Region Of Scene Research Articles

Visible and Infrared Object Tracking Based on Multimodal Hierarchical Relationship Modeling

Visible RGB and Thermal infrared (RGBT) object tracking has emerged as a prominent area of focus within the realm of computer vision. Nevertheless, the majority of existing RGBT tracking methods, which predominantly rely on Transformers, primarily emphasize the enhancement of features extracted by convolutional neural networks. Unfortunately, the latent potential of Transformers in representation learning has been inadequately explored. Furthermore, most studies tend to overlook the significance of distinguishing between the importance of each modality in the context of multimodal tasks. In this paper, we address these two critical issues by introducing a novel RGBT tracking framework centered on multimodal hierarchical relationship modeling. Through the incorporation of multiple Transformer encoders and the deployment of self-attention mechanisms, we progressively aggregate and fuse multimodal image features at various stages of image feature learning. Throughout the process of multimodal interaction within the network, we employ a dynamic component feature fusion module at the patch-level to dynamically assess the relevance of visible information within each region of the tracking scene. Our extensive experimentation, conducted on benchmark datasets such as RGBT234, GTOT, and LasHeR, substantiates the commendable performance of our proposed approach in terms of accuracy, success rate, and tracking speed.

Real-time 3D shape measurement with dual-frequency composite grating and motion-induced error reduction

Phase-shifting profilometry has been increasingly sought and applied in dynamic three-dimensional (3D) shape measurement. However, the object motion will lead to extra phase shift error and thus measurement error. In this paper, a real-time 3D shape measurement method based on dual-frequency composite phase-shifting grating and motion-induced error reduction is proposed for a complex scene containing dynamic and static objects. The proposed method detects the motion region of a complex scene through the phase relations of the dual-frequency composite grating and reduces the motion-induced error with the combination of the phase calculated by a phase-shifting algorithm and the phase extracted by Fourier fringe analysis. It can correctly reconstruct the 3D shape of a complex dynamic scene and ensure high measurement accuracy of its static object as well. With the aid of the phase-shifting image ordering approach, the dynamic 3D shape of complex scenes can be reconstructed and the motion-induced error can also be suppressed in real time. Experimental results well proved that the proposed method is effective and practical.

Superpixel-based online wagging one-class ensemble for feature selection in foreground/background separation

In the last decades, researchers in the field of Background Subtraction (BS) have developed methods to handle the different type of challenges. However, at the present time, no traditional algorithm seems to be able to simultaneously address all the key BS challenges. This can mainly be attributed to the lack of systematic investigation concerning the role and the importance of features within background modeling and foreground detection. In this paper, we present a novel online one-class ensemble based on wagging to select suitable features to each region of a certain scene to distinguish the foreground objects from the background. In addition, we propose a mechanism to update the importance of each feature discarding insignificantly features over time. The experimental results on three challenging datasets (i.e MSVS, RGB-D object detection, CD.net 2014) show the pertinence of the proposed approach.

On the Performance and Implementation of Parallax Free Video See-Through Displays

In see-through systems an observer watches a (background) scene partially occluded by a display. In this display, usually positioned close to the observer, a region of the background scene is shown, yielding the sensation that the display is transparent. To achieve the transparency effect, it is very important to compensate the parallax error and other distortions caused by the image acquisition system. In this paper a detailed study of a video see-through methodology with parallax correction is performed. In a system composed by two cameras-one directed to the user and another to the background scene-and a display, the relative position between the user, the display and the scene is estimated using a feature detection algorithm and the parallax error is compensated assuming a planar scene model. The application of the proposed methodology on Driver Assistance Systems (DAS) is proposed. A theoretical assessment of the algorithm shows that although approximations are proposed to simplify the methodology and reduce the computational cost, such as the planar scene model and fixed working distance, on some practical situations their effects can be neglected without noticeable impact on the perceptual quality of the solution.

