Target Vehicle Research Articles

Enhancing YOLOv8’s Performance in Complex Traffic Scenarios: Optimization Design for Handling Long-Distance Dependencies and Complex Feature Relationships

In recent years, the field of deep learning and computer vision has increasingly focused on the problem of vehicle target detection, becoming the forefront of many technological innovations. YOLOv8, as an efficient vehicle target detection model, has achieved good results in many scenarios. However, when faced with complex traffic scenarios, such as occluded targets, small target detection, changes in lighting, and variable weather conditions, YOLOv8 still has insufficient detection accuracy and robustness. To address these issues, this paper delves into the optimization strategies of YOLOv8 in the field of vehicle target detection, focusing on the EMA module in the backbone part and replacing the original SPPF module with focal modulation technology, all of which effectively improved the model’s performance. At the same time, modifications to the head part were approached with caution to avoid unnecessary interference with the original design. The experiment used the UA-DETRAC dataset, which contains a variety of traffic scenarios, a rich variety of vehicle types, and complex dynamic environments, making it suitable for evaluating and validating the performance of traffic monitoring systems. The 5-fold cross-validation method was used to ensure the reliability and comprehensiveness of the evaluation results. The final results showed that the improved model’s precision rate increased from 0.859 to 0.961, the recall rate from 0.83 to 0.908, and the mAP50 from 0.881 to 0.962. Meanwhile, the optimized YOLOv8 model demonstrated strong robustness in terms of detection accuracy and the ability to adapt to complex environments.

Diversity-Representativeness Replay and Knowledge Alignment for Lifelong Vehicle Re-identification

Lifelong vehicle re-identification (LVReID) aims to match a target vehicle across multiple cameras, considering non-stationary and continuous data streams, which fits the needs of the practical application better than traditional vehicle re-identification. Nonetheless, this area has received relatively little attention. Recently, methods for lifelong person re-identification (LPReID) have been emerging, with replay-based methods achieving the best results by storing a small number of instances from previous tasks for retraining, thus effectively reducing catastrophic forgetting. However, these methods cannot be directly applied to LVReID because they fail to simultaneously consider the diversity and representativeness of replayed data, resulting in biases between the subset stored in the memory buffer and the original data. They randomly sample classes, which may not adequately represent the distribution of the original data. Additionally, these methods fail to consider the rich variation in instances of the same vehicle class due to factors such as vehicle orientation and lighting conditions. Therefore, preserving more informative classes and instances for replay helps maintain information from previous tasks and may mitigate the model’s forgetting of old knowledge. In view of this, we propose a novel Diversity-Representativeness Dual-Stage Sampling Replay (DDSR) strategy for LVReID that constructs an effective memory buffer through two stages, i.e. , Cluster-Centric Class Selection and Diverse Instance Mining. Specifically, we first perform class-level sampling based on density in the clustered class-centered feature space and then further mine the diverse, high-quality instances within the selected classes. In addition, we introduce Maximum Mean Discrepancy loss to align the feature distribution between replay data and the new arrivals and apply L2 regularisation in the parameter space to facilitate knowledge transfer, thus enhancing the model’s generalization ability to new tasks. Extensive experiments demonstrate effective improvements of our method compared to current state-of-the-art lifelong ReID methods on the VeRi-776, VehicleID, and VERI-Wild datasets.

Learning super-resolution and pyramidal convolution residual network for vehicle re-identification

