Crowd Density Research Articles

EXPRESS: Optimizing Service Operations with Price- and Density-Dependent Demand: A Copula-Based Approach

Exclusivity is a key component of certain service offerings such as elite members-only social or country clubs. Indeed, such services may only appeal to customers if they limit the crowd density, i.e., the maximal number of customers granted access to the service. We consider a service provider, who sets its crowd density cap in addition to its price. Customers are heterogenous in both their willingness to pay (WTP) for the service and their density tolerance. Each customer purchases the service only if the price is below her WTP and the density is below her tolerance. Importantly, customers’ WTP and density tolerance may be statistically dependent, and we develop a novel, copula-based framework to model the dependence structure of these two dimensions of heterogeneity. We analytically characterize the provider’s optimal price and density decisions. As long as the customer population is not too severely sensitive to density, (i) the provider optimally serves all segments of the market and sets the price and density so that the price, not the density, is the determining factor for customers (from a particular sub-population) on the margin between purchasing and not and (ii) as customers’ valuations and density tolerances become more positively dependent, the provider earns higher revenue and optimally increases its service density. By contrast, the optimal price is quasi-convex but not necessarily increasing. On the other hand, with severely density-sensitive customers, it may be optimal to partially cover the market and to make density the determining factor in customer decisions. Our findings offer prescriptive guidelines for service operations in the presence of density-sensitive demand, and also provide an explanation for the failed versus successful practices of prominent private clubs.

Discrete Bio-Hazard Model of EM Radiation for Crowded People

This study investigates the behaviour of electromagnetic (EM) waves in densely populated environments using a Multicomponent Discrete Propagation Model (MCDM). Our analysis focuses on four critical parameters: Specific Absorption Rate (SAR), Electrical Field, Skin Depth, and Effective Radiated Power (ERP) by working on crowd densities and frequencies relevant to mobile technologies. This study is the first to employ a Propagation Model for such an analysis, offering a novel approach. Through comprehensive simulations, we explore how variations in electrical field strengths impact SAR, skin depth, and ERP. The results are compared to established safety limits for whole-body SAR exposure, providing valuable insights for guidelines. This research aims to contribute to designing future electronic devices that minimize overall RF emissions. This innovative approach has the potential to improve public confidence in the safety of wireless technologies significantly

Predicting the impact of public events and mobility in Smart Cities

AbstractThe ubiquitous presence of smartphones and the ever‐expanding Internet of Things are generating a treasure trove of data on human movement. We harness the power of Artificial Intelligence to extract knowledge within this data, in particular for predicting people flows and density in a Smart City. This predictive ability holds immense potential for a multitude of applications, from optimising people flow to streamlining event planning, while offering a powerful tool for pre‐emptive identification of situations that may lead to crowd disasters. In this paper, we tackle two crucial aspects of people mobility using data from public events and an Italian mobile phone network: to predict both event attendance and future crowd density in specific areas. The event details (location, time etc.) are automatically gathered and stored in a structured format. Next, we handle these problems are treated in a “supervised learning” setting, and various state‐of‐art Machine Learning techniques are tested to find the best model for each task. The obtained models will be encapsulated into a Policy Support System contributing to foster planning actions of mobility services.

MGFNet: Cross-scene Crowd Counting via Multistage Gated Fusion Network

Existing crowd counting methods are mainly trained and tested in similar scenarios. When the testing and training scenarios of the model are different, the counting accuracy of these methods will sharply decrease, which seriously limits their practical application. To address this problem, we propose a multistage gated fusion network (MGFNet) for cross-scene crowd counting. MGFNet is primarily composed of dynamic gated convolution units (DGCU) and multilevel scale attention blocks (MSAB) modules. Specifically, DGCU uses a dynamic gating path to supplement detailed information to reduce the loss of crowd information and overestimation of background in different scenarios. MSAB calibrates crowd information at different scales and perspectives in different scenes by generating attention maps with discriminative information. In addition, we used a new global local consistency loss to optimize the model to adapt to changes in crowd density and distribution. Extensive experiments on four different types of scene counting benchmarks show that the proposed MGFNet achieves superior cross-scene counting performance.

