Salient Landmarks Research Articles

Evidence for discrete landmark use by pigeons during homing

Considerable efforts have been made to investigate how homing pigeons (Columba livia f. domestica) are able to return to their loft from distant, unfamiliar sites while the mechanisms underlying navigation in familiar territory have received less attention. With the recent advent of global positioning system (GPS) data loggers small enough to be carried by pigeons, the role of visual environmental features in guiding navigation over familiar areas is beginning to be understood, yet, surprisingly, we still know very little about whether homing pigeons can rely on discrete, visual landmarks to guide navigation. To assess a possible role of discrete, visual landmarks in navigation, homing pigeons were first trained to home from a site with four wind turbines as salient landmarks as well as from a control site without any distinctive, discrete landmark features. The GPS-recorded flight paths of the pigeons on the last training release were straighter and more similar among birds from the turbine site compared with those from the control site. The pigeons were then released from both sites following a clock-shift manipulation. Vanishing bearings from the turbine site continued to be homeward oriented as 13 of 14 pigeons returned home. By contrast, at the control site the vanishing bearings were deflected in the expected clock-shift direction and only 5 of 13 pigeons returned home. Taken together, our results offer the first strong evidence that discrete, visual landmarks are one source of spatial information homing pigeons can utilize to navigate when flying over a familiar area.

Auto‐segmentation of normal and target structures in head and neck CT images: A feature‐driven model‐based approach

Intensity modulated radiation therapy (IMRT) allows greater control over dose distribution, which leads to a decrease in radiation related toxicity. IMRT, however, requires precise and accurate delineation of the organs at risk and target volumes. Manual delineation is tedious and suffers from both interobserver and intraobserver variability. State of the art auto-segmentation methods are either atlas-based, model-based or hybrid however, robust fully automated segmentation is often difficult due to the insufficient discriminative information provided by standard medical imaging modalities for certain tissue types. In this paper, the authors present a fully automated hybrid approach which combines deformable registration with the model-based approach to accurately segment normal and target tissues from head and neck CT images. The segmentation process starts by using an average atlas to reliably identify salient landmarks in the patient image. The relationship between these landmarks and the reference dataset serves to guide a deformable registration algorithm, which allows for a close initialization of a set of organ-specific deformable models in the patient image, ensuring their robust adaptation to the boundaries of the structures. Finally, the models are automatically fine adjusted by our boundary refinement approach which attempts to model the uncertainty in model adaptation using a probabilistic mask. This uncertainty is subsequently resolved by voxel classification based on local low-level organ-specific features. To quantitatively evaluate the method, they auto-segment several organs at risk and target tissues from 10 head and neck CT images. They compare the segmentations to the manual delineations outlined by the expert. The evaluation is carried out by estimating two common quantitative measures on 10 datasets: volume overlap fraction or the Dice similarity coefficient (DSC), and a geometrical metric, the median symmetric Hausdorff distance (HD), which is evaluated slice-wise. They achieve an average overlap of 93% for the mandible, 91% for the brainstem, 83% for the parotids, 83% for the submandibular glands, and 74% for the lymph node levels. Our automated segmentation framework is able to segment anatomy in the head and neck region with high accuracy within a clinically-acceptable segmentation time.

Giving the ‘Right’ Route Directions: The Requirements for Pedestrian Navigation Systems

AbstractIt is more natural and effective to include a variety of landmarks when providing route directions. Many of the existing solutions, however, behave like car navigation systems and do not include references to such landmarks. This research provides empirical evidence to identify the information requirements for an automated pedestrian navigation system. The two experiments discussed within this article helped identify the criteria that govern the relative saliency of features of interest within an urban environment. There are a large variety of features of interest that can be used as directional aids within route descriptions (buildings, statues, monuments, hills, roads, etc). Different measures are therefore required in order to calculate the saliency of the different classes of features. The research revealed that the most important measures of saliency include name, size, age, and colour. The research confirmed that that relative visibility and the direction of approach to the feature also affect a landmark's saliency. The research also highlighted the role that features of interest play in reorientation, confirmatory, and direction‐giving tasks.

The Giessen virtual environment laboratory: human wayfinding and landmark salience

In our virtual environment laboratory, we focus on different topics in human spatial cognition with projects on landmark salience, route knowledge, and survey knowledge. Within this laboratory note, we provide an overview of previous, current, and future work with our virtual environment SQUARELAND.