Vehicle re-identification (Vehicle Re-ID) aims at retrieving and tracking the specified target vehicle with multiple other cameras, which can provide help in checking violations and catching fugitives, but there are still the following problems that need to be solved urgently. First, the existing collected Vehicle Re-ID data often have low resolution and blur in local regions, so that the Vehicle Re-ID algorithm cannot accurately extract subtle feature representations. In addition, small features are easy to cause the disappearance of features under the operation of a large convolution kernel, which makes the model unable to capture and learn subtle features, resulting in inaccurate judgment of vehicles. In this study, we propose a Vehicle Re-ID method based on super resolution and pyramidal convolution residual network. Firstly, a super-resolution image generation network leveraging generative adversarial networks (GANs) is proposed. This network employs both content loss and adversarial loss as optimization criteria, ensuring an efficient transformation from a low-resolution image into a super-resolution counterpart, while meticulously preserving intricate high-frequency details. Then, multi levels of pyramidal convolution operations are designed to generate multi-scale features, which can capture information on different scales. Moreover, the concept of residual learning is applied between the multi levels of pyramidal convolution operations to expedite model optimization and enhance recognition capabilities. Ultimately, the double pyramidal convolutions are meticulously employed on both the original image and the super-resolution image, yielding low-noise feature representations and intricate semantic information respectively. By seamlessly fusing these two diverse sources of information, the resultant combined features exhibit heightened discrimination capabilities and significantly bolster the robustness of image features. In order to verify the effectiveness of the proposed method, extensive experiments are carried out on VeRi-776 and VehicleID datasets. The experimental results show that the method proposed in this paper effectively captures the detail information of vehicle images, accurately distinguishes the subtle differences between different vehicles of the same type, and is superior to state-of-the-art methods.

PV-YOLO: A lightweight pedestrian and vehicle detection model based on improved YOLOv8

With the frequent occurrence of urban traffic accidents, fast and accurate detection of pedestrian and vehicle targets has become one of the key technologies for intelligent assisted driving systems. To meet the efficiency and lightweight requirements of smart devices, this paper proposes a lightweight pedestrian and vehicle detection model based on the YOLOv8n model, named PV-YOLO. In the proposed model, receptive-field attention convolution (RFAConv) serves as the backbone network because of its target feature extraction ability, and the neck utilizes the bidirectional feature pyramid network (BiFPN) instead of the original path aggregation network (PANet) to simplify the feature fusion process. Moreover, a lightweight detection head is introduced to reduce the computational burden and improve the overall detection accuracy. In addition, a small target detection layer is designed to improve the accuracy for small distant targets. Finally, to reduce the computational burden further, the lightweight C2f module is utilized to compress the model. The experimental results on the BDD100K and KITTI datasets demonstrate that the proposed PV-YOLO can achieve higher detection accuracy than YOLOv8n and other baseline methods with less model complexity.

Dynamic radar cross section similarity study based on dynamic time warping

Abstract Dynamic radar cross-section (RCS) represents the radar reflectance cross-section of an aircraft at different moments during flight and is a crucial criterion for evaluating the ability of a radar to detect an aircraft. In practice, it is challenging to construct an accurate 6-DOF dynamics model for the purpose of solving its dynamic RCS. Consequently, a 3-DOF dynamics model is employed for the analysis of dynamic RCS. This paper presents a methodology for the joint simulation of the dynamic RCS under two DOF models and subsequent comparison and analysis of the resulting sequence data via the dynamic time warping algorithm. It is also concluded that the dynamic RCS values calculated using the 3-DOF dynamics model can identify the target vehicle.

Risk-Aware Lane Change and Trajectory Planning for Connected Autonomous Vehicles Based on a Potential Field Model

To enhance the safety of lane changes for connected autonomous vehicles in an intelligent transportation environment, this study draws from potential field theory to analyze variations in the risks that vehicles face under different traffic conditions. The safe minimum vehicle distance is dynamically adjusted, and a comprehensive vehicle risk potential field model is developed. This model systematically quantifies the risks encountered by connected autonomous vehicles during the driving process, providing a more accurate assessment of safety conditions. Subsequently, vehicle motion is decoupled into lateral and longitudinal components within the Frenet coordinate system, with quintic polynomials employed to generate clusters of potential trajectories. To improve computational efficiency, trajectory evaluation metrics are developed based on vehicle dynamics, incorporating factors such as acceleration, jerk, and curvature. An initial filtering process is applied to these trajectories, yielding a refined set of candidates. These candidate trajectories are further assessed using a minimum safety distance model derived from potential field theory, with optimization focusing on safety, comfort, and efficiency. The algorithm is tested in a three-lane curved simulation environment that includes both constant-speed and variable-speed lane change scenarios. Results show that the collision risk between the target vehicle and surrounding vehicles remains below the minimum safety distance threshold throughout the lane change process, ensuring a high level of safety. Furthermore, across various driving conditions, the target vehicle’s acceleration, jerk, and trajectory curvature remained well within acceptable limits, demonstrating that the proposed lane change trajectory planning algorithm successfully balances safety, comfort, and smoothness, even in complex traffic environments.

