Search Results (17)

Risk-based prioritization frameworks such as the Intelligent Pedestrian Model (IPM) rank pedestrians by an instantaneous, reference-normalized risk score. They indicate which pedestrian requires attention first. This study examines the temporal stability of such rankings. We computed an observation-only kinematic Exposure proxy frame by frame on three ETH/UCY benchmark scenes. In these scenes, the highest-priority identity changes rapidly: the median top-1 persistence is two frames (0.8 s). We introduce a switch classification that separates established switches from entry-driven and forced switches. Grace-period exclusion is evaluated as a sensitivity variant and shown to remove up to 82% of evaluable time in short-track scenes. Established flicker rates range from one switch per 2.7 s in dense scenes to one per 25.4 s in sparse scenes, with established switches concentrated at small score differences between competing pedestrians. The results show that instantaneous rankings alone may be insufficient for sustained attention allocation and motivate future work on temporal priority management. Full article

This paper presents a real-time pedestrian situation detection framework for autonomous mobility platforms. The proposed approach extracts pedestrians from images acquired by a camera mounted on an autonomous mobility system, classifies their postures, tracks their trajectories, and subsequently detects pedestrian situations. A convolutional neural network (CNN) is employed for pedestrian detection and posture classification, where the YOLOv12 model is fine-tuned via transfer learning for this purpose. To improve detection and classification performance, a region of interest (ROI) is defined using camera calibration data, enabling robust detection of small-scale pedestrians over long distances. Using a custom-labeled dataset, the proposed method achieves a precision of 96.6% and a recall of 97.0% for pedestrian detection and posture classification. The detected pedestrians are tracked using the DeepSORT algorithm, and their situations are inferred through a rule-based analysis module. Experimental results demonstrate that the proposed system operates at an execution speed of 58.11 ms per frame, corresponding to 17.2 fps, thereby satisfying the real-time requirements for autonomous mobility applications. These results confirm that the proposed framework enables reliable real-time pedestrian extraction and situation awareness in real-world autonomous mobility environments. Full article

To address the limitations of conventional social force models in simulating high-density pedestrian crowds, this study proposes an enhanced model that incorporates visual perception constraints, group-type labeling, and collective avoidance mechanisms. Pedestrian trajectories were extracted from a bidirectional commercial street scenario using OpenCV, with YOLOv8 and DeepSORT employed for multiple object tracking. Analysis of pedestrian grouping patterns revealed that 52% of pedestrians walked in pairs, with distinct avoidance behaviors observed. The improved model integrates three key mechanisms: a restricted 120° forward visual field, group-type classification based on social relationships, and an exponentially formulated inter-group repulsive force. Simulation results in MATLAB R2023b demonstrate that the proposed model outperforms conventional approaches in multiple aspects: speed distribution (error < 8%); spatial density overlap (>85%); trajectory similarity (reduction of 32% in Dynamic Time Warping distance); and avoidance behavior accuracy (82% simulated vs. 85% measured). This model serves as a quantitative simulation tool and decision-making basis for the planning of pedestrian spaces, crowd organization management, and the optimization of emergency evacuation schemes in high-density pedestrian areas such as commercial streets and subway stations. Consequently, it contributes to enhancing pedestrian mobility efficiency and public safety, thereby supporting the development of a sustainable urban slow transportation system. Full article