Subjective and objective dimensions of spatial legibility

Reading space means understanding, analyzing or evaluating a certain space. Yet spatial legibility differs from the concept of reading. Legibility is one of the principles of urban design. Legibility means the possibility of organizing an environment within an imageable and coherent pattern. Reading an environment is a process that evolves with the obtaining of spatial information from the environment and by mentally processing that information and using it in a way appropriate to its purpose. Two components play a part in the process of obtaining spatial information: the characteristics of the space and the characteristics of the observer. The observer's perception and understanding of the characteristics of a space occurs as a result of spatialpsychological processes that happen in the mind. At the same time, legibility is influenced by spatial characteristics. The degree of legibility of a space depends on the plan layout in the second dimension and its complexity, and on the aliency of architectural components in the third dimension. There are many concepts in the literature that define legible environments: simple, coherent, understandable, perceivable, etc. All of these concepts point to characteristics deriving from the structure of the space. However, it is impossible to measure legibility by these concepts. It is discussed in this article that there are two main variables to devise a definition based on characteristics deriving from space: 1. the complexity of spatial layout and 2. the saliency of landmarks. The complexity of spatial layout describes the two-dimensional information about a space, while the saliency of landmarks refers to the three-dimensional information about a space. These two variables are also the elements of spatial information used while employing wayfinding behavior.

Towards Attentive Robots

AbstractThis paper introduces

Developmental changes in infants’ visual short-term memory for location

To examine the development of visual short-term memory (VSTM) for location, we presented 6- to 12-month-old infants ( N = 199) with two side-by-side stimulus streams. In each stream, arrays of colored circles continually appeared, disappeared, and reappeared. In the changing stream, the location of one or more items changed in each cycle; in the non-changing streams the locations did not change. Eight- and 12.5-month-old infants showed evidence of memory for multiple locations, whereas 6.5-month-old infants showed evidence of memory only for a single location, and only when that location was easily identified by salient landmarks. In the absence of such landmarks, 6.5-month-old infants showed evidence of memory for the overall configuration or shape. This developmental trajectory for spatial VSTM is similar to that previously observed for color VSTM. These results additionally show that infants’ ability to detect changes in location is dependent on their developing sensitivity to spatial reference frames.

Selecting landmarks in novel environments

People use salient landmarks when learning a route through a novel environment. However, it is not clear what makes a given landmark salient. In two experiments, subjects learned a route through a virtual museum, performed a recognition memory test for objects in the museum, and provided spatial descriptions and drew maps of the learned route. Objects with strong perceptual features occurred at decision points or at non-decision points along the route. Objects with both of these features were recognized faster and were included more often in the maps and written directions. When these features were separated, perceptual features maintained a strong influence on the recognition task, but had no influence on the spatial tasks, which were influenced only by spatial features. These findings challenge the idea that either a recognition task or descriptive task alone provides a complete account of landmark representation.

Lateral entorhinal neurons are not spatially selective in cue-rich environments.

The hippocampus is a brain region that is critical for spatial learning, context-dependent memory, and episodic memory. It receives major inputs from the medial entorhinal cortex (MEC) and the lateral EC (LEC). MEC neurons show much greater spatial firing than LEC neurons in a recording chamber with a single, salient landmark. The MEC cells are thought to derive their spatial tuning through path integration, which permits spatially selective firing in such a cue-deprived environment. In accordance with theories that postulate two spatial mapping systems that provide input to the hippocampus-an internal, path-integration system and an external, landmark-based system-it was possible that LEC neurons can also convey a spatial signal, but that the signal requires multiple landmarks to define locations, rather than movement integration. To test this hypothesis, neurons from the MEC and LEC were recorded as rats foraged for food in cue-rich environments. In both environments, LEC neurons showed little spatial specificity, whereas many MEC neurons showed a robust spatial signal. These data strongly support the notion that the MEC and LEC convey fundamentally different types of information to the hippocampus, in terms of their spatial firing characteristics, under various environmental and behavioral conditions.

Spaces or Scenes: Map-based Orientation in Urban Environments

Two experiments examined people's strategies when orienting with a map in outdoor scenes within unfamiliar urban environments. We investigated how the 3D visual scene and the 2D layout geometry influenced people's choices of features when matching the scene and the map, and studied the problems they encountered when doing so. Results support previous evidence that in geographically realistic contexts, visible salient landmarks bias people away from using optimal geometry-matching strategies. This implies that prediction of orientation difficulty merely from analysing the spatial layout (e.g., with space syntax isovist measures) may be highly problematic. Implications for future map design are discussed.

