Perspective Scenes Research Articles

A novel framework for crowd counting using video and audio

Crowd counting is significant in many areas. The existing methods have poor accuracy for perspective scenes and low illumination scenes. Besides, the existing audio-assisted methods only use local audio, which fails to provide the spatial feature information of sound in all directions in space. To alleviate the above problems, a novel framework named Video and Audio-assisted Crowd Counting Network (VACCNet) is proposed. The framework consists of two submodules: Video Crowd Counting (VCC) module and Audio-assisted Crowd Counting (ACC) module. The visual features from the VCC module and the fused audio features from the ACC module are further combined to get the final density map. To prove the effects of VACCNet, a new self-collected dataset named multiPle dIrection Assistance couNting netwOrk (PIANO) is built. The experimental results based on existing benchmarks and PIANO show that the proposed method has a 14.23% improvement averagely to the conventional methods.

One-shot classification-based tilled soil region segmentation for boundary guidance in autonomous tillage

In recent years, machine vision has been frequently used to estimate local paths for autonomous tractors. The incorporation of deep learning-based techniques to its methodology is expected to induce robust performance under practical field conditions. Typical tillage of autonomous tractors requires the tilled soil region detection to provide the local paths, which are usually determined based on the boundaries between tilled and non-tilled soil. However, tilled soil regions usually exhibit different feature distributions by field (domain) in terms of factors like soil characteristics, crop type, and method of tillage. This makes the development of deep learning-based systems with highly generalizable performance difficult. In this study, this issue is addressed by proposing a few-shot learning (FSL)-based effective tilled soil region segmentation, which is utilized to detect tilled soil region in 2D perspective scenes to provide tillage path guidance to autonomous tractors. The proposed approach used an image patch classification-based field region segmentation pipeline as the backbone and a one-shot classification model was designed to classify each image patch as tilled or non-tilled soil. The tilled soil region is then segmented by clustering the image patches belonging to the corresponding classes. The results demonstrate that one-shot classification exhibits a classification accuracy exceeding 0.9 over the entire dataset obtained from various fields, including new data, while the typical classification method exhibits a classification accuracy of approximately 0.74 on new data. Further, segmentation is successfully performed on the detected tilled soil regions belonging to various domains. In conclusion, the proposed method can provide the relative feature difference between tilled and non-tilled soils, allowing for expansion to new data without fine-tuning. This robust field region segmentation can be utilized practically as an essential skill to support the tillage path decision for an autonomous tractor.

Unimpaired perception of relative depth from perspective cues in strabismus.

Strabismus is a relatively common ophthalmological condition where the coordination of eye muscles to binocularly fixate a single point in space is impaired. This leads to deficits in vision and particularly in three-dimensional (3D) space perception. The exact nature of the deficits in 3D perception is poorly understood as much of understanding has relied on anecdotal reports or conjecture. Here, we investigated, for the first time, the perception of relative depth comparing strabismic and typically developed binocular observers. Specifically, we assessed the susceptibility to the depth cue of perspective convergence as well as the capacity to use this cue to make accurate judgements of relative depth. Susceptibility was measured by examining a 3D bias in making two-dimensional (2D) interval equidistance judgements and accuracy was measured by examining 3D interval equidistance judgements. We tested both monocular and binocular viewing of images of perspective scenes under two different psychophysical methods: two-alternative forced-choice (2AFC) and the method of adjustment. The biasing effect of perspective information on the 2D judgements (3D cue susceptibility) was highly significant and comparable for both subject groups in both the psychophysical tasks (all ps < 0.001) with no statistically significant difference found between the two groups. Both groups showed an underestimation in the 3D task with no significant difference between the group's judgements in the 2AFC task, but a small statistically significant difference (ratio difference of approx. 10%, p = 0.016) in the method of adjustment task. A small but significant effect of viewing condition (monocular versus binocular) was revealed only in the non-strabismic group (ratio difference of approx. 6%, p = 0.002). Our results show that both the automatic susceptibility to, and accuracy in the use of, the perspective convergence cue in strabismus is largely comparable to that found in typically developed binocular vision, and have implications on the nature of the encoding of depth in the human visual system.

An Interactive Perspective Scene Completion Framework Guided by Complanate Mesh

This paper presents an efficient interactive framework for perspective scene completion and editing tasks, which are available largely in the real world but rarely studied in the field of image completion. Considering that it is quite hard to extract perspective information from a single image, this work starts from a friendly and portable interactive platform to obtain the basic perspective data. Then, in order to make this interface less sensitive, easier and more flexible, a perspective-rectification based correction mechanism is proposed to iteratively update the locations of the initial points selected by users. At last, a complanate mesh is generated by the geometry calculations from these corrected initial positions. This mesh must approximate the perspective direction and the structure topology as much as possible so that the filling process can be conducted under the constraint of the perspective effects of the original image. Our experiments show the results with good qualities and performances, and also demonstrate the validity of our approaches by various perspective scenes and images.

