Crowd Flow Segmentation Using a Novel Region Growing Scheme

The advent of the COVID-19 pandemic has adversely affected the entire world and has put forth high demand for techniques that remotely manage crowd-related tasks. Video surveillance and crowd management using video analysis techniques have significantly impacted today's research, and numerous applications have been developed in this domain. This research proposed an anomaly detection technique applied to Umrah videos in Kaaba during the COVID-19 pandemic through sparse crowd analysis. Managing the Kaaba rituals is crucial since the crowd gathers from around the world and requires proper analysis during these days of the pandemic. The Umrah videos are analyzed, and a system is devised that can track and monitor the crowd flow in Kaaba. The crowd in these videos is sparse due to the pandemic, and we have developed a technique to track the maximum crowd flow and detect any object (person) moving in the direction unlikely of the major flow. We have detected abnormal movement by creating the histograms for the vertical and horizontal flows and applying thresholds to identify the non-majority flow. Our algorithm aims to analyze the crowd through video surveillance and timely detect any abnormal activity to maintain a smooth crowd flow in Kaaba during the pandemic.

Attributes now play a vital role for characterizing a crowded scene. Compared to low-level visual features, processing informed by attributes can capture rich semantic information. However, to effectively assign attributes to a crowded scene still remains a challenging task. In this letter, inspired by a recently proposed zero-shot learning framework, a novel attribute assignment method that maps low-level features to predefined attributes is proposed. In particular, we propose to exploit the attribute dependency during the phase of attribute assignment, which can be regarded as our main contribution. In addition, to further enhance the performance, an effective low-level feature extraction mechanism is also proposed. More precisely, appearance and motion features are first simultaneously extracted from several sampled video frames and corresponding optical flow fields via deep convolutional neural network and then, respectively, aggregated by using Fisher vector encoding to form the low-level representation of crowded scenes. Experimental results on the challenging WWW dataset demonstrate that both the proposed attribute assignment method and the low-level feature extraction mechanism outperform the state of the art.

Let Ω be a bounded and subanalytic domain in R n , n > 2. We show that the set of boundary points of Ω which are regular with respect to the Dirichlet problem is again subanalytic. Moreover, we give sharp upper bounds for the dimension of the set of irregular boundary points. This enables us to decide whether the domain has a classical Green function. In dimensions 2 and 3, this is the case, given some mild and necessary conditions on the topology of the domain.

Dense matching plays an important role in many fields, such as DEM (digital evaluation model) producing, robot navigation and 3D environment reconstruction. Traditional approaches may meet the demand of accuracy. But the calculation time and out puts density is hardly be accepted. Focus on the matching efficiency and complex terrain surface matching feasibility an aerial image dense matching method based on optical flow field is proposed in this paper. First, some high accurate and uniformed control points are extracted by using the feature based matching method. Then the optical flow is calculated by using these control points, so as to determine the similar region between two images. Second, the optical flow field is interpolated by using the multi-level B-spline interpolation in the similar region and accomplished the pixel by pixel coarse matching. Final, the results related to the coarse matching refinement based on the combined constraint, which recognizes the same points between images. The experimental results have shown that our method can achieve per-pixel dense matching points, the matching accuracy achieves sub-pixel level, and fully meet the three-dimensional reconstruction and automatic generation of DSM-intensive matching’s requirements. The comparison experiments demonstrated that our approach’s matching efficiency is higher than semi-global matching (SGM) and Patch-based multi-view stereo matching (PMVS) which verifies the feasibility and effectiveness of the algorithm.

Anomaly detection in crowd scene has attracted an increasing attention in video surveillance, but a precise detection still remains a challenge. This paper presents a novel online learning method to automatically detect abnormal behaviors in crowd scene. Our focus is mainly on the deviation between the real motion and the predicted one. Through online defining experts, analyzing their motions, and dynamically updating the learned model, anomaly can be identified by the final expert joint decision. The outputs are represented as the anomaly probability of an examined frame. Compared with most of existing methods, the proposed one needs neither tracking each individual straight to the end nor requires any complex training procedure. We test the proposed method on public datasets, and the results show its effectiveness.

Obstacle detection by evaluation of optical flow fields from image sequences

Performance evaluation of selected complex video processing algorithms, implemented on a parallel, embedded GPU platform Tegra X1, is presented. Three algorithms were chosen for evaluation: a GMM-based object detection algorithm, a particle filter tracking algorithm and an optical flow based algorithm devoted to people counting in a crowd flow. The choice of these algorithms was based on their computational complexity and parallel structure. The aim of the experiments was to assess whether the current generation of low-power, mobile GPUs has sufficient power for running live analysis of video surveillance streams, e.g. in smart cameras, while maintaining energy consumption at a reasonable level. Tests were performed with both a synthetic benchmark and a real video surveillance recording. It was found that the computational power of the tested platform is sufficient for running operations such as background subtraction, but in case of more complex algorithms, such as tracking with particle filters, performance is not satisfactory because of inefficient memory architecture which stalls the processing.

We present a method for unsupervised learning of classes of motions in video. We project optical flow fields to a complete, orthogonal, a-priori set of basis functions in a probabilistic fashion, which improves the estimation of the projections by incorporating uncertainties in the flows. We then cluster the projections using a mixture of feature-weighted Gaussians over optical flow fields. The resulting model extracts a concise probabilistic description of the major classes of optical flow present. The method is demonstrated on a video of a person's facial expressions.

