Detection and classification of dangerous maneuvers based on traffic camera recordings

A new method is proposed for analyzing the nonlinear dynamics and stability in lane changes on highways for tractor-semitrailer under rainy weather. Unlike most of the literature associated with a simulated linear dynamic model for tractor-semitrailers steady steering on dry road, a verified 5DOF mechanical model with nonlinear tire based on vehicle test was used in the lane change simulation on low adhesion coefficient road. According to Jacobian matrix eigenvalues of the vehicle model, bifurcations of steady steering and sinusoidal steering on highways under rainy weather were investigated using a numerical method. Furthermore, based on feedback linearization theory, taking the tractor yaw rate and joint angle as control objects, a feedback linearization controller combined with AFS and DYC was established. The numerical simulation results reveal that Hopf bifurcations are identified in steady and sinusoidal steering conditions, which translate into an oscillatory behavior leading to instability. And simulations of urgent step and single-lane change in high velocity show that the designed controller has good effects on eliminating bifurcations and improving lateral stability of tractor-semitrailer, during lane changing on highway under rainy weather. It is a valuable reference for safety design of tractor-semitrailers to improve the traffic safety with driver-vehicle-road closed-loop system.

<div class="section abstract"><div class="htmlview paragraph">The recent progress in camera-based technologies has prompted the development of prototype camera-based video systems, intended to replace conventional passenger vehicle mirrors. Given that a significant number of collisions during lane changes stem from drivers being unaware of nearby vehicles, these camera-based systems offer the potential to enhance safety. By affording drivers a broader field of view, they facilitate the detection of potential conflicts. This project was focused on analyzing naturalistic driving data in support of the Federal Motor Vehicle Safety Standard 111 regulatory endeavors. The goal was to assess the effectiveness and safety compatibility of prototype camera-based side-view systems as potential replacements for traditional side-view mirrors. The method employed involved extracting radar data from instances of lane changes conducted by 12 drivers for two pick-up trucks includes 10018 signal-indicated lane changes performed at speeds consistent with highway travel in a naturalistic driving study. Among these, 5029 lane changes were carried out using camera-based displays. The study delved into the influence of camera-based systems on lane change behavior, particularly examining lane change conflicts using time-to-collision data obtained from radar readings. The objective was to assess their impact on safety measures, such as vehicle detection, distance, closing speeds, and conflicts. These assessments spanned a range of conditions, including both daylight and nighttime operations. Upon analyzing the results, no compelling evidence emerged to indicate that the adoption of camera-based systems had a detrimental effect on lane change performance, leading to riskier lane changes in comparison to the traditional mirror systems.</div></div>

Advanced driver assistance systems are required to detect latent hazards posed by surrounding vehicles and generate an appropriate response to enhance safety. Lane changes constitute potentially risky maneuvers, as drivers involved encounter latent hazards due to surrounding vehicles. A careful study of lane change behavior is therefore essential in identifying potential abnormalities that may lead to various hazards, during the process of a lane change. In this study, an anomaly detection technique is used to compare snapshots of normal and dangerous lane change maneuvers, to identify the abnormal instances. A one-class support vector machine is used and tested for novelty identification of naturalistic driving study data. The results show that the technique is able to detect dangerous lane changes with high accuracy. In addition, results suggest that dangerous behavior could occur before, after or during a lane change maneuver.

More than 90% of traffic accidents are caused by driver behavior, with lane change behavior being a major contributor. Recently, driving assistance systems are being introduced on vehicles to reduce traffic accidents, and a reliable vehicle lane change collision detection system is a key component of these systems. Besides, the foundation of the vehicle lane change detection system is the effective vehicle lane change detection model. In this paper, based on the support vector machine, we propose a model for detecting driver lane change maneuvers and take into account the real-time vehicle dynamic features transmitted via Vehicle to X (V2X) Communication. The accuracy is ideal for lane keep and lane change situations, and it is also robust for zigzag driving situations, according to tests conducted using the NGSIM real traffic dataset. The detection accuracy for left and right lane change maneuvers is 97.5% and 99.09%, respectively, while the false alarm rate is 8.56%. Additionally, the average advance detection time is 1.7 s, which is suitable for actual driving application scenarios.

Detection of lane change maneuvers of other traffic participants is one key element looking at highly automated driving and future driver assistance systems. A wide number of publications have shown approaches exploiting solely high level features from processed sensor data. Looking at low level features, vision based systems offer high potential for the tasks of lane change detection in the image domain, like the positioning of objects relative to the lane markings or the evaluation of their movement towards adjacent lanes. Among conventional image processing algorithms deployed, optical flow is an appropriate approach towards the latter task, that is affected by translation and rotation of the ego vehicle. In this paper we introduce true flow as an addition to conventional optical flow. Translational and rotational influence of ego motion is eliminated, allowing to estimate solely the observed object behavior. Features for lane change detection are presented and their invariance is shown under variable conditions, using simulation data. Experimental evaluation shows feature performance and predictability based on a real world data test set.