Abstracts from the 7th Visual Science of Art Conference (VSAC) Leuven, Belgium, August 21st–24th, 2019

There are two types of perspective relevant to human vision, photographic and visual perspective. Photographic perspective results from projection of the three-dimensional world on two-dimensional surfaces such as canvases and retinae. Photographic perspective is widely applied in drawings and paintings to create an illusion of depth. Visual perspective is a property of the space that we, human beings, perceive through vision. Photographic perspective is two-dimensional, visual perspec- tive three-dimensional. Paintings of perspective scenes are interesting objects of study because both types of perspectives play a role in the creation of depth. A previous study showed that Canaletto and other vedutisti applied perspective related to visual space.Paintings of Piazza San Marco in Venice are analysed in the current study. These paintings are of particular interest because of the Piazza’s extraordinary geometry and the fact that Canaletto and his followers produced dozens of paintings from almost identical viewing positions. Buildings and patterns are directed to three horizontally separated vanishing points. Perspective is extremely veridical in a number of paintings. It is hard to imagine how such accuracy could have been obtained without the help of a camera. In other paintings the quality of perspective is poorer and perspective is completely wrong in number of paintings, suggesting that those paintings are not of Canaletto. People have been painted twice to four times too large on all the paintings. The large sizes suggest that people were added later, aiming to portray San Marco as a more intimate plaza than it is.

Multiple Photographs of a Perspective Scene Reveal the Principles of Picture Perception.

A picture is a powerful and convenient medium for inducing the illusion that one perceives a three-dimensional scene. The relative invariance of picture perception across viewing positions has aroused the interest of painters, photographers, and visual scientists. This study explores variables that may underlie the invariance. It presents a computational analysis of distances and directions in sets of two photographs of perspective scenes taken from different camera positions. Focal lengths of the lens and picture sizes are chosen such that the sizes of one of the familiar objects are equally large in both photographs. The selected object is perceived at the same distance in both photographs, independent of viewing distance, showing that pictorial distance is fully determined by angular size of the object. Pictorial distance is independent of camera position, focal length of the lens, and picture size. Distances and directions of pictorial objects are computed as a function of viewing distance, and compared with distances and directions of the physical objects as a function of camera position. The computations show that ratios between pictorial distances, directions, and angular sizes of objects in a photograph are constant, as a function of viewing distance. The constant ratios are proposed as the reason for invariance of picture perception over a range of viewing distances. Reanalysis of distance judgments obtained from the literature shows that perspective space, previously proposed as the model for visual space, is also a good model for pictorial space. The geometry of pictorial space contradicts some conceptions about picture perception.

Equidistant Intervals in Perspective Photographs and Paintings.

Human vision is extremely sensitive to equidistance of spatial intervals in the frontal plane. Thresholds for spatial equidistance have been extensively measured in bisecting tasks. Despite the vast number of studies, the informational basis for equidistance perception is unknown. There are three possible sources of information for spatial equidistance in pictures, namely, distances in the picture plane, in physical space, and visual space. For each source, equidistant intervals were computed for perspective photographs of walls and canals. Intervals appear equidistant if equidistance is defined in visual space. Equidistance was further investigated in paintings of perspective scenes. In appraisals of the perspective skill of painters, emphasis has been on accurate use of vanishing points. The current study investigated the skill of painters to depict equidistant intervals. Depicted rows of equidistant columns, tiles, tapestries, or trees were analyzed in 30 paintings and engravings. Computational analysis shows that from the middle ages until now, artists either represented equidistance in physical space or in a visual space of very limited depth. Among the painters and engravers who depict equidistance in a highly nonveridical visual space are renowned experts of linear perspective.

A Method for Generating Customizable Coordinate Axes in AxSceneControl

Coordinate axes have been used extensively in three-dimensional (3D) GIS processing to help visualize the location of points directly. Meanwhile, AxSceneControl is a visual control provided by ESRI®ArcGIS®Engine, which is suited to generating perspective scenes that allow user to navigate and interact with 3D feature and raster data. However, there is no direct solution given by ArcGIS Engine to set Coordinate axes in 3D scenes. In this paper, we proposed and implemented an approach to generate and customize coordinate axis in AxSceneControl. By this method, up to five coordinate axes, named as x axis, y axis, z axis, top axis and right axis, can be drawn based on the layer that is loaded to the AxSceneControl. Furthermore, users can modify the properties of coordinate axes such as changing the style of degree scale, determining which axis will be shown and resizing the degree density to achieve simple coordinate axis.

The Theatres of Inigo Jones and John Webb

List of plates Preface A note on terminology 1. Introduction 2. The theatre at Christ Church, Oxford 3. The Cockpit in Drury Lane 4. Perspective scenes at Somerset House 5. The Cockpit-in-Court 6. The Paved Court Theatre at Somerset House 7. The Florimene Theatre at Whitehall 8. The Masquing House at Whitehall 9. An unidentified theatre project 10. The Hall Theatre at Whitehall 11. Conclusion Notes Index.

Practical Automation In Commercial Digital Photogrammetry

The revolution promised by digital photogrammetry is dependent on extensive, successful automation at many stages of the production process. The progress of automation is uneven. Digital cameras are not in widespread use. Scanners are available which include roll film transports and automated orientation, but set up for optimum tonal transfer remains interactive. Extensive automation of several operations on digital photogrammetric workstations, such as generation of digital terrain models, orthophotographs, mosaics and perspective scenes, contrasts with feature extraction, where robust products lag behind success in research. Progress towards practical automation to assist, but not replace, the human operator is significant, however, and examples are given from the Leica‐Helava systems.

