The New Generation of Automatic Landmark Detection Systems: Challenges and Guidelines

Abstract

Landmarks are salient objects in an environment. They play an important role in navigation by serving as orientation aids and marking decision points. Recently, there have been several efforts to design methods to automatically designate certain buildings with salient features as landmarks. All of these methodologies consist of similar steps: (a) establishing a neighborhood, usually around an intersection, (b) performing statistical or data mining analysis to find the building with outlier characteristics, and (c) establishing this salient building as the local landmark. Although these advances are significant, we believe that there are still several key issues that need to be fully addressed in order to realize the new generation of Automatic Landmark Detection Systems (ALDSs). Currently, the main shortcomings in the domain of ALDSs is the lack of a thorough and systematic study of attributes of objects that are analyzed to select landmarks, and deficient experimental verification of the benefits of ALDSs to the end users. Unless, these shortcomings are thoroughly addressed, the viability, applicability, and usefulness of ALDSs are uncertain. On the other hand, automatic landmark detection has the potential to be a dynamic, fascinating, and interdisciplinary research topic with wide applicability. Therefore, the goal of this paper is to discuss the current shortcomings in the domain of landmark detection, propose some preliminary solutions, and provide general guidelines for implementation of the new generation of ALDSs. Specifically, we discuss and promote the importance of: (a) widening the types of attributes analyzed in the landmark detection process, (b) weighting each attribute relative to its significance, (c) extending the types of objects considered as landmark candidates beyond just buildings, (d) identifying landmarks outside the vicinity of intersections, (e) identifying false landmarks along routes, and (f) using virtual environments for experiments with ALDSs. Throughout the paper, we discuss several demonstrative examples and experiments to clarify and support the ideas and concepts that are being promoted.     

Spatial Vagueness and Second-Order

This paper presents a geometric characterization of geographic objects with vague boundaries. Since the framework is based on ordering geometry and does not rely on numerical concepts, it can be applied to qualitative spatial reasoning. The characterization describes vague regions and gradual transitions between two vague regions. This approach takes into account that the region of all points that definitely belong to an object is also vague, called higher-order vagueness. Higher-order vagueness is usually neglected by theories modeling spatial vagueness. The approach distinguishes firstand second-order vague regions and characterizes the gradual transitions of second-order vague regions. These gradual transitions allow transitions within transitions. In general, it is not possible to reduce higher-order vagueness to first-order vagueness. Using a spatial ordering structure the article shows when second-order vagueness is useful and examines the relationship between first- and second-order vagueness. An example shows how second-order vagueness can be used to represent geographic objects.     

Space patrol: Variants of the optical barrier scheme

Different variants of the space patrol system to be designed for discovering and cataloging space objects hazardous for the Earth have been investigated. The basic idea of this system is to create an optical barrier using the telescopes deployed in a heliocentric orbit. Difficulties (as well as ways of overcoming them) of this program are analyzed, associated with form and position of the orbit of a space object relative to the patrol spacecraft, determination of orbit parameters, and mutual motion of space objects and the telescopes on spacecraft. The barrier’s schemes with scanning vertical or horizontal belts are considered. Some examples of observational conditions are presented for space objects crossing the barrier region: angular positions, velocities, distances, and numbers of days during which they are observed in the barrier region. The barrier’s characteristics are given for telescopes deployed in the orbits of the Earth and Venus.     

Monitoring Object Orientation: Effects of Layout Complexity,Viewpoint Changes,and Object Function

Spatial short-term memory for single target positions is subject to distortions which depend on the spatial layout of visual landmarks. Here, participants had to reproduce the positions of briefly presented targets in the context of three-landmark configurations presented in various orientations. Symmetry properties of distortional patterns were determined by the intrinsic reference system of the landmark configuration as well as by the environment-or body-centered vertical axis. Symmetry was best about the cardinal axes of the landmark system irrespective of their orientation, but symmetry of non-cardinal axes was enhanced when these axes were aligned with the extrinsic vertical. Results are inconsistent with most current models of spatial memory distortions but in line with models explaining distortions in terms of attentional processes in topographical neuronal networks.     

Estimation of the effect of orientation of <Emphasis Type="Italic">International Space Station</Emphasis> on the dose rate in Station’s Service Module when passing through the South-Atlantic anomaly region

The influence of spatial orientation of the International Space Station (ISS) on the dose rate recorded during passages of the station through the South-Atlantic anomaly (SAA) zone is considered. The dose rates detected by dosimeters of the radiation control system of the ISS are compared with results of calculation-based estimates. It is shown that when crossing the SAA region in close trajectories, but with different spatial orientation of the station, the dose rate near cabins of the ISS Service Module can differ by more than a factor of two.     