Crowd Density Estimation via Global Crowd Collectiveness Metric

Drone-captured crowd videos have become increasingly prevalent in various applications in recent years, including crowd density estimation via measuring crowd collectiveness. Traditional methods often measure local differences in motion directions among individuals and scarcely handle the challenge brought by the changing illumination of scenarios. They are limited in their generalization. The crowd density estimation needs both macroscopic and microscopic descriptions of collective motion. In this study, we introduce a Global Measuring Crowd Collectiveness (GMCC) metric that incorporates intra-crowd and inter-crowd collectiveness to assess the collective crowd motion. An energy spread process is introduced to explore the related crucial factors. This process measures the intra-crowd collectiveness of individuals within a crowded cluster by incorporating the collectiveness of motion direction and the velocity magnitude derived from the optical flow field. The global metric is adopted to keep the illumination-invariance of optical flow for intra-crowd motion. Then, we measure the motion consistency among various clusters to generate inter-crowd collectiveness, which constitutes the GMCC metric together with intra-collectiveness. Finally, the proposed energy spread process of GMCC is used to merge the inter-crowd collectiveness to estimate the global distribution of dense crowds. Experimental results validate that GMCC significantly improves the performance and efficiency of measuring crowd collectiveness and crowd density estimation on various crowd datasets, demonstrating a wide range of applications for real-time monitoring in public crowd management.

Robust Estimation of Crowd Density Using Vision Transformers

Estimating and predicting crowd density at large events or during disasters is of paramount importance for enhancing emergency management systems. This includes planning effective evacuation routes, optimizing rescue operations, and ensuring efficient deployment of emergency services. Traditionally, surveillance systems that rely on cameras have been employed to monitor crowd movements. However, accurately estimating crowd density using such systems presents several challenges. These challenges stem primarily from the interaction between large crowds and the limitations of two-dimensional cameras in capturing the full scope of three-dimensional spaces. Optical distortions, environmental factors, and variations in camera angles further complicate the task, making accurate estimations difficult to achieve. To address these challenges, this paper introduces a robust method for calculating crowd density that leverages advanced vision transformers. By combining the output of these transformers with a two-stage neural network, the method effectively mitigates the limitations of traditional approaches. One of the key advantages of the proposed system is its robustness, which allows it to perform well across different camera specifications, installation locations, and image aspect ratios. The method applies and evaluates various deep learning techniques, introducing improvements to existing network structures that are better suited for the problem at hand. Extensive experimental verification demonstrates that the proposed method consistently produces accurate crowd density estimates, even in diverse and complex crowd environments. This robust performance underscores its potential for improving emergency management and crowd control in real-world situations.

Research on Improved Crowd Detection Based on YOLOv5

With the acceleration of the process of modern urbanization and the improvement of residents' material living standards, the flow of people in the public space is gradually becoming saturated. The monitoring equipment in public places records a huge amount of people flow information all the time, but due to the crowds tend to be dense and crowded. Traditional machine learning cannot make accurate and efficient identification of a large number of dense crowds, if the deep learning technology can be used to process the crowded crowd captured by the surveillance camera and accurately identify the number of people in public places, it provides an important guarantee for the flow of people in public areas and safety construction. However, for crowded targets with occlusions, the traditional target detection algorithm sometimes performs poorly. Based on the above background, this paper introduces an enhanced deep learning framework utilizing the YOLOv5 neural network for crowd detection research. aiming at the characteristics of dense and crowded crowds in public areas. By improving convolutional layer C3 in the backbone structure of YOLOv5 neural network and adding CBAM attention mechanism. Compared with the original YOLOv5, the improved model has increased the maximum F1 value of crowd recognition at near, middle and far distances. To sum up, the deep learning framework improved by YOLOv5 neural network proposed in this paper has significantly improved the recognition accuracy of crowded people in public areas.

Data governance for soundscape studies

Soundscape research involves a complex array of data types, as soundscapes are defined by how acoustic environments are perceived and experienced by individuals within specific contexts. These studies typically incorporate a variety of data, including acoustic measurements, psychometric evaluations (both qualitative and quantitative), such as assessments of loudness and sharpness, as well as visual data (images and videos), assessments of crowd density (evaluating both the number of individuals and their activities), and geospatial and temporal data (detailing location specifics, environmental conditions like temperature and humidity, and various times of day). The collection, analysis, and presentation of this data are crucial steps in the research process. This paper aims to describe the critical data governance issues and propose solutions for managing data in soundscape studies.