<div class="htmlview paragraph">This paper describes a forward looking on-board vehicle detection and driver alert system that provides a distance indication and alert tone to the driver. The system, called VORAD (Vehicular On-board Radar), is the first of its kind to be fielded. The system can be programmed to function in different operating modes, allowing customization to the user's requirements. Some possible operating modes include simply providing an alert to the driver, providing following distance indication measured in seconds based on the vehicles' speeds, or providing following distance measured in feet. The VORAD System also has optional features to enhance its usefulness, including a blind spot alert system and a built-in event recorder. The blind spot system provides additional information to the driver regarding the presence of vehicles in his blind spot for use in making lane changes. The event recorder can act as a “flight recorder” for accident reconstruction, safety training and fleet management activities.</div> <div class="htmlview paragraph">The current version of the system does not interface with any vehicle controls. New applications under development include advanced control systems such as Intelligent Cruise Control. In addition, future systems can be designed to provide vehicle to vehicle and vehicle to roadside communication functions.</div>

This research focuses on the development of a real-time vehicle detection and tracking system using OpenCV and deep learning models to address growing traffic management challenges. The project integrates the Gaussian Mixture-based Background Subtraction (MOG2), morphological operations, contour detection, and tracking methods like Kalman Filtering and Deep SORT. Additionally, object classification via YOLO and MobileNet-SSD significantly improves the accuracy of identifying and distinguishing vehicles. Evaluated on multiple real-world datasets under various environmental conditions, the system demonstrated strong performance with over 91% mAP, 92.5% precision, and robust multi-object tracking accuracy. This study serves as a scalable framework for intelligent traffic surveillance.

Objective Autonomous Emergency Braking (AEB) is a promising technology for crash avoidance or pre-crash impact speed reduction through the automatic application of braking force. Implementation of AEB technology on motorcycles (MAEB) is still problematic as its interaction with the rider may compromise the safety. In previous studies, MAEB interventions at low decelerations were shown to be easily manageable by common riders in straight line condition, but they were not previously tested in lateral maneuvers such as lane change and swerving, which are common in pre-crash situations. The objective of this paper is to assess the applicability of MAEB activation during lateral avoidance maneuver and to estimate its benefits in this scenario. Methods Field tests were carried out involving common riders as participants, using a test protocol developed on the experience of previous studies. The test vehicle was a sport-touring motorcycle equipped with an automatic braking system that could be activated remotely by researchers to simulate MAEB intervention. The motorcycle was equipped with outriggers to prevent capsizing. The Automatic Braking (AB) interventions using a nominal deceleration of 0.3 g were deployed at pseudo-random times in conditions of straight-line travel and a sharp lane-change maneuver emulating a pre-crash avoidance action. The straight-line trials were used as the reference condition for analysis. Results Thirty-one participants experienced AB interventions in straight-line and lane-change at an average speed of 44.5 km/h. The automatic braking was deployed in all the key phases of the avoidance maneuver. The system reached a deceleration of 0.3 g for a time of intervention of approximately 1 s. The participants were consistently able to control the vehicle during the automatic braking interventions and were always able to complete the lane-change maneuver. The speed reductions obtained with the AB interventions during lane change were very similar to those obtained in the straight-line conditions. Conclusions MAEB interventions with decelerations up to 0.3 g can be easily managed by motorcycle riders not only in straight-line conditions but also during an avoidance maneuver. Further investigations using higher deceleration values are now possible.

Traffic Flow Analysis System is an important part of Intelligent Transportation Systems. It mainly contains two parts: vehicle detection part and vehicle tracking part. The efficient video object segmentation algorithm is the basis of vehicle detection part that aims to segment each vehicle correctly. In this paper, according to the street traffic environment, we adopt a multiple video object segmentation algorithm in vehicle detection. The algorithm consists of training section and segmentation section. At the training section, it uses codebook method to train the vehicle images: First, it uses Scale-invariant feature transform method to extract the features of vehicle image samples. Secondly, it segments the images into small patches to cluster. Third, it uses a new way to activate the codebook that can largely reduce the numbers of vehicle sample images. At the segmentation section, it uses an Implicit Shape Model to combine the recognition knowledge and the segmentation knowledge together. At last, we use this algorithm to implement a prototype of traffic flow analysis system, and achieve the expected results.

This project investigates the potential of 5G Edge computing to improve vehicular safety through ultra-low latency communication and real-time data processing. With human error accounting for the majority of traffic accidents, our system leverages the proximity of edge servers to 5G base stations to transmit critical vehicle telemetry between nearby cars using MQTT protocols and Docker-based deployment. We designed and tested two key safety scenarios—following vehicle and head-on collision detection—demonstrating reliable, low-latency performance under simulated conditions using OBD-II emulators. Our work also includes a documented framework for scalable deployment and real-time data streaming. Looking forward, we aim to integrate machine learning for predictive maneuvers such as lane changes and environmental adaptation, enhancing the system's robustness in complex traffic environments.

