Scene-Adaptive Vehicle Detection Algorithm Based on a Composite Deep Structure

In order to achieve rapid and accurate detection of vehicle objects in complex traffic conditions, we propose a novel vehicle detection method. Firstly, more contextual and small-object vehicle information can be obtained by our Joint Feature Network (JFN). Secondly, our Evolved Region Proposal Network (EPRN) generates initial anchor boxes by adding an improved version of the region proposal network in this network, and at the same time filters out a large number of false vehicle boxes by soft-Non Maximum Suppression (NMS). Then, our Mask Network (MaskN) generates an example that includes the vehicle occlusion, the generator and discriminator can learn from each other in order to further improve the vehicle object detection capability. Finally, these candidate vehicle detection boxes are optimized to obtain the final vehicle detection boxes by the Fine-Tuning Network(FTN). Through the evaluation experiment on the DETRAC benchmark dataset, we find that in terms of mAP, our method exceeds Faster-RCNN by 11.15%, YOLO by 11.88%, and EB by 1.64%. Besides, our algorithm also has achieved top2 comaring with MS-CNN, YOLO-v3, RefineNet, RetinaNet, Faster-rcnn, DSSD and YOLO-v2 of vehicle category in KITTI dataset.

General object-detection methods based on deep learning have received considerable attention in the field of computer vision. However, when they are applied to vehicle detection (VD) in a straightforward manner to realize an intelligent vehicle (IV), a graphics processing unit (GPU) is required for their real-time implementation. The use of GPUs is unacceptable in commercial VD systems. A novel on-road VD method comprising the use of a multi-stage convolutional neural network (MSCNN) is proposed to solve this problem. In the MSCNN, the properties of the vehicles are exploited, and an efficient region proposal specialized for vehicles is developed. The proposed MSCNN comprises four stages: vehicle lower-boundary detection, vehicle upper-boundary detection, region proposal network (RPN), and vehicle classification. Effective anchor boxes are generated in the lower- and upper-boundary-detection stages with appropriate sizes for vehicles of all scales. The bounding box of the vehicle within the anchor box is determined in the RPN stage. In the last stage, the predicted bounding boxes are classified as vehicles or non-vehicles. Finally, the proposed method is applied to the KITTI, CrowdAI, and AUTTI datasets, and its advantages are demonstrated by comparing its performance with those presented in previous studies. The proposed MSCNN realizes an average precision (72.1%) on the KITTI dataset while running on a central processing unit (CPU).

In this paper, we propose a new exemplar-based multi-view domain generalization (EMVDG) framework for visual recognition by learning robust classifier that are able to generalize well to arbitrary target domain based on the training samples with multiple types of features (i.e., multi-view features). In this framework, we aim to address two issues simultaneously. First, the distribution of training samples (i.e., the source domain) is often considerably different from that of testing samples (i.e., the target domain), so the performance of the classifiers learnt on the source domain may drop significantly on the target domain. Moreover, the testing data are often unseen during the training procedure. Second, when the training data are associated with multi-view features, the recognition performance can be further improved by exploiting the relation among multiple types of features. To address the first issue, considering that it has been shown that fusing multiple SVM classifiers can enhance the domain generalization ability, we build our EMVDG framework upon exemplar SVMs (ESVMs), in which a set of ESVM classifiers are learnt with each one trained based on one positive training sample and all the negative training samples. When the source domain contains multiple latent domains, the learnt ESVM classifiers are expected to be grouped into multiple clusters. To address the second issue, we propose two approaches under the EMVDG framework based on the consensus principle and the complementary principle, respectively. Specifically, we propose an EMVDG_CO method by adding a co-regularizer to enforce the cluster structures of ESVM classifiers on different views to be consistent based on the consensus principle. Inspired by multiple kernel learning, we also propose another EMVDG_MK method by fusing the ESVM classifiers from different views based on the complementary principle. In addition, we further extend our EMVDG framework to exemplar-based multi-view domain adaptation (EMVDA) framework when the unlabeled target domain data are available during the training procedure. The effectiveness of our EMVDG and EMVDA frameworks for visual recognition is clearly demonstrated by comprehensive experiments on three benchmark data sets.