Dynamic response and vibration suitability analysis of the large-span double-connected structure under coupled wind and pedestrian loads

Modern buildings increasingly utilize lightweight, high-strength materials and feature high-rise, large-span structural designs. These structures often exhibit low natural frequencies and are susceptible to resonance from low-frequency dynamic loads such as wind and pedestrian loads. This paper focuses on a large-span double-connected structure and analyzes its dynamic response under the combined effects of wind and pedestrian loads. First, a finite element model of the structure was created using ANSYS, and model validity verification and modal analysis were performed. Second, a Fourier-based pedestrian model was used to simulate pedestrian loads and generate time-range data. The pulsating wind speed was generated from the Davenport spectrum using the harmonic superposition method. Wind load time-range data were calculated for different heights using Bernoulli’s theorem. Finally, the solution yields information about the dynamic response of the structure. The study revealed maximum vertical comfort ratings in the connecting corridor were achieved when crowd density did not exceed 0.3 persons/m2. The connecting corridor’s most unfavorable horizontal comfort level was evaluated as a medium, except for the 0-degree wind angle condition. This paper provides experience in studying the dynamic response and vibration suitability assessment of the large-span double-connected structure under wind and pedestrian loads.

Human behavior and decision-making in earthquake emergencies: insights from virtual reality and serious games

Earthquakes can severely impact public spaces with complex layouts and high crowd densities, resulting in devastating consequences and deaths. Gaining insights into the actions of individuals during an earthquake can provide valuable information about areas where their decision-making may be flawed. Addressing these areas can improve safety measures and better protect people during emergencies. Moreover, acquiring this knowledge can enhance evacuation strategies for future earthquakes. Immersive Virtual Reality and Serious Games (IVR SGs) present an effective method for examining and understanding how people behave in emergency scenarios, such as earthquakes. Using IVR SGs, we created several potential scenarios during an earthquake in a shopping mall to explore how people act in an emergency when the evacuees feel pressure, urgency, and stress. A total of 146 participants were involved in the study. We investigated the human tendency to follow authority figures’ instructions (rational and irrational), help injured people of different genders and ages, and social influence during an emergency. Moreover, we investigated how familiarity with the environment impacts human wayfinding performance when evacuating the building.

Accurate crowd counting for intelligent video surveillance systems

The paper presents a novel deep learning approach for crowd counting in intelligent video surveillance systems, addressing the growing need for accurate monitoring of public spaces in urban environments. The demand for precise crowd estimation arises from challenges related to security, public safety, and efficiency in urban areas, particularly during large public events. Existing crowd counting techniques, including feature-based object detection and regression-based methods, face limitations in high-density environments due to occlusions, lighting variations, and diverse human figures. To overcome these challenges, the authors propose a new deep encoder-decoder architecture based on VGG16, which incorporates hierarchical feature extraction with spatial and channel attention mechanisms. This architecture enhances the model’s ability to manage variations in crowd density, leveraging adaptive pooling and dilated convolutions to extract meaningful features from dense crowds. The model’s decoder is further refined to handle sparse and crowded scenes through separate density maps, improving its adaptability and accuracy. Evaluations of the proposed model on benchmark datasets, including Shanghai Tech and UCF CC 50, demonstrate superior performance over state-of-the-art methods, with significant improvements in mean absolute error and mean squared error metrics. The paper emphasizes the importance of addressing environmental variability and scale differences in crowded environments and shows that the proposed model is effective in both sparse and dense crowd conditions. This research contributes to the advancement of intelligent video surveillance systems by providing a more accurate and efficient method for crowd counting, with potential applications in public safety, transportation management, and urban planning.

SPCANet: congested crowd counting via strip pooling combined attention network.

