Driving Through Graphs: a Bipartite Graph for Traffic Scene Analysis

Automatic symbolic traffic scene analysis is essential to many areas of IVHS (Intelligent Vehicle Highway Systems). Traffic scene information can be used to optimize traffic flow during busy periods, identify stalled vehicles and accidents, and aid the decision-making of an autonomous vehicle controller. Improvements in technologies for machine vision-based surveillance and high-level symbolic reasoning have enabled the authors to develop a system for detailed, reliable traffic scene analysis. The machine vision component of the system employs a contour tracker and an affine motion model based on Kalman filters to extract vehicle trajectories over a sequence of traffic scene images. The symbolic reasoning component uses a dynamic belief network to make inferences about traffic events such as vehicle lane changes and stalls. In this paper, the authors discuss the key tasks of the vision and reasoning components as well as their integration into a working prototype. Preliminary results of an implementation on special purpose hardware using C-40 Digital Signal Processors show that near real-time performance can be achieved without further improvements.

Concerns automatic traffic scene analysis. Major improvements in performance and quality of results of machine vision based traffic surveillance systems allow connections to symbolic reasoning components that attain a high level of accuracy and reliability. We apply an approach for detecting and tracking vehicles in road traffic scenes using an explicit occlusion reasoning step. We represent moving vehicles by closed contours and employ a contour tracker based on intensity and motion boundaries. Motion and contour estimation is performed by linear Kalman filters based on an affine motion model. Occlusion detection is performed by intersecting the depth ordered regions associated to the objects. The intersection is then excluded in the motion and shape update. A contour associated to a moving region is initialized using a motion segmentation step which is based on differences between filter outputs of an acquired image and a continuously updated background image. Symbolic reasoning of the traffic scene based on the extracted car tracks is performed using a belief network. Belief networks provide a flexible and theoretically sound framework for traffic scene analysis because of their inherent ability to model uncertainties. We show the validity of our approach and present results of experiments with real world traffic scenes. Preliminary results of an implementation on special purpose hardware using C-40 Digital Signal Processors show that near real-time performance can be achieved without further improvements.

Traffic scene analysis is important for emerging technologies such as smart traffic management and autonomous vehicles. However, such analysis also poses potential privacy threats. For example, a system that can recognize license plates may construct patterns of behavior of the corresponding vehicles' owners and use that for various illegal purposes. In this paper we present a system that enables traffic scene analysis while at the same time preserving license plate privacy. The system is based on a multi-task model whose latent space is selectively compressed depending on the amount of information the specific features carry about analysis tasks and private information. Effectiveness of the proposed method is illustrated by experiments on the Cityscapes dataset, for which we also provide license plate annotations.

In this paper, upon the background of driving assistance on highway, we propose a real-time vehicle detection and tracking algorithm based on traffic scene analysis. We describe a general traffic scene analysis framework for vehicle detection and tracking based on roadside detection at first. On that basis, we present a new object detection algorithm via fusion of global classifier and part-based classifier and a vehicle detection algorithm integrating classifying confidence and local shadow. The local shadow is obtained by detecting the Maximally Stable Extremal Regions (MSER) using a multi-resolution strategy. Finally, we test our algorithm on several video sequence captured from highway and suburban roads. The test results show high efficiency and robustness when coping with environment transition, illumination variation and vehicle orientation change.

This paper presents a practical system for vision-based traffic scene analysis from a moving vehicle based on a cognitive feedback loop which integrates real-time geometry estimation with appearance-based object detection. We demonstrate how those two components can benefit from each other’s continuous input and how the transferred knowledge can be used to improve scene analysis. Thus, scene interpretation is not left as a matter of logical reasoning, but is instead addressed by the repeated interaction and consistency checks between different levels and modes of visual processing. As our results show, the proposed tight integration significantly increases recognition performance, as well as overall system robustness. In addition, it enables the construction of novel capabilities such as the accurate 3D estimation of object locations and orientations and their temporal integration in a world coordinate frame. The system is evaluated on a challenging real-world car detection task in an urban scenario.KeywordsObject DetectionStereo PairScene AnalysisReconstruction ModuleScene GeometryThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The major challenges for analyzing the situation of traffic scenes include defining proper metrics and achieving computation efficiency. This paper proposes two new situation metrics, a multimodality scene model, and a metrics computing network for traffic scene analysis. The main novelty is threefold. (1) The planning complexity and perception complexity are proposed as the situation metrics of traffic senes. (2) A multimodality model is proposed to describe traffic scenes, which combines the information of the static environment, dynamic objects, and ego-vehicle. (3) A deep neural network (DNN) based computing network is proposed to compute the two situation metrics based on scene models. Using the Nuscenes dataset, a high-level dataset for traffic scene analysis is developed to validate the scene model and the situation metrics computing network. The experiment results show that the proposed scene model is effective for situation analysis and the proposed situation metrics computing network outperforms than traditional CNN methods.

