map is the most straight-forward hypothesis, but human navigation has been shown to systematically deviate from the Euclidean ground truth. Vector navigation based on non-metric models can better explain the observed behavior, but also discards useful geometric properties such as fast shortcut estimation and cue integration. Here, we propose another alternative, a Euclidean metric map that is systematically distorted to account for the observed behavior. The map is found by embedding the non-metric model, a labeled graph, into 2D Euclidean coordinates. We compared these two models using data from a human behavioral study where participants had to learn and navigate a non-Euclidean maze (i.e., with wormholes) and perform direct shortcuts between different locations. Even though the Euclidean embedding cannot correctly represent the non-Euclidean environment, both models predicted the data equally well. We argue that the embedding naturally arises from integrating the local position information into a metric framework, which makes the model more powerful and robust than the non-metric alternative. It may therefore be a better model for the human cognitive map.

The ability to decompose large tasks into smaller subtasks allows humans to solve complex problems step-by-step.To transfer this ability to an automated system, we propose a spiking neural network inspired by the neurobiological mechanics of spatial cognition to represent space on multiple levels of abstraction.As behavioral experiments suggest that humans integrate spatial knowledge in a graph of places, neurons in the state-action network encode locations while connections between them represent transition actions.In a series of simulation experiments, the influence of hierarchy on planning speed and on the resulting route choice in comparison to single-level models is investigated.We find that the model chooses biased subgoals in line with experiments on human navigation.

In this work we present a new approach for the detection of moving objects observed by a mobile camera, which is a critical issue related to autonomous robot navigation as well as driver/pilot assistance systems. In order to separate individual object motions from the self-motion of the observing camera, we implement a linear method to recover the full set of 3D motion parameters undergone by the camera. Based on the recovered camera motion and reconstructed depth information of the detected scene points, a criterion has been derived to determine how well individual scene points agree with the estimated camera motion. The classification of scene points is achieved by statistical analysis of the probability distribution function of the points' motion characteristics. After the initial classification, the identified dynamic scene points are further clustered into different objects by taking into account the underlying geometric distribution in the image. The approach is unique in that it can detect moving objects using a single pair of images and is completely automated. Several experiments have been carried out in challenging environments using two different hardware setups. A comparative study shows that the proposed classification method generates fewer false alarms compared to a standard one.

In this paper, we propose a novel approach for the visual navigation of unmanned aerial vehicles (UAV). In contrast to most available methods, a single perspective camera is used to estimate the complete set of 3D motion parameters undergone by the UAV. We establish robust point correspondences between consecutive image frames captured by the flying vehicle. Based on the estimated motion parameters as well as the reconstructed relative scene depth, a visual steering algorithm has been realized so that the UAV is capable of avoiding obstacles during navigation. The advantage of our approach lies in the fact that decision for collision avoidance is made immediately, by using purely visual information extracted from the live video sequence. Furthermore, it eliminates the time-consuming steps of explicit obstacle recognition and global reconstruction of the environment. Experimental evaluation has been carried out based on computer simulation as well as using a commercially available flying drone. It has been shown that the UAV is capable of autonomous navigation in unknown environments with arbitrary configuration of obstacles.

Proposes an active vision system for saccadic camera gaze shifts and explorative scene analysis as a new integral approach to image understanding. The model consists of two sensory subsystems: preattentive peripheral feature detection and high resolution foveal image identification based on a hypercolumnar representation. Visual objects are non-explicitly stored in two sparsely coded associative memories separating fixation locations for identities of foveal views. An egocentric interest map integrates bottom-up and top-down information sources and decides when to generate a camera movement. A selective masking of preattentive processes supports a cooperation with cognitive object recognition. The system is easily extendible, copes with occlusions and distortions and can be driven in different modes for exploration tasks. This model is able to perform visual search and reproduce findings in the human visual system.>

A stereo obstacle detection system has been developed for automatically guided vehicles that operate on flat (factory) floors. The system does not attempt to reconstruct the 3D environment visually but simply tries to detect obstacles on the floor in the vehicle's path. The approach to stereo image processing uses inverse perspective mappings to facilitate matching of the binocular field of vision against the expected 3D structure of the environment. Assuming a known relative camera model, a geometrical image transformation is computed which essentially compensates the stereo disparities for the image points of the floor. After the mapping operation the images are compared and local mismatches are interpreted as possible obstacle locations. The system has been successfully tested in a factory environment. The implementation runs on standard microprocessor hardware in real time.>

Network equations for the cortical area network (CAN) are presented. The nodes are formed by cortical areas with their intrinsic connectivity and the according computational capabilities. Intrinsic processing is modeled by convolutions. The edges are formed by the mappings between the various cortex areas. With respect to the spatial organization, one can distinguish topographic maps (coordinate transforms), patchy maps that occur when multiple input converges to a common target area, and parametric maps (2-D histograms that encode stimulus into a spatial position). Applications include space-variant image processing and visual receptive field organization