Ground assisted rendezvous with geosynchronous satellites for the disposal of space debris by means of Earth-oriented tethers

Previous studies have shown that extended length Earth-oriented tethers in the geosynchronous (GEO) region can be used to re-orbit satellites to disposal orbits. One such approach involves the extension of a GEO based tether, collection of a debris object, and retraction of the tether, which transfers the retracted configuration to a higher energy orbit for debris disposal. The re-extension of the tether after debris disposal returns the configuration to the near-GEO altitude. The practical feasibility of such a system depends on the ability to collect GEO debris objects, attach them to a deployed tether system, and retract the tethers for transfer to the disposal orbits.This study addresses the collection and delivery of debris objects to the deployed tether system in GEO. The investigation considers the number, type and the characteristics of the debris objects as well as the collection tug that can be ground controlled to detect, rendezvous and dock with the debris objects for their delivery to the tethers system.A total of more than 400 objects are in drift orbits crossing all longitudes either below or above the geostationary radius. More than 130 objects are also known to librate around the stable points in GEO with periods of libration up to five or more years. A characterization of the position and velocity of the debris objects relative to the collection tug is investigated. Typical rendezvous performance requirements for uncooperative GEO satellites are examined, and the similarities with other approaches such as the ESA's CX-OLEV commercial mission proposal to extend the life of geostationary telecommunication satellites are noted.     

Editorial: Cognitive Semantics and Spatio-Temporal Ontologies

How do people describe the location of a target object to another? This task requires a reference object or frame and terms of reference. Traditional linguistic analyses have loosely organized perspectives around people, objects, or environments as reference objects, using reference terms based on a viewpoint or the intrinsic sides of an object, such as left, right, front, and back or based on the environment, such as north, south, east, and west. In actual communication, social, spatial, and cognitive factors may also affect perspective choice. We examine those factors by varying the spatial information (landmarks and cardinal directions), the communication task (relative cognitive burden to speakers and addressees), and the culture of participants (American and Japanese). Speakers used addressees' perspectives more when addressees had the greater cognitive burden. They also used landmarks and cardinal directions when they were available, especially to avoid difficult discriminations like left/right. Some cases appearing to be perspective taking can be interpreted as using a person as a landmark. Finally, terms like near indicating close proximity were preferred to far and to terms requiring projection of directions. Globally, perspective choices of American and Japanese samples were strikingly similar; that is, Japanese did not select addressees' perspectives more than Americans. The traditional linguistic analyses need to be enhanced to account for effects of cognitive, situational, and social factors.     

Spatial updating and common misinterpretations of spatial reference frames

Spatial memory plays an important role in everyday life, and a large amount of research has been devoted to understanding spatial coding and reference frames across many areas. The popular research paradigms to study spatial reference frames include novel shortcut, perspective change, and landmark control tests. However, the growing research on spatial updating challenges the logical foundation of these classical paradigms, and suggests that the experimental findings using these paradigms have usually been misinterpreted. That is, performance in these tasks is generally unrelated to whether the spatial representations themselves are egocentric or allocentric. This article reviews the traditional paradigms and their logic, summarizes the theories of spatial updating, analyzes the logical flaws in these popular paradigms, and discusses their implications.     

On early cognitive mapping

In this paper we examine the nature of theearly cognitive map – the beginnings of acognitive map formed from one's earlyimpressions of the environment one is in. Twodistinct paradigms have emerged from ourstudies of what information is initially identified in a cognitive map. The first, which weterm a space-based approach, emphasises makingexplicit the spatial extent of the currentlocal environment. The second emphasises makingexplicit the relationships between objects inthe local environment and we call this anobject-based approach. For both paradigms weexamine the psychological literature to findsupport for the approach and the roboticists'attempts at implementing the idea. We arguethat a space-based approach is the moreappropriate way to compute an early cognitivemap. In particular, we find that Siegel andWhite's (1975) object-based hypothesis, whichstates that the developmental progression of acognitive map is from landmark to route tosurvey map, is not supported. The space-basedparadigm underpins our own work in this areaand we outline our own space-based theory forcomputing an early cognitive map.     

The acquisition of survey knowledge for local and global landmark configurations under time pressure

The influence of stress states on cognition is widely recognized. However, the manner in which stress affects survey knowledge acquisition is still unresolved. For the present study, we investigated whether survey knowledge acquisition during a stressful task (i.e., under time pressure) is more accurate for the mental representation of global or local landmarks. Participants navigated through virtual cities with a navigation aid and explicit learning instructions for different landmark configurations. Participants’ judgments of relative direction (JRDs) suggest that global landmark configurations were not represented more accurately than local landmark configurations and that survey knowledge acquisition was not impaired under time pressure. In contrast to prior findings, our results indicate the limitations of the utility of global landmarks for spatial knowledge acquisition.     

