Vehicles In Traffic Scenes Research Articles

Research on vehicle detection based on improved YOLOX_S

Aiming at the problem of easy misdetection and omission of small targets of long-distance vehicles in detecting vehicles in traffic scenes, an improved YOLOX_S detection model is proposed. Firstly, the redundant part of the original YOLOX_S network structure is clipped using the model compression strategy, which improves the model inference speed while maintaining the detection accuracy; secondly, the Resunit_CA structure is constructed by incorporating the coordinate attention module in the residual structure, which reduces the loss of feature information and improves the attention to the small target features; thirdly, in order to obtain richer small target features, the PAFPN structure tail to add an adaptive feature fusion module, which improves the model detection accuracy; finally, the loss function is optimized in the decoupled head structure, and the Focal Loss loss function is used to alleviate the problem of uneven distribution of positive and negative samples. The experimental results show that compared with the original YOLOX_S model, the improved model proposed in this paper achieves an average detection accuracy of 77.19% on this experimental dataset. However, the detection speed decreases to 29.73 fps, which is still a large room for improvement in detection in real-time. According to the visualization experimental results, it can be seen that the improved model effectively alleviates the problems of small-target missed detection and multi-target occlusion.

Fast vehicle detection algorithm in traffic scene based on improved SSD

For autonomous driving systems, vehicle detection is an important part and challenging problem due to the complex traffic scenes and poor computing resources. This paper proposes an improved SSD(single shot mutlibox detector) algorithm for the fast detection of vehicles in traffic scenes. MobileNet v2 is selected as the backbone feature extraction network for SSD, which improves the real-time performance of the algorithm. To improve detection accuracy, the channel attention mechanism is utilized for feature weighting, and the deconvolution module is utilized to construct a bottom–top feature fusion structure. The experimental results show that the average precision of the proposed algorithm on the BDD100K and KITTI datasets is 82.59% and 84.83%, respectively. The single inference time of the algorithm is 73ms, which is only about 5/11 of the original model, realizing the improvement of inference speed and prediction accuracy.

Multi-target Tracking Based on Deep Sort in Traffic Scene

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.

An Efficient Optimal Neural Network-Based Moving Vehicle Detection in Traffic Video Surveillance System

This paper presents an effective traffic video surveillance system for detecting moving vehicles in traffic scenes. Moving vehicle identification process on streets is utilized for vehicle tracking, counts, normal speed of every individual vehicle, movement examination, and vehicle classifying targets and might be executed under various situations. In this paper, we develop a novel hybridization of artificial neural network (ANN) and oppositional gravitational search optimization algorithm (ANN–OGSA)-based moving vehicle detection (MVD) system. The proposed system consists of two main phases such as background generation and vehicle detection. Here, at first, we develop an efficient method to generate the background. After the background generation, we detect the moving vehicle using the ANN–OGSA model. To increase the performance of the ANN classifier, we optimally select the weight value using the OGSA algorithm. To prove the effectiveness of the system, we have compared our proposed algorithm with different algorithms and utilized three types of videos for experimental analysis. The precision of the proposed ANN–OGSA method has been improved over 3% and 6% than the existing GSA-ANN and ANN, respectively. Similarly, the GSA-ANN-based MVD system attained the maximum recall of 89%, 91%, and 91% for video 1, video 2, and video 3, respectively.

Simulating Dynamic Driving Behavior in Simulation Test for Unmanned Vehicles via Multi-Sensor Data.

Driving behavior is the main basis for evaluating the performance of an unmanned vehicle. In simulation tests of unmanned vehicles, in order for simulation results to be approximated to the actual results as much as possible, model of driving behaviors must be able to exhibit actual motion of unmanned vehicles. We propose an automatic approach of simulating dynamic driving behaviors of vehicles in traffic scene represented by image sequences. The spatial topological attributes and appearance attributes of virtual vehicles are computed separately according to the constraint of geometric consistency of sparse 3D space organized by image sequence. To achieve this goal, we need to solve three main problems: Registration of vehicle in a 3D space of road environment, vehicle’s image observed from corresponding viewpoint in the road scene, and consistency of the vehicle and the road environment. After the proposed method was embedded in a scene browser, a typical traffic scene including the intersections was chosen for a virtual vehicle to execute the driving tasks of lane change, overtaking, slowing down and stop, right turn, and U-turn. The experimental results show that different driving behaviors of vehicles in typical traffic scene can be exhibited smoothly and realistically. Our method can also be used for generating simulation data of traffic scenes that are difficult to collect.

Robust method of changes of light to detect and track vehicles in traffic scenes

Robust method of changes of light to detect and track vehicles in traffic scenes

Model-Based Localisation and Recognition of Road Vehicles

Objects are often constrained to lie on a known plane. This paper concerns the pose determination and recognition of vehicles in traffic scenes, which under normal conditions stand on the ground-plane. The ground-plane constraint reduces the problem of localisation and recognition from 6 dof to 3 dof. The ground-plane constraint significantly reduces the pose redundancy of 2D image and 3D model line matches. A form of the generalised Hough transform is used in conjuction with explicit probability-based voting models to find consistent matches and to identify the approximate poses. The algorithms are applied to images of several outdoor traffic scenes and successful results are obtained. The work reported in this paper illustrates the efficiency and robustness of context-based vision in a practical application of computer vision. Multiple cameras may be used to overcome the limitations of a single camera. Data fusion in the proposed algorithms is shown to be simple and straightforward.

Model-independent recovery of object orientations

A novel algorithm is presented for determining the orientation of road vehicles in traffic scenes using video images. The algorithm requires no specific 3-D vehicle models and only uses local image gradient values. It may easily be implemented in real-time. Experimental results with a variety of vehicles in routine traffic scenes are included to demonstrate the effectiveness of the algorithm.

Closed-form algorithms for object pose and scale recovery in constrained scenes

This paper concerns the recovery of pose and scale of vehicles in traffic scenes which, under normal conditions, are constrained to be in contact with the ground-plane. Several closed-form algorithms are described for pose and scale recovery using known 2D-to-3D line matches. The algorithms directly exploit the ground-plane constraint and are applicable to an arbitrary number of line matches. The algorithms are tested extensively with both synthetic and real outdoor traffic images. They are found to be robust and perform satisfactorily with real images.

Model-based object tracking in monocular image sequences of road traffic scenes

Moving vehicles are detected and tracked automatically in monocular image sequences from road traffic scenes recorded by a stationary camera. In order to exploit the a priori knowledge about shape and motion of vehicles in traffic scenes, a parameterized vehicle model is used for an intraframe matching process and a recursive estimator based on a motion model is used for motion estimation. An interpretation cycle supports the intraframe matching process with a state MAP-update step. Initial model hypotheses are generated using an image segmentation component which clusters coherently moving image features into candidate representations of images of a moving vehicle. The inclusion of an illumination model allows taking shadow edges of the vehicle into account during the matching process. Only such an elaborate combination of various techniques has enabled us to track vehicles under complex illumination conditions and over long (over 400 frames) monocular image sequences. Results on various real-world road traffic scenes are presented and open problems as well as future work are outlined.