Visual Enhancements and Geometric Field of View as Factors in the Design of a Three-Dimensional Perspective Display

This study investigated the effect of providing visual enhancements to a three-dimensional (3D) perspective display on the observers ability to judge the azimuth and elevation which separated two computer-generated images. The 3D perspective scenes were modeled after displays presented previously by McGreevy and Ellis (1986) but with several visual enhancements designed to assist users in performance of the experiment tasks. The visual enhancements included: (1) the capability to rotate the perspective scenes in near real-time and, (2) the presentation of solid shaded objects in the computer-generated scenes. The results provide information on the magnitude of the errors which occur when observers are required to make directional judgements using perspective displays and on the effectiveness of several visual enhancements on the accuracy of directional judgements using a 3D perspective display.

The horizontal-vertical illusion in photographs of concrete scenes with and without depth information.

On the basis of the hypothesis of misapplied constancy scaling, the perception of an abstract horizontal-vertical illusion figure embedded in photographs of natural scenes with depth cues is investigated. The effect is compared with that of a figure on photographs containing no depth information and with a figure on a neutral surface. It is shown that the magnitude of the illusion in the perspective scenes is greater than in the other two conditions. The results are considered compatible with a constancy theory of the illusion. Finally, the evidence for misapplied constancy scaling in the horizontal-vertical illusion in relation to a retinal theory is discussed.

PSYCHOMETRICAL STUDIES ON THE ROLE OF ROADSIDE BUILDINGS IN STREETSCAPE “UNITY”

Effects of roadside buildings height (H), frontage width (BW) and setback (SB) on streetscape unity were investigated. The authors drew 207 streetscape perspective scenes by XY-plotter assigning mean and standard deviation of three building attributes to each scene. These scenes were presented to respondents to rate the sense of oppression, enclosure and unity (disorder-uniformity). Curve fitting and multiple regression analysis techniques were applied to the obtained data. The results will help city planners and town designers predict and estimate the results of their plans.

An Aid to Hidden Surface Removal in Real Time CGI Systems

One of the limiting factors in the development of real-time Computer Generated Image systems for displaying perspective colour scenes is finding a suitable solution to the hidden surface problem. This paper presents a solution to this problem, whereby spatial properties of a three-dimensional 'model' held in the Computer Generated Image system database are utilized in order to minimize hidden surface computations for producing moving perspective scenes. This solution is particularly acceptable to situations involving a fixed model and moving eye point. Often Computer Generated Image systems resort to processing individual surfaces from objects in the model to obtain display priority levels; the solution presented is based upon the use of preprocessed relative priority levels which only require complete objects to be given priority levels thus reducing real-time computation. Once an absolute display priority level for an object is found it will be shown that all surfaces forming that object have known display priority levels. Techniques used to obtain object display priority levels are discussed, together with background information concerning the hidden surface problem.

A linear time exact hidden surface algorithm

This paper presents a new hidden surface algorithm. Its output is the set of the visible pieces of edges and faces, and is as accurate as the arithmetic precision of the computer. Thus calculating the hidden surfaces for a higher resolution device takes no more time. If the faces are independently and identically distributed, then the execution time is linear in the number of faces. In particular, the execution time does not increase with the depth complexity. This algorithm overlays a grid on the screen whose fineness depends on the number and size of the faces. Edges and faces are sorted into grid cells. Only objects in the same cell can intersect or hide each other. Also, if a face completely covers a cell then nothing behind it in the cell is relevant. Three programs have tested this algorithm. The first verified the variable grid concept on 50,000 intersecting edges. The second verified the linear time, fast speed, and irrelevance of depth complexity for hidden lines on 10,000 spheres. This also tested depth complexities up to 30, and showed that perspective scenes with the farther objects smaller are even faster to calculate. The third verified this for hidden surfaces on 3000 squares.

The perception of egomotion.

Egomotion is defined as any environmental displacement of the observer. Twenty stationary observers viewed computer-generated films that simulated rectilinear egomotion of constant speed and altitude over an endless plain. The heading point could be 0 degrees, 15 degrees, 30 degrees, 45 degrees, 60 degrees, 75 degrees, or 90 degrees to an observer's right of the screen center. The screen was 53 degrees wide. Hence, there was a constant heading toward a point on the horizon, but the heading point was not always in view. The typical observer's mean pointing bias was 5.56 degrees to his right and his SD pointing error was 5.18 degrees. The results indicate that these abstract view samples of a pure egomotion ambient optic array are equally effective in giving rise to a compelling perception of egomotion. The results also indicate that observers can perceive the direction of their heading whether or not the heading point is in the view, although accuracy did vary slightly with the specific view. A pointing control using linear perspective scenes whose vanishing points fell off-view showed the same pattern of pointing errors with an overall rightward pointing bias of 3.19 degrees and with an SD pointing error of 3.04 degrees.