Pedestrian safety at signalized intersections remains a pressing concern in rapidly urbanizing cities. This study introduces a trajectory–signal behavior spectrum, grounded in Behavior Spectrum Theory (BST), to quantify crossing risk using readily observable data. Unmanned aerial vehicle (UAV) video is employed to record pedestrian movements, which are then detected with YOLOv8 and tracked with ByteTrack, producing frame-level trajectories without dependence on line-of-sight instrumentation. Five spatiotemporal features—speed, acceleration, crossing time, remaining pedestrian-signal green time, and red-phase duration—are compiled into the spectrum. Features are normalized using the interquartile range (IQR) method, and objective weights are determined with an improved CRITIC (Criteria Importance Through Intercriteria Correlation) scheme that incorporates a median-based coefficient of variation and absolute correlation for conflict measurement. The resulting risk eigenvalues are clustered with K-means into four levels: no risk, low, medium, and high. A case study of 1210 crossings at a two-way eight-lane intersection in Harbin, China (576 compliant, 634 non-compliant) demonstrates the approach. Results show greater variability among non-compliant speeds (mean 1.29 m/s) compared with compliant crossings (mean 1.40 m/s), with more extreme deviations. Clustering achieved silhouette coefficients of 0.60 for compliant and 0.69 for non-compliant groups, while expert validation on 20 samples yielded substantial agreement (Fleiss’ Kappa = 0.87). This study provides a systematic and interpretable method for risk classification, which supports both theoretical understanding and applied traffic safety management. Full article

To address the challenges of real-time monitoring via highway vehicle-mounted cameras—specifically, the difficulty in detecting distant pedestrians and vehicles in real time—this study proposes an enhanced object detection algorithm, YOLOv5s-F. Firstly, the FasterNet network structure is adopted to improve the model’s runtime speed. Secondly, the attention mechanism BRA, which is derived from the Transformer algorithm, and a 160 × 160 small-object detection layer are introduced to enhance small target detection performance. Thirdly, the improved upsampling operator CARAFE is incorporated to boost the localization and classification accuracy of small objects. Finally, Focal EIoU is employed as the localization loss function to accelerate model training convergence. Quantitative experiments on high-speed sequences show that Focal EIoU reduces bounding box jitter by 42.9% and improves tracking stability (consecutive frame overlap) by 11.4% compared to CIoU, while accelerating convergence by 17.6%. Results show that compared with the YOLOv5s baseline network, the proposed algorithm reduces computational complexity and parameter count by 10.1% and 24.6%, respectively, while increasing detection speed and accuracy by 15.4% and 2.1%. Transfer learning experiments on the VisDrone2019 and Highway-100k dataset demonstrate that the algorithm outperforms YOLOv5s in average precision across all target categories. On NVIDIA Jetson Xavier NX, YOLOv5s-F achieves 32 FPS after quantization, meeting the real-time requirements of in-vehicle monitoring. The YOLOv5s-F algorithm not only meets the real-time detection and accuracy requirements for small objects but also exhibits strong generalization capabilities. This study clarifies core challenges in highway small-target detection and achieves accuracy–speed improvements via three key innovations, with all experiments being reproducible. If any researchers need the code and dataset of this study, they can consult the author through email. Full article

DEXTER (detection of explosives and firearms to counter terrorism) is a project funded by NATO’s Science for Peace and Security (SPS) program with the goal of developing an integrated system capable of remotely and accurately detecting explosives and firearms in public places without impeding the flow of pedestrians. While body scanner systems in secure areas of public places are becoming more and more efficient, the attack at Brussels airport on 22 March 2016, upstream of these systems, in the middle of the crowd of passengers, demonstrated the lack of discreet and real-time security against threats of mass terrorism. The NATO-SPS international and multi-year DEXTER project aims to provide new technical and strategic solutions to fill this gap. This project is based on multi-sensor coordination and fusion, from hyperspectral remote laser to smart glasses, artificial algorithms, and suspect identification and tracking. One of these sensors is dedicated to threat detection (large weapon or explosive belt) using the clothing of pedestrians by means of an active microwave component. This project is referred to as MIC (Microwave Imaging Curtain), also supported by the French SGDSN (General Secretariat of Defense and National Security), and utilizes a radar system capable of generating 3D images in real-time to address non-checkpoint detection of explosives and firearms. The project, led by ONERA (France), is based on a radar imaging system developed by the Fraunhofer FHR institute, using a MIMO architecture with an Ultra-Wide Band waveform. Although high-resolution 3D microwave imaging is already being used in expensive body scanners to detect firearms concealed under clothing, MIC’s innovative approach lies in utilizing a high-resolution 3D imaging device that can detect larger dangerous objects carried by moving individuals at a longer range, in addition to providing discrete detection in pedestrian flow. Automatic detection and classification of these dangerous objects is carried out on 3D radar images using a deep-learning network. This paper will outline the project’s objectives and constraints, as well as the design, architecture, and performance of the final system. Additionally, it will present real-time imaging results obtained during a live demonstration in a relevant environment. Full article