Crowd counting aims to estimate the number and distribution of the population in crowded places, which is an important research direction in object counting. It is widely used in public place management, crowd behavior analysis, and other scenarios, showing its robust practicality. In recent years, crowd-counting technology has been developing rapidly. However, in highly crowded and noisy scenes, the counting effect of most models is still seriously affected by the distortion of view angle, dense occlusion, and inconsistent crowd distribution. Perspective distortion causes crowds to appear in different sizes and shapes in the image, and dense occlusion and inconsistent crowd distributions result in parts of the crowd not being captured completely. This ultimately results in the imperfect capture of spatial information in the model. To solve such problems, we propose a strip pooling combined attention (SPCANet) network model based on normed-deformable convolution (NDConv). We model long-distance dependencies more efficiently by introducing strip pooling. In contrast to traditional square kernel pooling, strip pooling uses long and narrow kernels (1×N or N×1) to deal with dense crowds, mutual occlusion, and overlap. Efficient channel attention (ECA), a mechanism for learning channel attention using a local cross-channel interaction strategy, is also introduced in SPCANet. This module generates channel attention through a fast 1D convolution to reduce model complexity while improving performance as much as possible. Four mainstream datasets, Shanghai Tech Part A, Shanghai Tech Part B, UCF-QNRF, and UCF CC 50, were utilized in extensive experiments, and mean absolute error (MAE) exceeds the baseline, which is 60.9, 7.3, 90.8, and 161.1, validating the effectiveness of SPCANet. Meanwhile, mean squared error (MSE) decreases by 5.7% on average over the four datasets, and the robustness is greatly improved.

A novel extended social force model for studying the impact of the heterogeneity of pedestrian physical fitness on emergency evacuation efficiency

Emergencies in public places, particularly confined crowded areas, will disrupt the stability of dense crowds and consequently lead to accidents. To promote public emergency safety, there is a pressing need for efficient modeling methods to investigate the evacuation mechanism in these places and improve the social public safety. This study proposes a Physical Fitness Heterogeneity based Social Force Model (PFH-SFM) that takes into account the heterogeneous desired evacuation velocity caused by the heterogeneity of pedestrian physical fitness, by means of developing the normalized desired velocity ratio. Then, we use PFH-SFM to investigate the relationships between the escape rate and the desired velocity, and between the evacuation duration and the desired velocity in terms of various group sizes with heterogeneous physical fitness, the relationship between the percentage of reduction in evacuation duration and desired velocity when including weak pedestrians, the pedestrian distribution in the evacuation process, the relationship between the total evacuation duration and the desired velocity in terms of various proportions of weak pedestrians and the relationship between the evacuation duration and the desired velocity in terms of various normalized starting and ending velocity ratios by considering various group sizes, respectively. The findings of this study show that the existence of a certain small proportion of pedestrians with weak physical fitness can promote global evacuation dynamics, especially in the case of high crowded density, and can reduce evacuation duration by up to 20% in our experiments. Additionally, when the percentage of pedestrians with weak physical fitness is relatively high, they tend to have a detrimental effect on the evacuation efficiency. Furthermore, there exists a moderate normalized desired starting velocity ratio that maximizes the overall evacuation efficiency; on the other hand, the lower the normalized desired ending velocity ratio is, the more efficient the evacuation is. To the best of the authors’ knowledge, this study is the first time to introduce the concepts of normalized desired starting and ending velocity ratios and innovatively analyzes the impact of the continuously changing desired velocity of pedestrians on the evacuation efficiency in multi-exit scenarios. The results offer valuable insights for relevant stakeholders to formulate effective evacuation plans, so as to enhance urban emergency capacity and minimize social and economic losses.

Study on the propagation characteristics of shockwave in dense crowd in corner passage

Crowd shockwaves are prone to occur when evacuating dense crowds, especially in areas with corner passages, leading to congestion and accidents. The characteristics of shockwaves in corner passages need further exploration. This study used AnyLogic software based on the social force model to construct scenarios. The propagation mechanism of crowd shockwaves was revealed through simulation analysis of the changes in wave amplitude, duration, and propagation velocity of crowd density waves under different corner angles (0–90°). It was found that a 45° corner passage has advantages compared to other corner passages for evacuation. At higher desired velocities, pedestrians form a bottleneck-like area at the corner of the passage, resulting in crowd shockwaves. Because of social forces, congestion occurs primarily in the corner area. To alleviate this pressure, pedestrians actively adjust the distance between themselves and others. The time interval between the two shockwaves decreases as the desired velocity increases. The propagation time of crowd shockwaves before the corner is generally more significant than the duration after the corner. This study could guide the elimination or reduction of crowd shockwaves, improving pedestrians' evacuation efficiency and safety in evacuation passages.