HPV guided object tracking: Theoretical advances on fast pattern matching technique

Summary Pattern matching is a fundamental machine vision problem that deals with searching an object in a comparatively large scene. It can use to solve many vision problems ranging from typical human detection to searching defective parts in industrial automation. This paper reports a fast pattern matching technique which makes use of cumulative subtraction and cumulative division operations based on Image Integral model. The idea is to use both the cumulative subtraction and division operations to evaluate the image values on a very small rectangular region of the image scene as well as on the input pattern to be searched for. Image values are transformed to Haar Projection Values (HPVs) using Haar transform in order to achieve pattern matching on sliding window of the image scene. Computation of HPV needs seven arithmetic operations, including two addition and five subtraction operations, which are found to be same as that of Image Integral technique. Besides, the proposed pattern matching technique is identified as computationally effective in terms of both time and memory.

Improved synthetic aperture radar micro‐Doppler jamming method based on phase‐switched screen

Target micro-motions such as rotation and vibration introduce phase modulation, termed as micro-Doppler (m-D) effect, onto synthetic aperture radar (SAR) signals. This causes ghost targets in the reconstructed SAR images. Inspired by this unique characteristic, a passive-jamming method based on m-D was developed for SAR. The m-D jamming method utilised a rotating reflector to intentionally generate ghost targets as jamming strips on the SAR image so as to protect certain areas. The m-D jamming method based on a single rotating reflector can only generate a long jamming strip along the azimuth direction but located in a finite number of range cells, which restricts the region of protected scene. In this study, an improved m-D jamming method based on phase-switched screen (PSS) is proposed, which combines the advantage induced by the PSS modulation and the usefulness of the m-D jamming. By controlling the modulating frequency and waveform of the PSS, the jamming strip is enlarged along the range direction. Theoretical analyses and simulation results verify the validity of the proposed method.

A Class of Visual Neurons with Wide-Field Properties Is Required for Local Motion Detection

Visual motion cues are used by many animals to guide navigation across a wide range of environments. Long-standing theoretical models have made predictions about the computations that compare light signals across space and time to detect motion. Using connectomic and physiological approaches, candidate circuits that can implement various algorithmic steps have been proposed in the Drosophila visual system. These pathways connect photoreceptors, via interneurons in the lamina and the medulla, to direction-selective cells in the lobula and lobula plate. However, the functional architecture of these circuits remains incompletely understood. Here, we use a forward genetic approach to identify the medulla neuron Tm9 as critical for motion-evoked behavioral responses. Using in vivo calcium imaging combined with genetic silencing, we place Tm9 within motion-detecting circuitry. Tm9 receives functional inputs from the lamina neurons L3 and, unexpectedly, L1 and passes information onto the direction-selective T5 neuron. Whereas the morphology of Tm9 suggested that this cell would inform circuits about local points in space, we found that the Tm9 spatial receptive field is large. Thus, this circuit informs elementary motion detectors about a wide region of the visual scene. In addition, Tm9 exhibits sustained responses that provide a tonic signal about incoming light patterns. Silencing Tm9 dramatically reduces the response amplitude of T5 neurons under a broad range of different motion conditions. Thus, our data demonstrate that sustained and wide-field signals are essential for elementary motion processing.

An Exploration with Novel Shape Signature of GMSC Distance Function to Track the Object

The process explicitly dedicated to estimate the path of the object as it moves along the region of scene in the image plane is the principle of tracking. In other words, it is a strategy to detect and track moving object through a sequence of frames. In this work, optical flow based on Horn–Schunk with Barren, Fleet and Beuchemin (BFB) kernel has been employed to estimate the motion vectors. The peripheries of moving objects are extracted for different shape signatures such as boundary, edge, area, curvature and centroid distance functions. Fourier descriptors (FD) of particular shape signature for each of the candidate templates and model template are computed. Similarity between the model template and candidate templates is confirmed by corresponding minimum Minkowski distance (MD). Subsequently, best match candidate template will be updated by model template in view of tracking process. However, centroid distance function has remarked some potentials and hence it has further motivated to mine it to throw in the proposed novel criteria such as the geometric mean of segmented centroid (GMSC) distance function to track the object. The proposed method of GMSC distance function has shown significant change in the tracking outcome.