With the advancement of science and technology, the development and application of unmanned mobile vehicles (UMVs) have emerged as topics of crucial concern in the global industry. The development goals and directions of UMVs vary according to their industrial uses, which include navigation, autonomous driving, and environmental recognition; these uses have become the priority development goals of researchers in various fields. UMVs employ sensors to collect environmental data for environmental analysis and path planning. However, the analysis function of a single sensor is generally affected by natural environmental factors, resulting in poor identification results. Therefore, this study introduces fusion technology that employs heterogeneous sensors in the Ackerman UMV, leveraging the advantages of each sensor to enhance accuracy and stability in environmental detection and identification. This study proposes a fusion technique involving heterogeneous imaging and LiDAR (laser imaging, detection, and ranging) sensors in an Ackerman UMV. A camera is used to obtain real-time images, and YOLOv4-tiny and simple online real-time tracking are then employed to detect the location of objects and conduct object classification and object tracking. LiDAR is simultaneously used to obtain real-time distance information of detected objects. An inertial measurement unit is used to gather odometry information to determine the position of the Ackerman UMV. Static maps are created using simultaneous localization and mapping. When the user commands the Ackerman UMV to move to the target point, the vehicle control center composed of the robot operating system activates the navigation function through the navigation control module. The Ackerman UMV can reach the destination and instantly identify obstacles and pedestrians when in motion. Full article

The advances in positioning techniques have generated massive trajectory data that represent the mobility of objects, e.g., pedestrians and mobile phones. It is important to integrate information from various modalities for subject tracking or trajectory prediction. Our work attempts to match a face with a corresponding mobile phone based on the heterogeneous trajectories. We propose a framework which associates face trajectories with their corresponding mobile phone trajectories using elaborate and explainable features. Our solution includes two stages: an initial selection of phone trajectories for a given face trajectory and a subsequent identification of which phone trajectory provides an exact match to the given face trajectory. In the first stage, we propose a Multi-Granularity SpatioTemporal Window Searching (MGSTWS) algorithm to select candidate mobile phones that are spatiotemporally close to a given face. In the second stage, we first build an affinity function to score face–phone trajectory point pairs selected by MGSTWS, and construct a feature set for building a face–phone trajectory matching determinator which determines whether a phone trajectory matches a given face trajectory. Our well-designed features guarantee high model simplicity and interpretability. Among the feature set, BGST intelligently leverages disassociation between a face and a mobile phone even if there exists some co-occurence for a non-matching face–phone pair. Based on the feature set, we represent the face–phone matching task as a binary classification problem and train various models, among which LightGBM achieves the best performance with 92.6% accuracy, 96.9% precision, 88.5% recall, and 92.5% F1. Our framework is acceptable in most application scenarios and may benefit some downstream tasks. The preselection-refining architecture of our framework guarantees the applicability and efficiency of the face–phone trajectory pair matching frame. Full article

