Search Results (114)

Cyber-physical infrastructures such as hospitals and smart campuses face hybrid threats that target both digital and physical domains. Traditional security solutions separate surveillance from network monitoring, leaving blind spots when attackers combine these vectors. This paper introduces ARGUS, an autonomous robotic platform designed to close this gap by correlating cyber and physical anomalies in real time. ARGUS integrates computer vision for facial and weapon detection with intrusion detection systems (Snort, Suricata) for monitoring malicious network activity. Operating through an edge-first microservice architecture, it ensures low latency and resilience without reliance on cloud services. Our evaluation covered five scenarios—access control, unauthorized entry, weapon detection, port scanning, and denial-of-service attacks—with each repeated ten times under varied conditions such as low light, occlusion, and crowding. Results show face recognition accuracy of 92.7% (500 samples), weapon detection accuracy of 89.3% (450 samples), and intrusion detection latency below one second, with minimal false positives. Audio analysis of high-risk sounds further enhanced situational awareness. Beyond performance, ARGUS addresses GDPR and ISO 27001 compliance and anticipates adversarial robustness. By unifying cyber and physical detection, ARGUS advances beyond state-of-the-art patrol robots, delivering comprehensive situational awareness and a practical path toward resilient, ethical robotic security. Full article

Social media promotion of harmful products (e.g., combustible tobacco) poses a public health threat. However, strategies that amplify exposure to and engagement with such content remain understudied. This study aims to characterize strategies boosting cigar, little cigar, and cigarillo (CLCC) marketing visibility, referrals, and engagement on Instagram. Using keyword rules, we collected publicly available CLCC-related Instagram posts from CrowdTangle for a six-year period from August 2016 to October 2021. Posts were categorized as commercial (e.g., posts by tobacco brands or vendors) or organic and were coded for consumer engagement (CE) strategies (e.g., presence of prompts to like/share) using a combination of machine learning methods and human coding. Temporal engagement trends were analyzed using metadata. A total of 320,488 CLCC-related public posts were collected, with 44.6% (n = 142,875) identified as overtly commercial. Of these, 33.5% (n = 47,832) contained CE cues, including discounts and giveaways for tagging peers, liking, commenting, or following CLCC brands and spokesperson/influencers accounts, as well as calls to participate in contests and polls. Overtly commercial CE messages consistently garnered more comments per post and likes per post than non-CE commercial posts. There was a significant upward trend in the rate of comments on CE posts, suggesting growing effectiveness in eliciting user interaction. The proliferation of and high level of engagement with cigar-related promotional messages on Instagram demonstrate the need for public health surveillance and regulation of the evolving strategies promoting CLCC marketing exposure, reach, and engagement on social media. Full article

Ongoing changes in the distribution and abundance of several tick species of medical relevance in Canada have prompted the development of the eTick platform—an image-based crowd-sourcing public surveillance tool for Canada enabling rapid tick species identification by trained personnel, and public health guidance based on tick species and province of residence of the submitter. Considering that more than 100,000 images from over 73,500 identified records representing 25 tick species have been submitted to eTick since the public launch in 2018, a partial automation of the image processing workflow could save substantial human resources, especially as submission numbers have been steadily increasing since 2021. In this study, we evaluate an end-to-end artificial intelligence (AI) pipeline to support tick identification from eTick user-submitted images, characterized by heterogeneous quality and uncontrolled acquisition conditions. Our framework integrates (i) tick localization using a fine-tuned YOLOv7 object detection model, (ii) resolution enhancement of cropped images via super-resolution Generative Adversarial Networks (RealESRGAN and SwinIR), and (iii) image classification using deep convolutional (ResNet-50) and transformer-based (ViT) architectures across three datasets (12, 6, and 3 classes) of decreasing granularities in terms of taxonomic resolution, tick life stage, and specimen viewing angle. ViT consistently outperformed ResNet-50, especially in complex classification settings. The configuration yielding the best performance—relying on object detection without incorporating super-resolution—achieved a macro-averaged F1-score exceeding 86% in the 3-class model (Dermacentor sp., other species, bad images), with minimal critical misclassifications (0.7% of “other species” misclassified as Dermacentor). Given that Dermacentor ticks represent more than 60% of tick volume submitted on the eTick platform, the integration of a low granularity model in the processing workflow could save significant time while maintaining very high standards of identification accuracy. Our findings highlight the potential of combining modern AI methods to facilitate efficient and accurate tick image processing in community science platforms, while emphasizing the need to adapt model complexity and class resolution to task-specific constraints. Full article