Prioritized Spatial Updating in the Intrinsic Frame of Reference

In order to maintain valid situation awareness, people need to update the spatial representations of their surroundings as objects, including themselves, move. The present study investigates the properties of spatial updating in the intrinsic frame of reference, where spatial relations are represented with respect to an external object (other than the viewer self) with an intrinsic reference direction. Three experiments were conducted using a task of direction pointing. It was found that given a relatively stable intrinsic reference direction, responses to a small number of salient objects were faster than responses to non-salient objects (Experiment 1 and Experiment 3). The salience effect disappeared when the intrinsic reference direction was no longer stable (Experiment 2). Furthermore, all three experiments revealed a type of orientation dependence similar to that found in egocentric spatial updating. Our results indicate that spatial updating in the intrinsic reference system can be easy only if a fixed reference direction is maintained and the number of objects that need to be tracked is limited.     

Selecting a reference frame

Reference frames are representations that parse space. In the case of spatial terms, reference frames mediate the mapping of linguistic expressions onto spatial configurations of objects. In the sentence ``The fly is above the cat,' ``above' is defined with respect to a reference frame that is imposed on the cat. Different types of reference frames can be used to define spatial terms, each based on a different source of information. For example, gravity, the orientation of objects in the scene or the orientation of the viewer can all be used to set the orientation of a reference frame. When these reference frames disagree (because the viewer is reclining or because the objects in the scene are overturned), there are competing definitions for the spatial term, resulting in the need for reference frame selection. The purpose of this paper is to review a line of research that examines reference frame selection in the context of spatial language. This work shows that all reference frames are initially active and assign a direction to a spatial term. Moreover, this activation is automatic, and is followed by the selection of a single reference frame, with selection accompanied by inhibition of the non-selected frames. Parallels between reference frame selection in language and in perception and attention are discussed.     

The Effect of an Extra Object on the Linguistic Apprehension of the Spatial Relationship between Two Objects

A sentence describing a spatial relationship includes at least two objects: a located object and a reference object. The presence of an extra object in a scene that is recognized as neither a located nor a reference object, reduces the acceptability of spatial terms related to the spatial relationship between reference and located objects under given conditions (Carlson & Logan, 2001). However, it is still unclear how the overall acceptability distribution of a spatial term changes in a scene when an extra object is present in various locations. In this study, we examined this point, using the paired comparison method with Thurstone's law of judgment (Case V). The results showed that an extra object had both relative increase and reduction effects on the overall acceptability distribution of a spatial term. This suggests that our linguistic apprehension of the spatial relationship is biased intricately by the presence of extra objects.     

A study of people's sketching habits in GIS

Sketching is traditionally associated withdoodling simple strokes on a piece of paper.Only few professionals outside of design andthe fine arts have recognized the expressivepower of this intuitive modality. However,sketching seems particularly well suited tocapture objects and situations in a spatialenvironment, such as geographic space. To learnmore about the techniques and strategies peopleuse when sketching, a survey of sketching wasconducted. The study showed that paper andpencil sketches contain mostly simple andabstract objects that are composed of only fewstrokes. The spatial configuration of a sceneis primarily expressed through the topologicalordering of objects relative to each other.Metric relationships are used to refine spatialconfigurations. These and other findingssuggest that sketching is an appropriatemodality to interact with a computer where onewants to describe and capture objectconfigurations in a spatial environment, suchas a geographic information system (GIS).     

Spatial Reference in Linguistic Human-Robot Interaction: Iterative,Empirically Supported Development of a Model of Projective Relations

The aim of our research is to enable spontaneous and efficient spatial reference to objects in human-robot interaction situations. This paper presents the iterative, empirically based design of a robotic system that uses a computational model for identifying objects on the basis of a range of spatial reference systems. The efficiency of the system is evaluated by two successive empirical studies involving uninformed users. The linguistic analysis points to the striking variability in speakers' spontaneous strategies and preferences, and it motivates a number of modifications of the computational model.     

Multiple Systems for Spatial Imagery: Transformations of Objects and Bodies

Problem-solving often requires imagining spatial changes. Object-based transformations allow imagining an object in a different orientation. Perspective transformations allow imagining changes in one's viewpoint. Three experiments tested the hypothesis that these two transformations are dissociable and specialized for different situations, by manipulating instructions and task parameters and measuring response times, errors, and introspective reports. Human experience with small objects such as telephones and clothes irons consists mostly of manipulation or observed manipulation, which is characterized by object-based transformations. Consistent with this experience, when participants made judgments about small manipulable objects, they showed a strong tendency to use object-based transformations. Experience with human bodies is more varied, including both object-like interactions and interactions in which one must estimate another's perspective. Accordingly, when making judgments about pictures of bodies, participants' selection of a spatial transformation depended on the type of judgment that needed to be made. When instructions violated these natural mappings, performance was impaired. These data argue for the view that multiple spatial transformation systems evolved to solve different spatial reasoning problems.