An intelligent, accurate, and powerful object detection system is required for automated driving systems to keep these vehicles aware of their surrounding objects. Thus, vehicles adapt their speed and operations to avoid crashing with the existing objects and follow the driving rules around the existence of emergency vehicles and installed traffic signs. The objects considered in this work are summarized by regular vehicles, big trucks, emergency vehicles, pedestrians, bicycles, traffic lights, and traffic signs on the roadside. Autonomous vehicles are equipped with high-quality sensors and cameras, LiDAR, radars, and GPS tracking systems that help to detect existing objects, identify them, and determine their exact locations. However, these tools are costly and require regular maintenance. This work aims to develop an intelligent object classification mechanism for autonomous vehicles. The proposed mechanism uses machine learning technology to predict the existence of investigated objects over the road network early. We use different datasets to evaluate the performance of the proposed mechanism. Accuracy, Precision, F1-Score, G-Mean, and Recall are the measures considered in the experiments. Moreover, the proposed object classification mechanism is compared to other selected previous techniques in this field. The results show that grouping the dataset based on their mobility nature before applying the classification task improved the results for most of the algorithms, especially for vehicle detection. Full article

One of the primary tasks undertaken by autonomous vehicles (AVs) is object detection, which comes ahead of object tracking, trajectory estimation, and collision avoidance. Vulnerable road objects (e.g., pedestrians, cyclists, etc.) pose a greater challenge to the reliability of object detection operations due to their continuously changing behavior. The majority of commercially available AVs, and research into them, depends on employing expensive sensors. However, this hinders the development of further research on the operations of AVs. In this paper, therefore, we focus on the use of a lower-cost single-beam LiDAR in addition to a monocular camera to achieve multiple 3D vulnerable object detection in real driving scenarios, all the while maintaining real-time performance. This research also addresses the problems faced during object detection, such as the complex interaction between objects where occlusion and truncation occur, and the dynamic changes in the perspective and scale of bounding boxes. The video-processing module works upon a deep-learning detector (YOLOv3), while the LiDAR measurements are pre-processed and grouped into clusters. The output of the proposed system is objects classification and localization by having bounding boxes accompanied by a third depth dimension acquired by the LiDAR. Real-time tests show that the system can efficiently detect the 3D location of vulnerable objects in real-time scenarios. Full article

People detection in images has many uses today, ranging from face detection algorithms used by social networks to help the users tag other people, to surveillance systems that can create a statistic of the population density in an area, or identify a suspect, or even in the automotive industry as part of the Pedestrian Crash Avoidance Mitigation (PCAM) system. This work focuses on creating a fast and reliable object detection algorithm that will be trained on scenes that depict people in an indoor environment, starting from an existing state-of-the-art approach. The proposed method improves upon the You Only Look Once version 4 (YOLOv4) network by adding a region of interest classification and regression branch such as Faster R-CNN’s head. The candidate bounding boxes proposed by YOLOv4 are ranked based on their confidence score, the best candidates being kept and sent as input to the Faster Region-Based Convolutional Neural Network (R-CNN) head. To keep only the best detections, non-maximum suppression is applied to all proposals. This decreases the number of false-positive candidate bounding boxes, the low-confidence detections of the regression and classification branch being eliminated by the detections of YOLOv4 and vice versa in the non-maximum suppression step. This method can be used as the object detection algorithm in an image-based people tracking system, namely Tracktor, having a higher inference speed than Faster R-CNN. Our proposed method manages to achieve an overall accuracy of 95% and an inference time of 22 ms. Full article

One of the major challenges for blind and visually impaired (BVI) people is traveling safely to cross intersections on foot. Many countries are now generating audible signals at crossings for visually impaired people to help with this problem. However, these accessible pedestrian signals can result in confusion for visually impaired people as they do not know which signal must be interpreted for traveling multiple crosses in complex road architecture. To solve this problem, we propose an assistive system called CAS (Crossing Assistance System) which extends the principle of the BLE (Bluetooth Low Energy) RSSI (Received Signal Strength Indicator) signal for outdoor and indoor location tracking and overcomes the intrinsic limitation of outdoor noise to enable us to locate the user effectively. We installed the system on a real-world intersection and collected a set of data for demonstrating the feasibility of outdoor RSSI tracking in a series of two studies. In the first study, our goal was to show the feasibility of using outdoor RSSI on the localization of four zones. We used a k-nearest neighbors (kNN) method and showed it led to 99.8% accuracy. In the second study, we extended our work to a more complex setup with nine zones, evaluated both the kNN and an additional method, a Support Vector Machine (SVM) with various RSSI features for classification. We found that the SVM performed best using the RSSI average, standard deviation, median, interquartile range (IQR) of the RSSI over a 5 s window. The best method can localize people with 97.7% accuracy. We conclude this paper by discussing how our system can impact navigation for BVI users in outdoor and indoor setups and what are the implications of these findings on the design of both wearable and traffic assistive technology for blind pedestrian navigation. Full article