We present a novel method for vehicle detection and classification in aerial imagery. First, change detection analyzes a pair of mutually aligned images captured at the same location but at different time instants to generate vehicle proposals. Next, a trained Convolutional Neural Network (CNN) classifier is applied to (1) determine if a proposal truly contains a vehicle, and (2) classify any vehicle present into major categories. Experimental results using infrared (IR) data demonstrate the efficacy of our method: vehicle detection rate is over 99%; light-duty vehicles (e.g. sedan) are classified with 89% accuracy, medium-duty vehicles (e.g., van and pickup) with 79% and heavy-duty vehicles (e.g., trucks and buses) with 73% accuracies. The classification performance difference among different vehicle types is attributed to the size of training samples available in each category. Our system can process data at video rate making it feasible for real-time traffic monitoring.

Vision-based traffic surveillance has been one of the most promising fields for improvement and research. Still, many challenging problems remain unsolved, such as addressing vehicle occlusions and reducing false detection. In this work, a method for vehicle detection and tracking is proposed. The proposed model considers background subtraction concept for moving vehicle detection but unlike conventional approaches, here numerous algorithmic optimization approaches have been applied such as multi-directional filtering and fusion based background subtraction, thresholding, directional filtering and morphological operations for moving vehicle detection. In addition, blob analysis and adaptive bounding box is used for Detection and Tracking. The Performance of Proposed work is measured on Standard Dataset and results are encouraging.

Deep learning-based classification and detection algorithms have emerged as a powerful tool for vehicle detection in intelligent transportation systems. The limitations of the number of high-quality labeled training samples makes the single vehicle detection methods incapable of accomplishing acceptable accuracy in road vehicle detection. This paper presents detection and classification of vehicles on publicly available datasets by utilizing the YOLO-v5 architecture. This paper’s findings utilize the concept of transfer learning through fine tuning the weights of the pre-trained YOLO-v5 architecture. To employ the concept of transfer learning, extensive data sets of images and videos of the congested traffic patterns were collected by the authors. These datasets were made more comprehensive by pointing various attributes, for instance high- and low-density traffic patterns, occlusions, and different weather circumstances. All of these gathered datasets were manually annotated. Ultimately, the improved YOLO-v5 structure becomes accustomed to any difficult traffic patterns. By fine-tuning the pre-trained network through our datasets, our proposed YOLO-v5 has exceeded several other traditional vehicle detection methods in terms of detection accuracy and execution time. Detailed simulations performed on the PKU, COCO, and DAWN datasets demonstrate the effectiveness of the proposed method in various challenging situations.

An adaptive confidence algorithm for vehicle detection is proposed. Shortcomings of traditional vehicle detection method of deep learning with fixed confidence threshold are analyzed. To deal with false alarms and false negatives brought by these shortcomings, an adaptive confidence threshold algorithm for vehicle detection is raised. Two datasets are set up to test and compare the detection performance of the adaptive confidence algorithm with fixed confidence on MS-CNN. The experimental results show that the proposed algorithm could effectively reduce the false alarm rate and improve the detection performance of MS-CNN. It is of certain significance for the application of deep learning technology to vehicle detection.

Occlusion is an important problem of the moving target detection. This paper proposed a new method of vehicle detection according to the deficiencies of common vehicle detection methods. Firstly, the background is modeled through the improved histogram-mean model to extract more accurate background model and update in real-time; then we obtained the background through background subtraction and supplement edge information; at last get the complete foreground information of vehicle by the morphological to denoise and fill the empty positions of windows. Based on the occlusion of target in the foreground, this paper presents an effective method to obtain the accurate vehicles using adhesion point detection. The experimental results show that this method can accurately detect various types of vehicles, the computational complexity is low, be able to extract traffic information in real-time, and solved the problem of vehicle occlusion.