Perceptual load in different regions of the visual scene and its relevance for driving.

The aim of this study was to better understand the role played by perceptual load, at both central and peripheral regions of the visual scene, in driving safety. Attention is a crucial factor in driving safety, and previous laboratory studies suggest that perceptual load is an important factor determining the efficiency of attentional selectivity. Yet, the effects of perceptual load on driving were never studied systematically. Using a driving simulator, we orthogonally manipulated the load levels at the road (central load) and its sides (peripheral load), while occasionally introducing critical events at one of these regions. Perceptual load affected driving performance at both regions of the visual scene. Critically, the effect was different for central versus peripheral load: Whereas load levels on the road mainly affected driving speed, load levels on its sides mainly affected the ability to detect critical events initiating from the roadsides. Moreover, higher levels of peripheral load impaired performance but mainly with low levels of central load, replicating findings with simple letter stimuli. Perceptual load has a considerable effect on driving, but the nature of this effect depends on the region of the visual scene at which the load is introduced. Given the observed importance of perceptual load, authors of future studies of driving safety should take it into account. Specifically, these findings suggest that our understanding of factors that may be relevant for driving safety would benefit from studying these factors under different levels of load at different regions of the visual scene.

A computer vision system for rapid search inspired by surface-based attention mechanisms from human perception

Humans are highly efficient at visual search tasks by focusing selective attention on a small but relevant region of a visual scene. Recent results from biological vision suggest that surfaces of distinct physical objects form the basic units of this attentional process. The aim of this paper is to demonstrate how such surface-based attention mechanisms can speed up a computer vision system for visual search. The system uses fast perceptual grouping of depth cues to represent the visual world at the level of surfaces. This representation is stored in short-term memory and updated over time. A top-down guided attention mechanism sequentially selects one of the surfaces for detailed inspection by a recognition module. We show that the proposed attention framework requires little computational overhead (about 11 ms), but enables the system to operate in real-time and leads to a substantial increase in search efficiency.

A Novel Shape Signature of Geometric Mean of Segmented Centroid Distance Function to Track the Object through Fourier Descriptors

The process explicitly dedicated to estimate the path of the object as it moves along the region of scene in the image plane is principle of tracking. In other words it is a strategy to detect and track moving object through a sequence of frames. In this work, optical flow based on Horn-Schunk with Barren, Fleet and Beuchemin (BFB) kernel has been employed to estimate the motion vectors. The peripheries of moving objects are extracted for different shape signatures such as boundary, edge, area, curvature and centroid distance functions. Fourier descriptors of particular shape signature for each of the candidate and model template are extracted. Similarity between the model and candidate templates is confirmed by corresponding minimum Minkowski distance. Subsequently, best match candidate will be updated. However, centroid distance function has remarked some potential. Thus, it further motivated to mine exhaustively to throw in the proposed novel criteria such as the Geometric Mean of Segmented Centroid (GMSC) distance function to track the object. The proposed method of GMSC distance function has shown significant change in the tracking outcome.

Region specific magnification on digital holograms at 100 frames per second

We report a fast method for magnifying a selected region of the pictorial content that has been recorded in a digital hologram, without the need of regenerating the latter from the original object scene. In essence, the hologram is first back-projected to a complex, two dimensional virtual diffraction plane (VDP) which is located at close proximity to the original object points. Next, the fringe patterns in the selected region of the VDP, which corresponds to the complex light wave emerged from the part of the object within the coverage of the region, is transformed to achieve the magnification effect. Subsequently, the processed field distribution on the VDP is expanded into a full hologram. We have applied our proposed method to realize a magnifying glass effect, so that the reconstructed image of the hologram will zoom into a specific region of the object scene. For a medium size digital hologram comprising of 2048 × 2048 pixels, the magnifying process can be realized in less than 10ms (equivalent to a rate of around 100 frames per second). To the best of our knowledge, this is the first time region specific magnification at such a high frame rate is considered and achieved in the context of digital holography.