Dense pedestrian face detection in complex environments

To address the problem of dense crowd face detection in complex environments, this paper proposes a face detection model named Deep and Compact Face Detection (DCFD), which adopts an improved lightweight EfficientNetV2 network to replace the backbone network of RetinaFace. A large kernel attention mechanism is introduced to address the face detection task more accurately. The backbone network, an improved efficient channel attention (ECA) mechanism, is added to further improve the algorithm performance. The feature fusion module is an improved neural architecture search feature pyramid network (NAS-FPN) that significantly improves the face detection accuracy in different scenes. To balance the training process of positive and negative samples, we use the focus loss function to replace the traditional cross-entropy loss function. In different environments, the DCFD algorithm has shown efficient face detection performance. This algorithm provides not only a feasible and effective solution for solving the problem of face detection in dense groups but also an important basis for improving the accuracy of face detection models in practical applications.

The use of convolutional neural networks for abnormal behavior recognition in crowd scenes

This study introduces the Abnormality Converging Scene Analysis Method (ACSAM) to detect abnormal group behavior using monitored videos or CCTV images in crowded scenarios. Abnormal behavior recognition involves classifying activities and gestures in continuous scenes, which traditionally presents significant computational challenges, particularly in complex crowd scenes, leading to reduced recognition accuracy. To address these issues, ACSAM employs a convolutional neural network (CNN) enhanced with Abnormality and Crowd Behavior Training layers to accurately detect and classify abnormal activities, regardless of crowd density. The method involves extracting frames from the input scene and using CNN to perform conditional validation of abnormality factors, comparing current values with previous high values to maximize detection accuracy. As the abnormality factor increases, the identification rate improves with higher training iterations. The system was tested on 26 video samples and trained on 34 samples, demonstrating superior performance to other approaches like DeepROD, MSI-CNN, and PT-2DCNN. Specifically, ACSAM achieved a 12.55% improvement in accuracy, a 12.97% increase in recall, and a 10.23% enhancement in convergence rate. These results suggest that ACSAM effectively overcomes the computational challenges inherent in crowd scene detection, offering a robust solution for real-time abnormal behavior recognition in crowded environments.

Crowd Counting Using Meta-Test-Time Adaptation.

Machine learning algorithms are commonly used for quickly and efficiently counting people from a crowd. Test-time adaptation methods for crowd counting adjust model parameters and employ additional data augmentation to better adapt the model to the specific conditions encountered during testing. The majority of current studies concentrate on unsupervised domain adaptation. These approaches commonly perform hundreds of epochs of training iterations, requiring a sizable number of unannotated data of every new target domain apart from annotated data of the source domain. Unlike these methods, we propose a meta-test-time adaptive crowd counting approach called CrowdTTA, which integrates the concept of test-time adaptation into the meta-learning framework and makes it easier for the counting model to adapt to the unknown test distributions. To facilitate the reliable supervision signal at the pixel level, we introduce uncertainty by inserting the dropout layer into the counting model. The uncertainty is then used to generate valuable pseudo labels, serving as effective supervisory signals for adapting the model. In the context of meta-learning, one image can be regarded as one task for crowd counting. In each iteration, our approach is a dual-level optimization process. In the inner update, we employ a self-supervised consistency loss function to optimize the model so as to simulate the parameters update process that occurs during the test phase. In the outer update, we authentically update the parameters based on the image with ground truth, improving the model's performance and making the pseudo labels more accurate in the next iteration. At test time, the input image is used for adapting the model before testing the image. In comparison to various supervised learning and domain adaptation methods, our results via extensive experiments on diverse datasets showcase the general adaptive capability of our approach across datasets with varying crowd densities and scales.