Perception of Space by Multiple Intrinsic Frames of Reference

It has been documented that when memorizing a physical space, the person's mental representation of that space is biased with distortion and segmentation. Two experiments reported here suggest that distortion and segmentation arise due to a hierarchical organization of the spatial representation. The spatial relations associated with salient landmarks are more strongly encoded and easier to recall than those associated with non-salient landmarks. In the presence of multiple salient landmarks, multiple intrinsic frames of reference are formed and spatial relations are anchored to each individual frame of reference. Multiple such representations may co-exist and interactively determine a person's spatial performance.

Including landmarks in routing instructions

This article addresses the problem of incorporating cognitively salient landmarks in computer-generated navigation instructions. On the basis of a review of the existing literature in the domain of navigation with landmarks, the article develops algorithms for generating routing instructions that include references to landmarks. The most basic algorithm uses a new weighting model to annotate simple routes with references to landmarks. A key novel feature of this algorithm is that it depends only on commonly available data and generic capabilities of existing web mapping environments. A suite of extensions are also proposed for improving the cognitive ergonomics of the basic landmark instructions. A case study, implemented within a national online routing system, demonstrates practicality of the approach. The article then concludes by reviewing a range of further issues for future work.

When directions fail: Investigation of getting lost behaviour in adults with acquired brain injury

Primary objective: To investigate and describe getting lost behaviour and wayfinding strategies among acquired brain injury (ABI) survivors and matched controls.Research design: Matched control group comparison design.Methods and procedures: This study compared wayfinding performance of 18 adults with acquired brain injury to controls matched for gender, age and education. Participants followed written directions along an eight-step route in an unfamiliar neighbourhood, with three intentionally challenging choice-points. They used a cellular phone to request assistance if they became lost. Dependent measures included accuracy, directness and wayfinding strategy. Statistical and qualitative analyses explored group themes and differences.Main results: Participants with ABI demonstrated significantly greater on-route wayfinding errors and hesitancy than matched controls. The ABI group requested assistance over the cell phone more frequently than controls and required more attempts at re-orientation with concrete, salient directions in order to re-orient in the field. Participants in the control group anticipated errors with greater frequency than those with ABI.Conclusions: ABI survivors demonstrated greater challenges with wayfinding than matched controls. Re-orientation required concrete, explicit redirection with reference to salient landmarks. Implications for clinical practice and assistive technology are discussed.

Biologically Inspired Mobile Robot Vision Localization

We present a robot localization system using biologically inspired vision. Our system models two extensively studied human visual capabilities: (1) extracting the ldquogistrdquo of a scene to produce a coarse localization hypothesis and (2) refining it by locating salient landmark points in the scene. Gist is computed here as a holistic statistical signature of the image, thereby yielding abstract scene classification and layout. Saliency is computed as a measure of interest at every image location, which efficiently directs the time-consuming landmark-identification process toward the most likely candidate locations in the image. The gist features and salient regions are then further processed using a Monte Carlo localization algorithm to allow the robot to generate its position. We test the system in three different outdoor environments-building complex (38.4 m times 54.86 m area, 13 966 testing images), vegetation-filled park (82.3 m times 109.73 m area, 26 397 testing images), and open-field park (137.16 m times 178.31 m area, 34 711 testing images)-each with its own challenges. The system is able to localize, on average, within 0.98, 2.63, and 3.46 m, respectively, even with multiple kidnapped-robot instances.

Which portion of the natural panorama is used for view-based navigation in the Australian desert ant?

Ants that forage in visually rich environments often develop idiosyncratic routes between their nest and a profitable foraging ground. Such route knowledge is underpinned by an ability to use visual landmarks for guidance and place recognition. Here we ask which portions of natural visual scenes are essential for visually guided navigation in the Australian desert ant Melophorus bagoti whose foragers navigate through a habitat containing grass tussocks, shrubs and trees. We captured M. bagoti foragers after they had returned to their nest from a feeder, but before they had entered their nest, and tested their ability to home accurately from a series of release locations. We used this simple release paradigm to investigate visually guided navigation by monitoring the accuracy of nestwards orientation when parts of the ants' visual field were obscured. Results show that the lower portion of the visual panorama is more important for visually guided homing than upper portions. Analysis of panoramic images captured from the release and nest locations support the hypothesis that the important visual information is provided by the panoramic contour, where terrestrial objects contrast against sky, rather than by a limited number of salient landmarks such as tall trees.