Automatic traffic analysis is very important in the modern world with heavy traffic. It can be achieved in numerous ways, among them, detection and analysis through video system, being able to provide affluent information and having little disturbance to the traffic, is an ideal choice. The proposed traffic vision analysis system uses Image Acquisition Card to capture real time images of the traffic scene through video camera, and then exploits the sequence of traffic scene and the image processing and analysis technique to detect the presence and movement of vehicles. First getting rid of the complex traffic background, which is always changing, the system segment each vehicle in the region the user interested. The system extracts features from each vehicle and tracks them through the image sequence. Combined with calibration, the system calculates information of the traffic, such as the speed of the vehicles, their types, the volume of flow, the traffic density, the waiting length of the lanes, the turning information of the vehicles, and so. Traffic congestion and vehicles’ shadows are disturbing problems of the vehicle detection, segmentation and tracking. So we make great effort to investigate on methods to dealing with them. Keywords: Traffic analysis, Object Tracking, Vehicle Detection

A model based approach to monocular image sequence analysis of road traffic scenes is presented. Within this framework a vision system for applications like autonomous driving and collision avoidance was developed. The approach takes part in problems of selective and active vision. The fully automatic system MOSAIK recognizes and describes all visual vehicles on or near the road. It solves the problem to compute a robust scene description under egomotion nearly in realtime on a standard monoprocessor workstation. MOSAIK has been tested by using typical German 'Autobahn' and road scenes. This paper describes the vision approach and the interaction of vehicle recognition and tracking and the influence of attention control.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Analyzing motion patterns in traffic videos can be exploited directly to generate high-level descriptions of the video contents. Such descriptions may further be employed in different traffic applications such as traffic phase detection and abnormal event detection. One of the most recent and successful unsupervised methods for complex traffic scene analysis is based on topic models. In this paper, a two-level Sparse Topical Coding (STC) topic model is proposed to analyze traffic surveillance video sequences which contain hierarchical patterns with complicated motions and co-occurrences. The first level STC model is applied to automatically cluster optical flow features into motion patterns. Then, the second level STC model is used to cluster motion patterns into traffic phases. Experiments on a real world traffic dataset demonstrate the effectiveness of the proposed method against conventional one-level topic model based methods. The results show that our two-level STC can successfully discover not only the lower level activities but also the higher level traffic phases, which makes a more appropriate interpretation of traffic scenes. Furthermore, based on the two-level structure, either activity anomalies or traffic phase anomalies can be detected, which cannot be achieved by the one-level structure.

This paper presents the development of a vision-based system for microscopic road traffic scene analysis and understanding using computer vision and computational intelligence techniques. The traffic flow model is calibrated using the information obtained from the road-side cameras. It aims to demonstrate an understanding of different levels of traffic scene analysis from simple detection, tracking, and classification of traffic agents to a higher level of vehicular and pedestrian dynamics, traffic congestion build-up, and multi-agent interactions. The study used a video dataset suitable for analysis of a T-intersection. Vehicle detection and tracking have 88.84% accuracy and 88.20% precision. The system can classify private cars, public utility vehicles, buses, and motorcycles. Vehicular flow of every detected vehicles from origin to destination are also monitored for traffic volume estimation, and volume distribution analysis. Lastly, a microscopic traffic model for a T-intersection was developed to simulate a traffic response based on actual road scenarios.

Homography has wide applications in aerial photographic surveys, camera calibration, traffic scene analysis, and sports science, such as player and team performance evaluation. Unlike the mainstream homography that utilizes points as matching features, homography estimation for sports and traffic video can achieve higher accuracy and speed by utilizing straight lines in the scenes, which convey more information than points. Owing to the more stringent requirement of accuracy and computational speed for advanced video analysis, this paper presents a novel homography computational algorithm. Three major novelties are proposed and validated, which are multiple points Hough transform for straight line extraction, correspondence initialization by angle to estimate a set of quasi-optimal solutions, and the feature correspondences optimization to achieve a minimized error using genetic algorithm. With these contributions, the experiments have shown that the proposed algorithm can improve the homography computational accuracy by up to 130% and reduce the processing time by up to 96% over the state-of-the-art algorithms for the same purposes.

Analyzing motion patterns in traffic videos can directly generate some high-level descriptions of the video content which can be further employed in rule mining and abnormal event detection. The most recent and successful unsupervised approaches for complex traffic scene analysis are based on topic models. However, most existing topic models share some key characteristics which could limit their utility. In this paper, based on extracted optical flow features from video clips, we employ Sparse Topical Coding (STC) framework to automatically discover typical motion patterns in traffic scenes. For this purpose, we improve the STC to overcome one of the drawbacks of topic models with the aim of learning the semantic traffic motion patterns. We go beyond the usual word-document paradigm in topic models by taking into account the order of optical flow words during learning. Experimental results show that our proposed method can learn better motion patterns to analyse the traffic video scenes.

The Government of India develops the road and transport facilities to reduce traffic congestion in mass gatherings at rush time. Based on the public needs, the video scene detection system provides a video classification technique to analyze the video sequences from various crowd and traffic congestion excellently. This classification technique is most significant in many applications such as video surveillance, traffic control analysis and crowd monitoring. A number of applications for detecting and classifying videos have been proposed in the literature. Scene analysis in the video is a big challenge in many Crowd monitoring and Traffic controlling system. The proposed system presents an automatic video scene detection and analysis method for detecting and classifying crowd and traffic video scenes. Histogram Oriented Gradients (HOG) feature descriptor is extracted features from the video scenes. K Nearest Neighbour (KNN) and Support Vector Machine (SVM) classifiers are used to classify the video scenes. The video scenes are collected from various crowd videos of Tamil Nadu. The experimental results show that KNN with HOG features performs well with 97% accuracy.

A Parallel Pipeline Based Multiprocessor System For Real-Time Measurement of Road Traffic Parameters

A new representation of scene layout improves saliency detection in traffic scenes