The Development of Marketing Plan for a Virtual Cycling Application: A Case Study of China (Beijing)

Background and Aims: Due to the COVID-19 pandemic, people's social activities have been greatly affected, so online activities have seen rapid development and user growth. become the norm. In cycling, the China Cycling Association canceled or postponed several races, affecting cyclists across the country. This is because the mass gatherings and high crowd density associated with such events increase the risk of virus transmission. Because of this, bicycle virtual riding software has developed rapidly, and users have increased rapidly. In addition, the evolving metaverse concept has prompted increased investment in bicycle virtual riding software development. This study explores and analyzes the operation plan of bicycle virtual riding software in China (Beijing) and analyzes the specific needs of Chinese people. Methodology: The main research methods include SWOT analysis, user stories, and population research. Use competitive product analysis such as SWOT analysis to study the specific reasons why other products cannot be promoted and popular in China (Beijing). Collect the main channels for users to obtain information and determine product promotion plans. And conduct user demand research, establish user portraits, and finally propose feasible solutions. Results: The main findings are the difficulties in promoting virtual bicycle riding software in Beijing, China, the reasons why the currently popular software is difficult to promote, the travel habits and consumption willingness of residents, and a complete promotion plan. Conclusion: The study sheds light on factors impeding the adoption of popular software already in use and clarifies the travel habits and consumption attitudes of the locals in Beijing, China. It also highlights challenges in promoting virtual bicycle riding software in Beijing. It provides insights into creating a strong promotion strategy to overcome the obstacles found and increase the adoption of virtual bicycle riding technology through in-depth analysis.

A Dual-Branch Convolutional Neural Network-Based Bluetooth Low Energy Indoor Positioning Algorithm by Fusing Received Signal Strength with Angle of Arrival

Indoor positioning is the key enabling technology for many location-aware applications. As GPS does not work indoors, various solutions are proposed for navigating devices. Among these solutions, Bluetooth low energy (BLE) technology has gained significant attention due to its affordability, low power consumption, and rapid data transmission capabilities, making it highly suitable for indoor positioning. Received signal strength (RSS)-based positioning has been studied intensively for a long time. However, the accuracy of RSS-based positioning can fluctuate due to signal attenuation and environmental factors like crowd density. Angle of arrival (AoA)-based positioning uses angle measurement technology for location devices and can achieve higher precision, but the accuracy may also be affected by radio reflections, diffractions, etc. In this study, a dual-branch convolutional neural network (CNN)-based BLE indoor positioning algorithm integrating RSS and AoA is proposed, which exploits both RSS and AoA to estimate the position of a target. Given the absence of publicly available datasets, we generated our own dataset for this study. Data were collected from each receiver in three different directions, resulting in a total of 2675 records, which included both RSS and AoA measurements. Of these, 1295 records were designated for training purposes. Subsequently, we evaluated our algorithm using the remaining 1380 unseen test records. Our RSS and AoA fusion algorithm yielded a sub-meter accuracy of 0.79 m, which was significantly better than the 1.06 m and 1.67 m obtained when using only the RSS or the AoA method. Compared with the RSS-only and AoA-only solutions, the accuracy was improved by 25.47% and 52.69%, respectively. These results are even close to the latest commercial proprietary system, which represents the state-of-the-art indoor positioning technology.

Convective heat transfer and drag coefficients of human body in multiple crowd densities and configurations in semi-outdoor scenarios

This study numerically assessed the impact of human crowd density and outdoor wind conditions (average velocity, its profile, and direction) on the convective heat transfer and drag coefficients (hc and Cd). Five different configurations of standing computer-simulated persons (CSPs) were tested in a semi-outdoor environment. A single isolated CSP, nine CSPs in a block array (with three representative crowd densities), and eighteen randomly allocated CSPs were used. The results indicated a significant impact of crowd density on the overall and local hc values. As the density increases, the body's obstruction against wind increases, resulting in lower heat loss. Newly proposed formulas for hc as a function of the average wind velocity (UAVE.) are (7.56 × UAVE.0.65), (8.02 × UAVE.0.64), and (8.26 × UAVE.0.63) for the high, medium, and low crowd densities, respectively. This reveals an overestimation of hc when an isolated human body is used. The hc values of the upper segments were the most affected by a 22 % reduction in the predicted hc. Moreover, when the crowd density increased, local hc and Cd decreased simultaneously, particularly in the chest, pelvis, and thigh segments. Oblique wind angles (60° and 150°) resulted in the highest hc and Cd values compared to other angles. The chest and pelvis were most affected by shifting the wind direction, indicating the dominance of these segments in concurrently controlling the thermal and drag performances. These results provide valuable insights into the optimization of human thermal and physical comfort models.