Face recognition (FR) in unconstrained conditions remains an open research topic and an ongoing challenge. The facial images exhibit diverse expressions, occlusions, variations in illumination, and heterogeneous backgrounds. This work aims to produce an accurate and robust system for enhanced Security and Surveillance. A parallel hybrid deep learning model for feature extraction and classification is proposed. An ensemble of three parallel extraction layer models learns the best representative features using CNN and RNN. 2D LBP and 3D Mesh LBP are computed on face images to extract image features as input to two RNNs. A stacked autoencoder (SAE) merged the feature vectors extracted from the three CNN-RNN parallel layers. We tested the designed 2D/3D CNN-RNN framework on four standard datasets. We achieved an accuracy of $98.9\%$. The hybrid deep learning model significantly improves FR against similar state-of-the-art methods. The proposed model was also tested on an unconstrained conditions human crowd dataset, and the results were very promising with an accuracy of $95\%$. Furthermore, our model shows an 11.5% improvement over similar hybrid CNN-RNN architectures, proving its robustness in complex environments where the face can undergo different transformations. Full article

The Segment Anything Model 2 (SAM2) has achieved state-of-the-art performance in pixel-level object segmentation for both static and dynamic visual content. Its streaming memory architecture maintains spatial context across video sequences, yet struggles with long-term tracking due to its static inference framework. SAM 2’s fixed temporal window approach indiscriminately retains historical frames, failing to account for frame quality or dynamic motion patterns. This leads to error propagation and tracking instability in challenging scenarios involving fast-moving objects, partial occlusions, or crowded environments. To overcome these limitations, this paper proposes SAM2Plus, a zero-shot enhancement framework that integrates Kalman filter prediction, dynamic quality thresholds, and adaptive memory management. The Kalman filter models object motion using physical constraints to predict trajectories and dynamically refine segmentation states, mitigating positional drift during occlusions or velocity changes. Dynamic thresholds, combined with multi-criteria evaluation metrics (e.g., motion coherence, appearance consistency), prioritize high-quality frames while adaptively balancing confidence scores and temporal smoothness. This reduces ambiguities among similar objects in complex scenes. SAM2Plus further employs an optimized memory system that prunes outdated or low-confidence entries and retains temporally coherent context, ensuring constant computational resources even for infinitely long videos. Extensive experiments on two video object segmentation (VOS) benchmarks demonstrate SAM2Plus’s superiority over SAM 2. It achieves an average improvement of 1.0 in J&F metrics across all 24 direct comparisons, with gains exceeding 2.3 points on SA-V and LVOS datasets for long-term tracking. The method delivers real-time performance and strong generalization without fine-tuning or additional parameters, effectively addressing occlusion recovery and viewpoint changes. By unifying motion-aware physics-based prediction with spatial segmentation, SAM2Plus bridges the gap between static and dynamic reasoning, offering a scalable solution for real-world applications such as autonomous driving and surveillance systems. Full article