In an optical flow field, the background and moving objects present different vector groups with different directions, velocities and region areas. The idea optical flow field is not easy to obtain for some kinds of reasons; in practical field, the motion vectors present confusion and uncertainty to some extent. The fuzzy clustering provides an effective way to process unclear classification. It maps every vector into every group, and the ascription presents a degree a vector belongs to a group. However, conventional fuzzy clustering method needs to determine the group number, namely the moving objects number in the view field. Before all samples are processed and the group number is fixed during iteration. The unsuitable number easily results in inaccurate segmentation. In view of this problem, an enhanced detection algorithm using fuzzy clustering with elastic grouping logic is proposed. To be called elastic grouping logic, it means that in the process of optical flow field detection, according to the ascription the vector to each group, together with the vector's location, direction and magnitude, the group number, namely the moving object number, is selfadaptively generated, and further to achieve the moving objects segmentation with precision. A stability model of motion vectors for an object group and the group's partition is also established. The experimental results illustrate the proposed algorithm is able to satisfy the need of multi-objects detection and locate the moving objects successfully.

We propose the fast optical flow extractor, a filtering scheme that recovers artifact-free optical flow fields from HEVC-compressed video. To extract accurate optical flow fields, we form a regularized optimization problem that considers the smoothness of the solution and the pixelwise confidence weights of the artifact-ridden HEVC motion field. Solving such an optimization problem is slow, so we first convert the problem into a confidence-weighted filtering task. We then use the extracted optical flow fields for fast synthetic refocus of video using camera motion. By leveraging the already-available HEVC motion information, we achieve a 10-fold speed-up in the running times compared to similar methods. Our fast optical flow extractor is useful since video signals encountered in real life are typically available already in coded formats. Our fast optical flow extractor is general, and works on motion fields of any video coder.

The notion of a boundary belongs in the canon of the most important notions of mereotopology, the topological theory induced by mereological structures; the importance of this notion rests not only in its applications to practical spatial reasoning, e.g., in geographical information systems, where it is usually couched under the term of a contour and applied in systems related to economy, welfare, climate, wildlife etc., but also in its impact on reasoning schemes elaborated for reasoning about spatial objects, represented as regions, about spatial locutions etc. The difficulty with this notion lies primarily in the fact that boundaries are things not belonging in mereological universa of things of which they are boundaries. Various authors, from philosophers through mathematicians to logicians and computer scientists proposed schemes for defining and treating boundaries. We propose two approaches to boundaries; the first aims at defining boundaries as things possibly in the universe in question, i.e., composed of existing things, whereas the second defines them as things in a meta–space built over the mereological universe in question, i.e., we assume a priori that boundaries are in a sense 'things at infinity', in an agreement with the topological nature of boundaries. Of the two equivalent topological definitions of a boundary, the first, global, defining the boundary as the difference between the closure and the interior of the set, and the second, local, defining it as the set of boundary points whose all neighborhoods transect the set, the first calls for the first type of the boundary and the second is best fitted for the meta–boundary. In the text that follows, we discuss mereology and rough mereology notions (sects. 2, 3), the topological approach to the notion of a boundary and the model ROM with which we illustrate our discussion (sect. 4), the mereology approach (sect. 5), and the approach based on rough mereology and granular computing in the framework of rough mereology (sect. 6).

This paper is about high-level modeling of analog circuits based on its complete feasible performance space. The behavioral model is a multi-dimensional description of all enclosing surfaces of this space, approximated by a set of boundary points. An exploration method, applying the Normal Boundary Intersection with a novel constraint rotation method, is shown to calculate these boundary points. The solutions are found by transistor level spice simulations, which leads to a highly accurate approximated model. A fully-differential transconductance amplifier with folded cascode and class-AB output stage is taken as example circuit, whose performance space model is calculated by the deterministic exploration scheme. The resulting model is used to perform an efficient system level optimization towards accuracy of a switched capacitor circuit.

We introduce a new kind of regularity in the Dirichlet problem for the heat equation, called mild regularity. The mildly regular boundary points include all types of regular points previously considered. Just as regular boundary points have barriers associated with them, the mildly regular boundary points have mild barriers, which are just as important. We consider whether the mild regularity of a boundary point for an open set is inherited by an open subset or superset. We detail the connection of mild regularity with the Green function. We consider, in particular, subdomains of an arbitrary open set that satisfy a certain connectivity condition which takes account of the special nature of the temporal variable, called zones of influence. Our main theorem states that, for any zone of influence, the set of boundary points that are not mildly regular is caloric measure null.

Identification methods for the dynamic systems are considered with respect to electric drives. It is shown that electric drives are complex electromechanical systems, the mechanical part of which is multimass and contains elements with elastic connections. The features of these systems that require considering the gaps in gear units when solving modeling and identification problems; elasticity arising in the shafts; stiffness change of elastic elements in couplings. It is noted that the identification problem of electric drives is a subproblem of managing the state of electromechanical systems. The basis of methods and algorithms for its solution is a set is information about the performance limits of the electric drive. It is shown that identification of the controlled parameters of the electric drive requires using statistical methods based on the expansion of the pulse transient functions to a series of the orthonormal Laguerre transformed system of functions. The method has been proposed in order to assess the condition of the electric drive during operation. This method involves setting the performance limits as a set of boundary points. A distinctive feature of the method is the high identification accuracy.

The paper is devoted to the study of Toeplitz operators with piecewise continuous symbols. We clarify the geometric regularities of the behavior of the essential spectrum of Toeplitz operators in dependence on their crucial data: the angles between jump curves of symbols at a boundary point of discontinuity and on the limit values reached by a symbol at that boundary point. We show then that the curves supporting the symbol discontinuities, as well as the number of such curves meeting at a boundary point of discontinuity, do not play any essential role for the Toeplitz operator algebra studied. Thus we exclude the curves of symbol discontinuity from the symbol class definition leaving only the set of boundary points (where symbols may have discontinuity) and the type of the expected discontinuity. Finally we describe the C*-algebra generated by Toeplitz operators with such symbols.