Visual Scene Perception in Navigating Wood Ants

Ants, like honeybees, can set their travel direction along foraging routes using just the surrounding visual panorama. This ability gives us a way to explore how visual scenes are perceived. By training wood ants to follow a path in an artificial scene and then examining their path within transformed scenes, we identify several perceptual operations that contribute to the ants' choice of direction. The first is a novel extension to the known ability of insects to compute the "center of mass" of large shapes: ants learn a desired heading toward a point on a distant shape as the proportion of the shape that lies to the left and right of the aiming point--the 'fractional position of mass' (FPM). The second operation, the extraction of local visual features like oriented edges, is familiar from studies of shape perception. Ants may use such features for guidance by keeping them in desired retinal locations. Third, ants exhibit segmentation. They compute the learned FPM over the whole of a simple scene, but over a segmented region of a complex scene. We suggest how the three operations may combine to provide efficient directional guidance.

An approach based on constrained nonnegative matrix factorization to unmix hyperspectral data

Nonnegative matrix factorization (NMF) has been recently applied to solve the hyperspectral unmixing problem because it ensures nonnegativity and needs no assumption for the presence of pure pixels. However, the algorithm has a large amount of local minima due to the obvious nonconvexity of the objective function. In order to improve its performance, auxiliary constraints can be introduced into the algorithm. In this paper, we propose a new approach named abundance separation and smoothness constrained NMF by introducing two constraints, namely, abundance separation and smoothness, into the NMF algorithm. These constraints are based on two properties of hyperspectral imagery. First, usually, every ground object presents dominance in a specific region of the entire image scene and the correlation is weak between different endmembers. Second, moving through various regions, ground objects usually vary slowly and abrupt changes rarely appear. We also propose a learning algorithm to further improve the performance of our method, from which the auxiliary constraints are removed at an appropriate time. The proposed algorithm retains all the advantages of NMF and effectively overcomes the shortcoming of local minima at the same time. Experimental results based on synthetic and real hyperspectral data show the superiority of the proposed algorithm with respect to other state-of-the-art approaches.

Optimization of the Digital Near-Field Beamforming for Underwater 3-D Sonar Imaging System

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> This paper proposes a novel approximation for the delays of digital near-field beamforming based on a planar array. An optimized algorithm for the underwater 3-D sonar imaging system based on the novel approximation will also be described. The algorithm is applied to both an evenly spaced planar array and an unevenly spaced planar array. A data path for the optimized algorithm will be illustrated. Comparisons of the memory and the computational requirements between the optimized algorithm and the direct method (DM) beamforming will be presented. Comparisons of the proposed approximation and the Fresnel approximation will also be demonstrated. This paper illustrates that the proposed approximation enlarges the validity region of the system's view scene as compared with the Fresnel approximation. The optimized algorithm has the advantage of reducing the memory requirements for parameter storage comparing with DM beamforming. </para>

Sustained division of the attentional spotlight.

By voluntarily directing attention to a specific region of a visual scene, we can improve our perception of stimuli at that location. This ability to focus attention upon specific zones of the visual field has been described metaphorically as a moveable spotlight or zoom lens that facilitates the processing of stimuli within its 'beam'. A long-standing controversy has centred on the question of whether the spotlight of spatial attention has a unitary beam or whether it can be divided flexibly to disparate locations. Evidence supporting the unitary spotlight view has come from numerous behavioural and electrophysiological studies. Recent experiments, however, indicate that the spotlight of spatial attention may be divided between non-contiguous zones of the visual field for very brief stimulus exposures (&<100 ms). Here we use an electrophysiological measure of attentional allocation (the steady-state visual evoked potential) to show that the spotlight may be divided between spatially separated locations (excluding interposed locations) over more extended time periods. This spotlight division appears to be accomplished at an early stage of visual-cortical processing.