Comparing black-capped (Poecile atricapillus) and mountain chickadees (Poecile gambeli): use of geometric and featural information in a spatial orientation task

Since Cheng (Cognition 23:149-178, 1986) first proposed the "geometric module" in rats, a great deal of research has focused on how other species use geometric information and how geometric encoding may differ across species. Here, hand-reared and wild-caught black-capped chickadees and wild-caught mountain chickadees searched for food hidden in one corner in a rectangular environment. Previous research has shown that mountain chickadees do not spontaneously encode geometric information when a salient feature is present near the goal location. Using a slightly different training and testing procedure, we found that both hand-reared and wild-caught black-capped chickadees encoded geometric information, even in the presence of a salient landmark. Some, but not all, mountain chickadees also encoded geometric information. Overall, our results suggest that use of geometric information may be a less preferred strategy for mountain chickadees than for either wild-caught or hand-reared black-capped chickadees. To our knowledge, this is the first direct interspecies comparison of use of geometric information in a spatial orientation task.

Calibrating color vision

Why do some colors look special? Color vision is our ability to distinguish the spectral characteristics of light. Wavelength varies continuously, yet we perceive spectra in terms of a small number of dimensions anchored by salient landmarks. These dimensions define a color space, within which colors appear to vary in hue (direction) or saturation (distance) relative to a neutral gray (Figure 1). Coding relative to a norm may reflect a general strategy in perception. For example, individual faces may be judged by how they differ from a neutral or average face.

Influences on the First-Perspective Alignment Effect from Text Route Descriptions

Four experiments investigated the more efficient recall of routes learned from text descriptions when the imagined orientation at test was in alignment with the first experienced perspective. Experiments 1 and 2 replicated the effect, but found little evidence for the influence of an external frame of reference provided either by describing a salient landmark external to the route, or by employing cardinal directions in the descriptions. In Experiment 3, the first-perspective alignment (FPA) effect was relatively unaffected by elaboration of spatial information or more experience of reading the text. Experiment 4 found attenuation of the FPA effect when participants made active spatial judgements from imagined key locations while learning. The results are discussed in relation to theories of spatial reference frames and the influence of location salience.

Honeybees can recognise images of complex natural scenes for use as potential landmarks

The ability to navigate long distances to find rewarding flowers and return home is a key factor in the survival of honeybees (Apis mellifera). To reliably perform this task, bees combine both odometric and landmark cues, which potentially creates a dilemma since environments rich in odometric cues might be poor in salient landmark cues, and vice versa. In the present study, honeybees were provided with differential conditioning to images of complex natural scenes, in order to determine if they could reliably learn to discriminate between very similar scenes, and to recognise a learnt scene from a novel distractor scene. Choices made by individual bees were modelled with signal detection theory, and bees demonstrated an ability to discriminate between perceptually similar target and distractor views despite similar spatiotemporal content of the images. In a non-rewarded transfer test bees were also able to recognise target stimuli from novel distractors. These findings indicate that visual processing in bees is sufficiently accurate for recognising views of complex scenery as potential landmarks, which would enable bees flying in a forest to use trees both as landmark and/or odometric cues.

On the assessment of landmark salience for human navigation

In this paper, we propose a conceptual framework for assessing the salience of landmarks for navigation. Landmark salience is derived as a result of the observer's point of view, both physical and cognitive, the surrounding environment, and the objects contained therein. This is in contrast to the currently held view that salience is an inherent property of some spatial feature. Salience, in our approach, is expressed as a three-valued Saliency Vector. The components that determine this vector are Perceptual Salience, which defines the exogenous (or passive) potential of an object or region for acquisition of visual attention, Cognitive Salience, which is an endogenous (or active) mode of orienting attention, triggered by informative cues providing advance information about the target location, and Contextual Salience, which is tightly coupled to modality and task to be performed. This separation between voluntary and involuntary direction of visual attention in dependence of the context allows defining a framework that accounts for the interaction between observer, environment, and landmark. We identify the low-level factors that contribute to each type of salience and suggest a probabilistic approach for their integration. Finally, we discuss the implications, consider restrictions, and explore the scope of the framework.