The rapid advancement in the field of deep learning and high performance computing has highly augmented the scope of video-based vehicle counting system. In this paper, the authors deploy several state-of-the-art object detection and tracking algorithms to detect and track different classes of vehicles in their regions of interest (ROI). The goal of correctly detecting and tracking vehicles’ in their ROI is to obtain an accurate vehicle count. Multiple combinations of object detection models coupled with different tracking systems are applied to access the best vehicle counting framework. The models’ addresses challenges associated to different weather conditions, occlusion and low-light settings and efficiently extracts vehicle information and trajectories through its computationally rich training and feedback cycles. The automatic vehicle counts resulting from all the model combinations are validated and compared against the manually counted ground truths of over 9 h’ traffic video data obtained from the Louisiana Department of Transportation and Development. Experimental results demonstrate that the combination of CenterNet and Deep SORT, and YOLOv4 and Deep SORT produced the best overall counting percentage for all vehicles.

In this paper, we use the deep sort multi-target tracking algorithm to achieve multi-target tracking of pedestrians and vehicles in traffic scenes. In this paper, firstly, yolov4 is used to train the pedestrian and vehicle detection model in traffic scenes. Then, according to the detection frame predicted by yolov4, multi-target tracking is carried out for specific targets. The multi-target tracking algorithm uses deep Sort, which can be combined with yolov4, can achieve less ID switching in real-time reasoning and deal with the loss of occlusion, so as to achieve more stable tracking effect.

It is dangerous that changing lane without knowing the information of the other lane in the blind-spot area. We propose a vision based lane changing assistance system to monitor the vehicle in the blind-spot area. So far in the literature, only few results are found using the features of the vehicle to detect the vehicle. Without using features from vehicle, to conclude that vehicles do appear in that area with strong evidence is hard. We use the image features which are directly obtained from vehicle images to detect vehicles possibility in the area. In order to overcome large variation problem due to significant difference in view angle during the process of detecting vehicles in the blind-spot area, we propose a method to combine two kinds of part-based features. After building all the features from training images, we use Adaboost algorithm to choose the best features with better geometric information for detection. The experiments show that our system is reliably in detecting the vehicles in the blind-spot area.

This paper presents a novel approach to automatically detect and count cars in different aerial images, which can be satellite or unmanned aerial vehicle (UAV) images. Variations in satellite and/or UAV data make it particularly challenging to have a robust method that works properly on a variety of images. A solution based on the bag-of-words (BoW) model is explored in this paper due to its invariance characteristic and highly stable performance in object/scene categorization. Different from categorization tasks, vehicle detection needs to localize the positions of cars in images. To make BoW suitable for this purpose, we extensively improve the methodology in three aspects, namely, by introducing a recently proposed feature representation, i.e., the local steering kernel descriptor, adding spatial structure constraints, and developing an orientation aware scanning mechanism to produce detection with “one-window-one-car” results. Experiments are conducted on various aerial images with large variations, which consist of data from two public databases, e.g., the Overhead Imagery Research Data Set and Vehicle Detection in Aerial Imagery, as well as other satellite and UAV images. The results demonstrate the effectiveness and robustness of the proposed method. Compared with existing techniques, the proposed method is applicable to a wider range of aerial images.

This research focuses on assessing and developing a real-time detection system using the YOLOv8 algorithm. Accurate and fast vehicle detection is a big challenge in modern traffic management, especially in various environmental conditions such as bad weather, low lighting, and high traffic density. The aim of this study was to evaluate the performance of YOLOv8 under these conditions and identify potential improvements. The dataset used consists of 16,990 vehicle images with various variations and environmental conditions. After being trained, the model is evaluated using metrics such as precision, recall, and F1-score, as well as Intersection over Union (IoU) with a threshold of 0.8 on IoU. The results show that YOLOv8 is superior with a fairly high detection accuracy of 78%, with precision of 82% and recall above 90%, and is able to detect vehicles in real-time conditions. However, the challenge of detecting small objects or irregularly shaped vehicles such as tractors still needs to be optimized. This research also compared the performance of YOLOv8 with the SSD (Single Shot Detector) algorithm, where YOLOv8 was proven to be superior in terms of accuracy, precision, recall and F1-score. The research results obtained provide valuable insights for the development of traffic management systems based on deep learning technology. The main contribution of this research is to provide a more efficient and effective vehicle detection solution, which can be applied in modern traffic management systems. Thus, it is hoped that the results of this research can increase the efficiency of traffic management and have a positive impact on the development of intelligent transportation systems in the future.