A short-term vehicle speed prediction approach considering dynamic traffic scene

Vehicle speed serves as a crucial indicator in traffic flow efficiency and safety evaluation. Previous researches typically utilize the time-series data of target vehicle speed or the relative speed and distance between the target and leading vehicles as speed prediction model input, which are suitable for stable and single-vehicle scenes. However, the surrounding traffic scenes dynamically changes during the vehicle’s driving process, the driver needs to allocate attention to the key scene vehicles in the surrounding traffic scenes and adjust the current speed based on these vehicles’ behavior. Therefore, this paper proposes a short-time vehicle speed prediction model considering the dynamic traffic scenes. A dynamic grid method is adopted to divide the target vehicle’s surrounding areas into left front, front, and right front regions. And key scene vehicles are identified based on each region’s state. Then the relative distance, speed, and acceleration data between the target vehicle and key scene vehicles are selected as model inputs. Vehicle trajectory data from the South-North No. 1 Expressway in Hefei and Ranger optimized Temporal Convolutional Network (TCN) are utilized to train the speed prediction model. Experiment results indicated that under 1 s, 2 s, and 3 s prediction lengths, the model’s average Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) were 0.430 and 0.271 respectively. Compared to the speed prediction model considering the static traffic scenes, the average RMSE and MAE for the three prediction lengths have been reduced by 49.15 % and 56.02 % respectively.

Research on the detection and recognition system of target vehicles based on fusion algorithm

<p>In modern society, the monitoring of vehicles is paramount for upholding traffic safety and order. This paper delves into a Graphical User Interface (GUI)-driven system for vehicle target detection and integration. This system seamlessly integrates three pivotal functionalities: vehicle type recognition, license plate recognition, and driver face recognition. It automates the vehicle inspection assessment process through an intuitive user interface. Specifically, for license plate recognition, the system leverages traditional computer vision techniques, including color and grayscale processing, radon transform, morphological operations, edge detection, character segmentation, and character recognition. The driver face recognition module incorporates methods such as image gray scaling, Gaussian filtering for denoising, face detection, and feature extraction. Meanwhile, vehicle type recognition is achieved by extracting Histogram of Oriented Gradients (HOG) features and utilizing a Support Vector Machine (SVM) classifier. Experimental findings highlight the system’s remarkable recognition accuracy, offering a highly efficient and dependable solution for the automated inspections of vehicles.</p>

Vehicle Localization Method in Complex SAR Images Based on Feature Reconstruction and Aggregation.

Due to the small size of vehicle targets, complex background environments, and the discrete scattering characteristics of high-resolution synthetic aperture radar (SAR) images, existing deep learning networks face challenges in extracting high-quality vehicle features from SAR images, which impacts vehicle localization accuracy. To address this issue, this paper proposes a vehicle localization method for SAR images based on feature reconstruction and aggregation with rotating boxes. Specifically, our method first employs a backbone network that integrates the space-channel reconfiguration module (SCRM), which contains spatial and channel attention mechanisms specifically designed for SAR images to extract features. The network then connects a progressive cross-fusion mechanism (PCFM) that effectively combines multi-view features from different feature layers, enhancing the information content of feature maps and improving feature representation quality. Finally, these features containing a large receptive field region and enhanced rich contextual information are input into a rotating box vehicle detection head, which effectively reduces false alarms and missed detections. Experiments on a complex scene SAR image vehicle dataset demonstrate that the proposed method significantly improves vehicle localization accuracy. Our method achieves state-of-the-art performance, which demonstrates the superiority and effectiveness of the proposed method.