Vehicle detection and tracking plays a major role in military and civilian applications. The accurate detection of multiple vehicles in a complicated traffic environment is too difficult. This process is made more difficult, if there are occlusions between vehicles. The optimal feature selection of a vehicle is also a challenging task. These issues are focused in the existing work namely convolution neural network (CNN) based vehicle detection. However in this work, feature selection is not concentrated and with the lesser feature information classification accuracy is reduced considerably. This is focused and resolved in the proposed research by introducing the method namely Enhanced Convolution neural network with Support Vector Machine (ECNN-SVM) based vehicle detection. In this work initially feature extraction is done. Haar like features are used for this purpose. This has compact representation, capture information from multiple scales, encoded edge and structural information and can be efficient computation. Haar like feature pool is large scale in nature. Enhanced bat optimization is used to select features from this pool. This selection is done by combining sample’s class label and feature value. Confidence level between two thresholds of tracking rectangles is used for classification. Support vector machine (SVM) classifier is combined with local binary pattern to perform this classification. The interference areas between moving objects and vehicles are removed by Enhanced Convolutional Neural Network (ECNN) classifier. The efficiency of the proposed method is computed by using experimental and theoretical analysis.

Intelligent transportation systems (ITS) depend on precise vehicle identification and speed estimation to provide real-time traffic monitoring, accident avoidance, and effective traffic flow management. This paper, a novel deep neural network (DNN)-based method for real-time vehicle detection and speed estimate is presented. For accurate vehicle detection, we use cutting-edge object detection models like YOLOv5, and we use computer vision algorithms to determine a vehicle's speed from video frames. The technology operates reliably in a range of environmental circumstances and processes real-time traffic video streams. High detection accuracy (above 95%) and accurate speed estimation with an average error margin of less than 5 km/h are demonstrated by the experimental findings. A scalable and economical architecture for traffic surveillance applications is provided by the suggested approach.

This research develops an advanced vehicle detection method, which improves the original Viola-Jones (V-J) object detection scheme for better vehicle detections from low-altitude unmanned aerial vehicle (UAV) imagery. The original V-J method is sensitive to objects’ in-plane rotation, and therefore has difficulties in detecting vehicles with unknown orientations in UAV images. To address this issue, this research proposes a road orientation adjustment method, which rotates each UAV image once so that the roads and on-road vehicles on rotated images will be aligned with the horizontal direction and the V-J vehicle detector. Then, the original V-J can be directly applied to achieve better efficiency and accuracy. The enhanced V-J method is further applied for vehicle tracking. Testing results show that both vehicle detection and tracking methods are competitive compared with other existing methods. Future research will focus on expanding the current methods to detect other transport modes, such as buses, trucks, motorcycles, bicycles, and pedestrians.

With the rapid development of intelligent transportation systems, vehicle detection and traffic flow statistics have become increasingly critical in traffic management and the construction of smart cities. However, complex weather conditions lead to challenges in vehicle detection algorithms, increasing false negatives and false positives. Additionally, the rapid growth of vehicle ownership makes traditional algorithms inadequate for meeting practical application needs. Therefore, this paper proposes a collaborative optimization-based method combining YOLOv11 for vehicle detection and traffic flow analysis under complex weather conditions. Vehicle information is accurately extracted through object detection and tracking methods, and further analyzed using Logistic Regression to assess traffic congestion. The main contributions of this study are: (1) Integration of the CBAM attention mechanism into YOLOv11, enhancing the vehicle feature extraction ability through collaborative channel and spatial attention, thereby improving interference resistance under complex weather conditions. (2) Adoption of the Shape-IoU loss function, with dynamic weight allocation to optimize the accuracy of detecting smaller vehicles and the precision of bounding box regression in images where vehicles are located at a greater distance. (3) Construction of a traffic congestion prediction model based on the detection results, integrated with Logistic Regression. The proposed method's effectiveness is validated on the public dataset UA-DETRAC. Through the synergistic effect of the CBAM attention mechanism and Shape-IoU loss function, the improved model's mAP50-95 shows a 0.3 percentage point increase over the YOLOv11 baseline, providing efficient and reliable technical support for intelligent transportation systems. This study enhances the prediction performance of traffic states under complex weather conditions by effectively calculating traffic flow parameters such as vehicle count and density, thus offering foundational data support for traffic management decision-making. The research conclusions are of practical significance for urban traffic improvement.