Two-dimensional human pose estimation (2D HPE) has become a fundamental task in computer vision, driven by growing demands in intelligent surveillance, sports analytics, and healthcare. The rapid advancement of deep learning has led to the development of numerous methods. However, the resulting diversity in research directions and model architectures has made systematic assessment and comparison difficult. This review presents a comprehensive overview of recent advances in 2D HPE, focusing on method classification, technical evolution, and performance evaluation. We classify mainstream approaches by task type (single-person vs. multi-person), output strategy (regression vs. heatmap), and architectural design (top-down vs. bottom-up) and analyze their respective strengths, limitations, and application scenarios. Additionally, we summarize commonly used evaluation metrics and benchmark datasets, such as MPII, COCO, LSP, OCHuman, and CrowdPose. A major contribution of this review is the detailed comparison of the top six models on each benchmark, highlighting their network architectures, input resolutions, evaluation results, and key innovations. In light of current challenges, we also outline future research directions, including model compression, occlusion handling, and cross-domain generalization. This review serves as a valuable reference for researchers seeking both foundational insights and practical guidance in 2D human pose estimation. Full article

In response to the decreased accuracy in person detection caused by densely populated areas and mutual occlusions in public spaces, a human head-detection approach is employed to assist in detecting individuals. To address key issues in dense scenes—such as poor feature extraction, rough label assignment, and inefficient pooling—we improved the YOLOv7 network in three aspects: adding attention mechanisms, enhancing the receptive field, and applying multi-scale feature fusion. First, a large amount of surveillance video data from crowded public spaces was collected to compile a head-detection dataset. Then, based on YOLOv7, the network was optimized as follows: (1) a CBAM attention module was added to the neck section; (2) a Gaussian receptive field-based label-assignment strategy was implemented at the junction between the original feature-fusion module and the detection head; (3) the SPPFCSPC module was used to replace the multi-space pyramid pooling. By seamlessly uniting CBAM, RFLAGauss, and SPPFCSPC, we establish a novel collaborative optimization framework. Finally, experimental comparisons revealed that the improved model’s accuracy increased from 92.4% to 94.4%; recall improved from 90.5% to 93.9%; and inference speed increased from 87.2 frames per second to 94.2 frames per second. Compared with single-stage object-detection models such as YOLOv7 and YOLOv8, the model demonstrated superior accuracy and inference speed. Its inference speed also significantly outperforms that of Faster R-CNN, Mask R-CNN, DINOv2, and RT-DETRv2, markedly enhancing both small-object (head) detection performance and efficiency. Full article

Tunnels are commonly found in small and enclosed environments on highways, roads, or city streets. They are constructed to pass through mountains or beneath crowded urban areas. To prevent accidents in these confined environments, lane changes, slow driving, or speeding are prohibited on single- or multi-lane one-way roads. We developed a foreground detection algorithm based on the K-nearest neighbor (KNN) and Gaussian mixture model and 400 collected images. The KNN was used to gather the first 200 image data, which were processed to remove differences and estimate a high-quality background. Once the background was obtained, new images were extracted without the background image to extract the vehicle’s foreground. The background image was processed using Canny edge detection and the Hough transform to calculate road lines. At the same time, the oriented FAST and rotated BRIEF (ORB) algorithm was employed to track vehicles in the foreground image and determine positions and lane deviations. This method enables the calculation of traffic flow and abnormal movements. We accelerated image processing using xfOpenCV on the PYNQ-Z2 and FPGA Xilinx platforms. The developed algorithm does not require pre-labeled training models and can be used during the daytime to automatically collect the required footage. For real-time monitoring, the proposed algorithm increases the computation speed ten times compared with YOLO-v2-tiny. Additionally, it uses less than 1% of YOLO’s storage space. The proposed algorithm operates stably on the PYNQ-Z2 platform with existing surveillance cameras, without additional hardware setup. These advantages make the system more appropriate for smart traffic management than the existing framework. Full article

Video violence detection has gained significant attention in recent years due to its applications in surveillance and security. This paper proposes a two-stage framework for detecting violent actions in video sequences. The first stage leverages GMFlow, a pre-trained optical flow network, to capture the temporal motion between consecutive frames, effectively encoding motion dynamics. In the second stage, we integrate these optical flow images with RGB frames and feed them into a CBAM-enhanced ResNet3D network to capture complementary spatiotemporal features. The attention mechanism provided by CBAM enables the network to focus on the most relevant regions in the frames, improving the detection of violent actions. We evaluate the proposed framework on three widely used datasets: Hockey Fight, Crowd Violence, and UBI-Fight. Our experimental results demonstrate superior performance compared to several state-of-the-art methods, achieving AUC scores of 0.963 on UBI-Fight and accuracies of 97.5% and 94.0% on Hockey Fight and Crowd Violence, respectively. The proposed approach effectively combines GMFlow-generated optical flow with deep 3D convolutional networks, providing robust and efficient detection of violence in videos. Full article