Vehicle object detection and ranging in vehicle images based on deep learning

In order to improve the accuracy of vehicle target detection and the stability of ranging in driving environments, a vehicle target detection and ranging method based on deep learning is proposed. The YOLOX-S algorithm is used as the vehicle target detection framework for improvement: the CBAM attention module is introduced on the basis of the original algorithm to enhance the network feature expression ability, and the confidence loss function is replaced by Focal Loss to reduce the training weight of simple samples and improve the attention of positive samples. The vehicle ranging model is established according to the imaging principle and geometric relationship of the vehicle camera, and the ranging feature point coordinates and camera internal parameters are input to obtain the ranging results. The self-made Tlab dataset and BDD 100K dataset are used to train and evaluate the improved YOLOX-S algorithm, and a static ranging experimental scene is built to verify the vehicle ranging model. The experimental results show that the improved YOLOX-S algorithm has a detection speed of 70.14 frames per second on the experimental data set. Compared with the original algorithm, the precision, recall, F1 value, and mAP are improved respectively 0.86%、1.32%、1.09%、1.54% ; within the measurement range of 50 m in the longitudinal direction and 11.25 m in the lateral direction, the average ranging error is kept within 3.20% . It can be seen that the proposed method has good vehicle ranging accuracy and stability while meeting the real-time requirements of vehicle detection.

Research on vehicle trajectory prediction methods in dense and heterogeneous urban traffic

ABSTRACT In autonomous driving, accurately predicting the trajectories of surrounding vehicles is essential, particularly in dense and heterogeneous urban traffic. We propose a graph-structured model with a category layer to efficiently forecast the target vehicle’s trajectory. The model enables flexible selection of interacting objects based on environmental interactions and extracts spatial-temporal features using a graph convolutional network. A categorical layer is introduced to account for the different influences of dynamic agents, while vehicle dynamics constraints ensure the feasibility of predicted trajectories. We developed a new heterogeneous and dense urban unsignalized intersection dataset (HID), capturing complex urban interactions, and conducted extensive experiments on HID, ApolloScape, and TRAF datasets. Results demonstrate that our model outperforms benchmark methods across diverse urban scenarios, and the integration of key modules significantly enhances prediction accuracy and performance.

Optimal powertrain system design and V2V and V2I based hierarchical control strategy of FRIDEV under vehicle-following scenarios

In this paper, the front- and rear-independent-drive electric vehicle (FRIDEV) is selected as the target vehicle for its good dynamic and economy performance. Based on the analysis of the cost, efficiency, and loss characteristics of different motors, the interior permanent magnet synchronous motor (IPMSM) and induction motor (IM) are applied in the target powertrain system. Then, the proposed powertrain system is optimized based on the analysis of driving cycles. The optimal control strategy for the target powertrain system is developed, which consists of the driving mode switching algorithm and the driving torque distribution based on adaptive particle swarm optimization (APSO). Furthermore, in order to achieve the efficient autonomous driving of the target vehicle under the vehicle-following scenarios, a hierarchical control strategy is proposed with the combination of upper-level control of the speed planning strategy and the lower-level control of powertrain system control strategy. The upper-level control of speed planning strategy is designed, which can determine and track the target vehicle speed based on the information from vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. Finally, simulation analysis in MATLAB/Simulink and virtual driving experiments are conducted to verify the superiority of the proposed powertrain system and hierarchical control strategy.

Research on the strategy of cruise control system for urban traffic jam assistant

This paper proposes a hierarchical cruise control strategy that can adapt to urban traffic scenarios. Firstly, the perception layer defines a target selection strategy by using a novel trapezoidal dangerous area. Then, the upper controller contains speed follow mode and distance follow mode. An improved PI mode with integral separation and saturation terms is developed to achieve the desired speed following control, a safe distance mode via variable time headway is developed to facilitate the safe and smooth distance following control. The lower controller is designed based on feedforward and PI feedback. Finally, the experimental results demonstrate that the strategy can effectively select suitable target vehicles in complex urban scenarios. With the improved safe distance model and the third-order Bessel actuator response delay model, the strategy exhibits adaptability and stability. This study offers valuable suggestions and guidance for designing and applying the cruise control strategy for urban traffic scenarios.

An Ontology-Based Vehicle Behavior Prediction Method Incorporating Vehicle Light Signal Detection.