Attentional modulation in visual cortex depends on task timing

Paying attention to a stimulus selectively increases the ability to process it. For example, when subjects attend to a specific region of a visual scene, their sensitivity to changes at that location increases. A large number of studies describe the behavioural consequences and neurophysiological correlates of attending to spatial locations. There has, in contrast, been little study of the allocation of attention over time. Because subjects can anticipate predictable events with great temporal precision, it seems probable that they might dynamically shift their attention when performing a familiar perceptual task whose constraints changed over time. We trained monkeys to respond to a stimulus change where the probability of occurrence changed over time. Recording from area V4 of the visual cortex in these animals, we found that the modulation of neuronal responses changed according to the probability of the change occurring at that instant. Thus, we show that the attentional modulation of sensory neurons reflects a subject's anticipation of the timing of behaviourally relevant events.

Semianalytic Moderate‐Resolution Imaging Spectrometer algorithms for chlorophyll a and absorption with bio‐optical domains based on nitrate‐depletion temperatures

This paper describes algorithms for retrieval of chlorophyll a concentration and phytoplankton and gelbstoff absorption coefficients for the Moderate‐Resolution Imaging Spectrometer (MODIS) or sensors with similar spectral channels. The algorithms are based on a semianalytical, bio‐optical model of remote sensing reflectance, Rrs(λ). The Rrs(λ) model has two free variables, the absorption coefficient due to phytoplankton at 675 nm, aϕ(675), and the absorption coefficient due to gelbstoff at 400 nm, ag(400). The Rrs model has several parameters that are fixed or can be specified based on the region and season of the MODIS scene. These control the spectral shapes of the optical constituents of the model. Rrs(λi) values from the MODIS data processing system are placed into the model, the model is inverted, and aϕ(675), ag(400), and chlorophyll a are computed. The algorithm also derives the total absorption coefficients a(λi) and the phytoplankton absorption coefficients aϕ(λi) at the visible MODIS wavelengths. MODIS algorithms are parameterized for three different bio‐optical domains: (1) high photoprotective pigment to chlorophyll ratio and low self‐shading, which for brevity, we designate as “unpackaged”; (2) low photoprotective pigment to chlorophyll ratio and high self‐shading, which we designate as “packaged”; and (3) a transitional or global‐average type. These domains can be identified from space by comparing sea‐surface temperature to nitrogen‐depletion temperatures for each domain. Algorithm errors of more than 45% are reduced to errors of less than 30% with this approach, with the greatest effect occurring at the eastern and polar boundaries of the basins.

Quantitative calculation of characteristic volume of surface fire behavior

In the description of the characteristic volume of fire behavior, fire intensityI, rate of spreadR, and flame lengthL, etc. all have their scientific definitions, while the bound between the firclinc and the residual region of a fire scene is not yet scientifically defined. This paper tries to find out theoretically the basis for dividing the bound, to give the fireline thicknessD a scientific definition, and to propose a method of quantitative calculation. According to the generality of fuels, the time-dependent curves of fuels combustion rate are utilized to work out a formula by, experience. On the basis of defining scientifically the fireline thickness, the Byram’s formula is corrected, and a new one for calculating the fire intensity of surface fire, i. e.I = 0.788HWr, is presented. In this formula,H is the fuel low heat of combustion, the calculating method of the thermal current intensityJ is given, andJ max = 1.25HWR/D, the calculating formula of flame length is revised, andL = 0.077I 0.46. In the spreading process of the forest fire, when the time needed for getting over a fireline thickness isT, it is called residence time. It is given out in this paper the method for estimating and determining the rate of spreadr and the residence timeT in the combustion chamber. And on this basis the characteristic volume (W, I, R, D, T, L andJ max, etc.) of surface fire behavior can be evaluated.