Tuberculosis (TB) is a global health challenge, particularly in the Central African Republic (CAR), which is classified as a high TB burden country. In the CAR, factors like poverty, limited healthcare access, high HIV prevalence, malnutrition, inadequate sanitation, low measles vaccination coverage, and conflict-driven crowded living conditions elevate TB risk. Improved AI-driven surveillance is hypothesized to address under-reporting and underdiagnosis. Therefore, we created an epidemiological digital representation of TB in Bangui by employing passive data collection, spatial analysis using a 100 × 100 m grid, and mapping TB treatment services. Our approach included estimating undiagnosed TB cases through the integration of TB incidence, notification rates, and diagnostic data. High-resolution predictions are achieved by subdividing the area into smaller units while considering influencing variables within the Bayesian model. By designating moderate and high-risk hotspots, the model highlighted the potential for precise resource allocation in TB control. The strength of our model lies in its adaptability to overcome challenges, although this may have been to the detriment of precision in some areas. Research is envisioned to evaluate the model’s accuracy, and future research should consider exploring the integration of multidrug-resistant TB within the model. Full article

Recent transportation systems are operated by cooperative factors including mobile robots, smart vehicles, and intelligent management. It is highly anticipated that the surveillance using mobile robots can be utilized in complex transportation areas where the high accuracy is required. In this paper, we introduce a crowd surveillance system using mobile robots and intelligent vehicles to provide obstacle avoidance in transportation stations with a consideration of different moving strategies of the robots in an existing 2D area supported by line-based barriers and surveillance formations. Then, we formally define a problem that aims to minimize the distance traveled by a mobile robot, while also considering the speed of the mobile robot and avoiding the risk of collisions when the mobile robot moves to specific locations to fulfill crowd surveillance. To solve this problem, we propose two different schemes to provide improved surveillance that can be used even when considering speed. After that, various ideas are gathered to define conditions, set various settings, and modify them to evaluate their performances. Full article

For applications such as home surveillance systems and assisted living for elderly care, sensing capabilities are essential for tasks such as locating, determining the approximate position of a person, or identifying the status of a person (static or moving), since the effects caused by the presence of people can be captured in the power received by signals in an infrastructure deployed for these purposes. Human interference in Received Signal Strength Indicator (RSSI) measurements between different pairs of wireless nodes can vary depending on whether the target is moving or static. To test these ideas, an experiment was conducted using four nodes equipped with the ZigBee protocol in each corner of an empty 6.9 m × 8.1 m × 3.05 m room. These nodes were configured as routers, communicating with a coordinator outside the room that instructed the nodes to send back their pairwise RSSI measurements. The coordinator was connected to a computer in order to log the measurements, as well as the time at which the measurements were generated. The code was run for every iteration of the experiment, whether the target was static, moving, or when the number of targets was increased to five. The data were then statistically analyzed to extract patterns and other target relational parameters. There was a correlation between the change in the pairwise RSSI and the path described by the target when moving through the room. The data presented by the results can aid algorithms for device-free localization and crowd classification, with a low infrastructure cost for both, and shed light on the relevant characteristics correlated with the path and crowd size in indoor settings. Full article