Although deep learning techniques have potential in vehicle behavior prediction, it is difficult to integrate traffic rules and environmental information. Moreover, its black-box nature leads to an opaque and difficult-to-interpret prediction process, limiting its acceptance in practical applications. In contrast, ontology reasoning, which can utilize human domain knowledge and mimic human reasoning, can provide reliable explanations for the speculative results. To address the limitations of the above deep learning methods in the field of vehicle behavior prediction, this paper proposes a front vehicle behavior prediction method that combines deep learning techniques with ontology reasoning. Specifically, YOLOv5s is first selected as the base model for recognizing the brake light status of vehicles. In order to further enhance the performance of the model in complex scenes and small target recognition, the Convolutional Block Attention Module (CBAM) is introduced. In addition, so as to balance the feature information of different scales more efficiently, a weighted bi-directional feature pyramid network (BIFPN) is introduced to replace the original PANet structure in YOLOv5s. Next, using a four-lane intersection as an application scenario, multiple factors affecting vehicle behavior are analyzed. Based on these factors, an ontology model for predicting front vehicle behavior is constructed. Finally, for the purpose of validating the effectiveness of the proposed method, we make our own brake light detection dataset. The accuracy and mAP@0.5 of the improved model on the self-made dataset are 3.9% and 2.5% higher than that of the original model, respectively. Afterwards, representative validation scenarios were selected for inference experiments. The ontology model created in this paper accurately reasoned out the behavior that the target vehicle would slow down until stopping and turning left. The reasonableness and practicality of the front vehicle behavior prediction method constructed in this paper are verified.

Efficient tuning of active sound design in electric vehicles - An interactive validation approach

The implementation of active sound design models in a target vehicle requires careful tuning of all synthetic sound elements with respect to existent vehicle interior noise, current driving conditions and general driving styles. Besides the creative aspects in sound design, this tuning process is highly time-consuming. It must be ensured that load feedback, speed feedback and general interactivity with the synthetic sound meet the driver's expectations and match the intended emotional expressions of the sound concept. Testing the tuning parameters in real vehicles requires access to the vehicle prototype and all involved system components, which is often not granted in early development stages. A key approach to overcome this difficulty is the application of acoustic driving simulator technology, directly linked to the main tuning parameters. Combining this methodology with an efficient procedure for perceptive assessment of the sound design model, an interactive approach to active sound design is established. This paper discusses current progress in the field of efficient tuning of active sound design models, presenting a case study for a high-performance electric vehicle.

Wireless sensor fusion localization method for target vehicles in a collaborative environment of vehicle road trajectory information

Abstract Vehicle positioning technology is the foundation for achieving traffic management. Vehicle positioning errors are significant in complex environments, making it challenging to ensure positioning accuracy. To reduce positioning errors, this paper explores a wireless sensor fusion localization method for target vehicles in a collaborative environment of vehicle road trajectory information. Based on wireless sensor data, we achieve collaborative environmental perception of vehicle and road trajectory information and construct a target vehicle trajectory optimization model. Through a fuzzy neural network, data from different sensors are integrated to overcome the problem of limited positioning accuracy of a single sensor in a complex environment. The experimental results show that both the heading angle and rolling angle errors float between −0.1° and 0.1°, and the positioning error is less than 0.1 m, which proves that the method has high positioning accuracy and good application value.

SPCF-YOLO: Dense Small Target Detection Algorithm for Autonomous Driving

Abstract In the automatic driving scene, the pedestrian and vehicle targets are dense and block each other, and the problem of target false detection and missing detection is easy to occur. In this paper, a dense small target detection algorithm SPCF-YOLO is proposed. A new SPCSA module has been designed to strengthen the capture of edge features in occluded small targets. MPDIoU replaces the original CIOU loss function to address cases where the predicted and real frames have the same aspect ratio, but the aspect ratio difference of anchor frame is large. Experimental results show that in KITTI 2D data set, compared with the original YOLOv7 algorithm, mAP@0.5 is 93.5%, an increase of 0.5%. mAP@0.5:0.95 is 66.9%, an increase of 2%; micro-target mAP@0.5:0.95 is 50.3%, an increase of 2.6%; and the computational load and parameter number remain basically unchanged, which proves the effectiveness of the design algorithm.