Pedestrian detection is a critical perception task for autonomous driving and intelligent vehicle, and it is challenging due to the potential variation of appearance and pose of human beings as well as the partial occlusion. In this paper, we present a novel pedestrian detection method via four-layer laser scanner. The proposed approach deals with the occlusion problem by fusing the segment classification results with past knowledge integration from tracking process. First, raw point cloud is segmented into the clusters of independent objects. Then, three types of features are proposed to capture the comprehensive cues, and 18 effective features are extracted with the combination of the univariate feature selection algorithm and feature correlation analysis process. Next, based on the segment classification at individual frame, the track classification is conducted further for consecutive frames using particle filter and probability data association filter. Experimental results demonstrate that both back-propagation neural network and Adaboost classifiers based on 18 selected features have their own advantages at the segment classification stage in terms of pedestrian detection performance and computation time, and the track classification procedure can improve the detection performance particularly for partially occluded pedestrians in comparison with the single segment classification procedure. Full article

Loitering detection can help recognize vulnerable people needing attention and potential suspects harmful to public security. The existing loitering detection methods used time or target trajectories as assessment criteria, and only handled some simple loitering circumstances because of complex track. To solve these problems, this paper proposes a loitering detection method based on pedestrian activity area classification. The paper first gave loitering definition from a new perspective using the size of pedestrian activity area. The pedestrian loitering behaviors were divided into three categories. The proposed algorithms dynamically calculate enclosing rectangle, ellipse, and sector of pedestrian activity areas through curve fitting based on trajectory coordinates within given staying time threshold. The loitering is recognized if the pedestrian activity is detected to be constrained in an area within a certain period of time. The algorithm does not need to calculate complex trajectories. The PETS2007 dataset and our own self-collected simulated test videos were tested. The experimental results show that the proposed method accurately detected the pedestrian loitering, not only detected some loitering that the existing methods could not detect, but also distinguishing different types of loitering. The proposed method is effectiveness, robust, and simplified in implementation. Full article

The emergence of new wearable technologies, such as action cameras and smart glasses, has driven the use of the first-person perspective in computer applications. This field is now attracting the attention and investment of researchers aiming to develop methods to process first-person vision (FPV) video. The current approaches present particular combinations of different image features and quantitative methods to accomplish specific objectives, such as object detection, activity recognition, user–machine interaction, etc. FPV-based navigation is necessary in some special areas, where Global Position System (GPS) or other radio-wave strength methods are blocked, and is especially helpful for visually impaired people. In this paper, we propose a hybrid structure with a convolutional neural network (CNN) and local image features to achieve FPV pedestrian navigation. A novel end-to-end trainable global pooling operator, called AlphaMEX, has been designed to improve the scene classification accuracy of CNNs. A scale-invariant feature transform (SIFT)-based tracking algorithm is employed for movement estimation and trajectory tracking of the person through each frame of FPV images. Experimental results demonstrate the effectiveness of the proposed method. The top-1 error rate of the proposed AlphaMEX-ResNet outperforms the original ResNet (k = 12) by 1.7% on the ImageNet dataset. The CNN-SIFT hybrid pedestrian navigation system reaches 0.57 m average absolute error, which is an adequate accuracy for pedestrian navigation. Both positions and movements can be well estimated by the proposed pedestrian navigation algorithm with a single wearable camera. Full article