With the increasing need for real-time crowd evaluation in military surveillance, public safety, and event crowd management, crowd counting using unmanned aerial vehicle (UAV) captured images has emerged as an essential research topic. While conventional RGB-based methods have achieved significant success, their performance is severely hampered in low-light environments due to poor visibility. Integrating thermal infrared (TIR) images can address this issue, but existing RGB-T crowd counting networks, which employ multi-stream architectures, tend to introduce computational redundancy and excessive parameters, rendering them impractical for UAV applications constrained by limited onboard resources. To overcome these challenges, this research introduces an innovative, compact RGB-T framework designed to minimize redundant feature processing and improve multi-modal representation. The proposed approach introduces a Partial Information Interaction Convolution (PIIConv) module to selectively minimize redundant feature computations and a Global Collaborative Fusion (GCFusion) module to improve multi-modal feature representation through spatial attention mechanisms. Empirical findings indicate that the introduced network attains competitive results on the DroneRGBT dataset while significantly reducing floating-point operations (FLOPs) and improving inference speed across various computing platforms. This study’s significance is in providing a computationally efficient framework for RGB-T crowd counting that balances accuracy and resource efficiency, making it ideal for real-time UAV deployment. Full article

Bird strikes are a substantial aviation safety issue that can result in serious harm to aircraft components and even passenger deaths. In response to this increased tendency, the implementation of new and more efficient detection and prevention technologies becomes urgent. The paper presents a novel deep learning model which is developed to detect and alleviate bird strike issues in airport conditions boosting aircraft safety. Based on an extensive database of bird images having different species and flight patterns, the research adopts sophisticated image augmentation techniques which generate multiple scenarios of aircraft operation ensuring that the model is robust under different conditions. The methodology evolved around the building of a spatiotemporal convolutional neural network which employs spatial attention structures together with dynamic temporal processing to precisely recognize flying birds. One of the most important features of this research is the architecture of its dual-focus model which consists of two components, the attention-based temporal analysis network and the convolutional neural network with spatial awareness. The model’s architecture can identify specific features nested in a crowded and shifting backdrop, thereby lowering false positives and improving detection accuracy. The mechanisms of attention of this model itself enhance the model’s focus by identifying vital features of bird flight patterns that are crucial. The results are that the proposed model achieves better performance in terms of accuracy and real time responses than the existing bird detection systems. The ablation study demonstrates the indispensable roles of each component, confirming their synergistic effect on improving detection performance. The research substantiates the model’s applicability as a part of airport bird strike surveillance system, providing an alternative to the prevention strategy. This work benefits from the unique deep learning feature application, which leads to a large-scale and reliable tool for dealing with the bird strike problem. Full article

Traffic congestion is a prevalent problem in modern civilizations worldwide, affecting both large cities and smaller communities. Emergency vehicles tend to group tightly together in these crowded scenarios, often masking one another. For traffic surveillance systems tasked with maintaining order and executing laws, this poses serious difficulties. Recent developments in machine learning for image processing have significantly increased the accuracy and effectiveness of emergency vehicle classification (EVC) systems, especially when combined with specialized hardware accelerators. The widespread use of these technologies in safety and traffic management applications has led to more sustainable transportation infrastructure management. Vehicle classification has traditionally been carried out manually by specialists, which is a laborious and subjective procedure that depends largely on the expertise that is available. Furthermore, erroneous EVC might result in major problems with operation, highlighting the necessity for a more dependable, precise, and effective method of classifying vehicles. Although image processing for EVC involves a variety of machine learning techniques, the process is still labor intensive and time consuming because the techniques now in use frequently fail to appropriately capture each type of vehicle. In order to improve the sustainability of transportation infrastructure management, this article places a strong emphasis on the creation of a hardware system that is reliable and accurate for identifying emergency vehicles in intricate contexts. The ResNet50 model’s features are extracted by the suggested system utilizing a Field Programmable Gate Array (FPGA) and then optimized by a multi-objective genetic algorithm (MOGA). A CatBoost (CB) classifier is used to categorize automobiles based on these features. Overtaking the previous state-of-the-art accuracy of 98%, the ResNet50-MOP-CB network achieved a classification accuracy of 99.87% for four primary categories of emergency vehicles. In tests conducted on tablets, laptops, and smartphones, it demonstrated excellent accuracy, fast classification times, and robustness for real-world applications. On average, it took 0.9 nanoseconds for every image to be classified with a 96.65% accuracy rate. Full article