Video hyperlinking represents a classical example of multimodal problems. Common approaches to such problems are early fusion of the initial modalities and crossmodal translation from one modality to the other. Recently, deep neural networks, especially deep autoencoders, have proven promising both for crossmodal translation and for early fusion via multimodal embedding. A particular architecture, bidirectional symmetrical deep neural networks, have been proven to yield improved multimodal embeddings over classical autoencoders, while also being able to perform crossmodal translation. In this work, we focus firstly at evaluating good single-modal continuous representations both for textual and for visual information. Word2Vec and paragraph vectors are evaluated for representing collections of words, such as parts of automatic transcripts and multiple visual concepts, while different deep convolutional neural networks are evaluated for directly embedding visual information, avoiding the creation of visual concepts. Secondly, we evaluate methods for multimodal fusion and crossmodal translation, with different single-modal pairs, in the task of video hyperlinking. Bidirectional (symmetrical) deep neural networks were shown to successfully tackle downsides of multimodal autoencoders and yield a superior multimodal representation. In this work, we extensively tests them in different settings, with different single-modal representations, within the context of video-hyperlinking. Our novel bidirectional symmetrical deep neural networks are compared to classical autoencoders and are shown to yield significantly improved multimodal embeddings that significantly (alpha=0.0001) outperform multimodal embeddings obtained by deep autoencoders with an absolute improvement in precision at 10 of 14.1% when embedding visual concepts and automatic transcripts and an absolute improvement of 4.3% when embedding automatic transcripts with features obtained with very deep convolutional neural networks, yielding 80% of precision at 10.

Most of the current vision sensor-based target detection is suitable for good weather conditions. Adverse weather conditions, especially foggy environments, significantly reduce visibility, which seriously affects the target detection performance. To improve driving safety in foggy environments, this paper proposes an improved YOLOX-based vehicle and pedestrian detection method in foggy environments. The method is based on the advanced YOLOX network model and introduces an attention mechanism in the feature extraction network to enhance the network's extraction of target features in foggy images. Some images in the training dataset are fogged to supplement the target-specific features in foggy environments and improve the robustness of the target detection network in foggy environments. The idea of migration learning is used in the training process to save training time and optimize the training effect. The experimental results show that the target detection method proposed in this paper has significantly improved the detection performance of vehicles and pedestrians in the foggy environment, with an 11.35% improvement in mAP, and the detection effect is better than the GCANet image defogging method. The effectiveness of the method improvement is proved.

Vehicle detection is an important technology in au-tonomous driving, for which high detection accuracy and real-time performance are often required. The YOLOv5-GE vehicle detection algorithm is proposed to address the situation that the YOLOv5 vehicle detection model has false detection and missed detection for small and dense targets in complex environments. The global attention mechanism is added to the backbone net-work of the YOLOOv5 model, which is composed of two inde-pendent submodules of channel attention and convolutional spa-tial attention, which prevents the loss of information to a certain extent and amplifies the interaction of global dimensions. Second-ly, the training process is optimized using the Focal-EloU loss function to replace the GloU loss function, which improves the accuracy of vehicle detection. Finally, the proposed YOLOv5-GE algorithm and the YOLOv5 algorithm are subjected to a con-trolled experiment on the KITTI dataset. The experimental re-sults show that the YOLOv5-GE algorithm achieves an average accuracy of 86% while maintaining real-time performance, which is 2.5% higher than that of the YOLOv5 algorithm, and can im-prove the detection accuracy of small and dense targets in com-plex environments.