UAV Vehicle Detection System Based on YOLOv8

Abstract To utilize the drone vehicle detection technology to observe the road traffic conditions in real-time, thereby improving the operational effectiveness and safety of the traffic system, this paper adopts a vehicle target detection algorithm based on YOLOv8. YOLOv8 is the latest iteration developed by Ultralytics, which has further improvements and enhancements compared to YOLOv5. In the backbone network, YOLOv8 employs a C2f structure with richer gradient flow, boosting feature selection capabilities, and enhancing target detection performance and accuracy; In the detection head, YOLOv8 utilizes a Decoupled Head design and uses an improved Anchor-Free method to reduce computational complexity while improving detection efficiency; In terms of loss function, YOLOv8 utilizes the The Task-Aligned Assigner positive sample allocation strategy, and adds the Distribution Focal Loss to improve the model’s generalization ability. This paper divides the vehicles data set into 5 types, trains it using a pre-training model, and optimizes the network parameters through multiple iterations to achieve better detection performance. The experimental results indicate that YOLOv8 reaches the mAP of over 97% on the experimental data set, which is an improvement of approximately 8% compared to YOLOv5, and achieves a good detection performance, and is highly practical.

A Target Collision Algorithm of Autonomous Mobile Robot for Two-player Competition

Abstract A variety of robot competitions are held to train and attract students everywhere. This paper studies the algorithm of autonomous robot with two-player competition. The autonomous robots in a two-player competition takes the scoring form of collision and wins first. The impact object is the enemy target. The key algorithms in a two-player competition are target detection, target path decision, and collision target motion control. First, the vehicle target data set was established and enhanced, the lightweight detection algorithm model was screened and trained, the model pruning optimization experiment and the model deployment optimization experiment were carried out, and then the TensorRT framework was used for inter-layer fusion and data quantification processing. After hitting the target path decision algorithm after detecting the target target, it finally integrates the decision in simple and complex situations. The collision target motion control is the scheme of more intelligent and self-adjustable parameters of fuzzy Radial Basis Function (RBF) network Proportional Integral Derivative (PID) controller. Finally, the collision target implementation algorithm is tested in a real environment, and the results show that the task requirements can be met.

A Multi-Scale Feature Fusion Based Lightweight Vehicle Target Detection Network on Aerial Optical Images

Vehicle detection with optical remote sensing images has become widely applied in recent years. However, the following challenges have remained unsolved during remote sensing vehicle target detection. These challenges include the dense and arbitrary angles at which vehicles are distributed and which make it difficult to detect them; the extensive model parameter (Param) that blocks real-time detection; the large differences between larger vehicles in terms of their features, which lead to a reduced detection precision; and the way in which the distribution in vehicle datasets is unbalanced and thus not conducive to training. First, this paper constructs a small dataset of vehicles, MiVehicle. This dataset includes 3000 corresponding infrared and visible image pairs, offering a more balanced distribution. In the infrared part of the dataset, the proportions of different vehicle types are as follows: cars, 48%; buses, 19%; trucks, 15%; freight, cars 10%; and vans, 8%. Second, we choose the rotated box mechanism for detection with the model and we build a new vehicle detector, ML-Det, with a novel multi-scale feature fusion triple cross-criss FPN (TCFPN), which can effectively capture the vehicle features in three different positions with an mAP improvement of 1.97%. Moreover, we propose LKC–INVO, which allows involution to couple the structure of multiple large kernel convolutions, resulting in an mAP increase of 2.86%. We also introduce a novel C2F_ContextGuided module with global context perception, which enhances the perception ability of the model in the global scope and minimizes model Params. Eventually, we propose an assemble–disperse attention module to aggregate local features so as to improve the performance. Overall, ML-Det achieved a 3.22% improvement in accuracy while keeping Params almost unchanged. In the self-built small MiVehicle dataset, we achieved 70.44% on visible images and 79.12% on infrared images with 20.1 GFLOPS, 78.8 FPS, and 7.91 M. Additionally, we trained and tested our model on the following public datasets: UAS-AOD and DOTA. ML-Det was found to be ahead of many other advanced target detection algorithms.