Video-Based Vehicle Counting for Expressway: A Novel Approach Based on Vehicle Detection and Correlation-Matched Tracking Using Image Data from PTZ Cameras

Abstract

Vehicle counting plays a significant role in vehicle behavior analysis and traffic incident detection for established video surveillance systems on expressway. Since the existing sensor method and the traditional image processing method have the problems of difficulty in installation, high cost, and low precision, a novel vehicle counting method is proposed, which realizes efficient counting based on multivehicle detection and multivehicle tracking. For multivehicle detection tasks, a construction of the new expressway dataset consists of a large number of sample images with a high resolution (1920 × 1080) captured from real-world expressway scenes (including the diversity climatic conditions and visual angles) by Pan-Tilt-Zoom (PTZ) cameras, in which vehicle categories and annotation rules are defined. Moreover, a correlation-matched algorithm for multivehicle tracking is proposed, which solves the problem of occlusion and vehicle scale change in the tracking process. Due to the discontinuity and unsmooth of the trajectories that occurred during the tracking process, we designed a trajectory optimization algorithm based on least square method. Finally, a new vehicle counting method is designed based on the tracking results, in which the driving direction information of the vehicle is added in the counting process. The experimental results show that the proposed counting method in this research can achieve more than 93% accuracy and an average speed of 25 frames per second in expressway video sequence.

Dataset of Annotated Virtual Detection Line for Road Traffic Monitoring

“…Moving object detection using video surveillance systems [1] is an important research area for various computer vision applications, where it plays an important role in intelligent video surveillance [2,3], traffic monitoring [4], and pedestrian detection [5]. Although moving object detection methods have been extensively studied to achieve higher detection performance, i.e., accuracy [6], since videos from surveillance cameras are captured in an uncontrolled environment, the performance may be degraded due to illumination problems, complex backgrounds, occlusions, moving shadows, unpredictable motion, changes in the appearance of moving objects, and camera problems.…”

An Optimal Deep Learning for Cooperative Intelligent Transportation System

“…In addition, frame variation model has been influenced by vehicle speed as well as the time period of prominent frames. Moreover, optical flow technology is defined as pixel-level density evaluation which is not applicable for practical domains because of the huge computation [7]. Recently, the effect of solving complex scenarios to gain précised target prediction, Machine Learning (ML) models and classifiers are employed extensively prior to applying Deep Learning (DL) which is considered as the major stream of computer vision.…”

A Traffic Event Detection Method Based on Random Forest and Permutation Importance

“…The vehicle detector uses a video camera as a sensor, and sets a virtual coil in the field view. When the vehicle enters the detection area, the background gray value will change, so as to know the existence of the vehicle and detect the density and speed of the traffic flow [4][5][6][7]. Detector can be installed above the traffic lane or roadside.…”

“…Events and events not occurring in essence can be summarized as the classification problem in machine learning category. Therefore, many classification algorithms in the field of artificial intelligence can be introduced into the application, and improve efficiency and precision [7,8].…”