Neural-edge-based vehicle detection and traffic parameter extraction

Abstract

Vehicle detection is a fundamental component of image-based traffic monitoring system. In this paper, we propose a neural-edge-based vehicle detection method to improve the accuracy of vehicle detection and classification. In this method, the feature information is extracted by the seed-filling-based method and is presented to the input of neural network for vehicle detection and classification. The neural-edge-based vehicle detection method is effective and the correct rate of vehicle detection is higher than 96%, independent of environmental conditions. Also, traffic parameters, such as vehicle count, vehicle class, and vehicle speed, are extracted via vehicle tracking method

Introduction

In recent years, laying new pavement or adding more lanes is becoming less and less feasible, thus that is no longer efficient solution for serious traffic congestion problem due to consistent increment of vehicles. One of the realistic solutions is to use the existing infrastructure more efficiently. Road traffic monitoring and control is the essential component of this solution. To monitor road traffic, it is necessary to extract traffic parameters that describe the characteristics of vehicles and their movement on the road. Vehicle counts, vehicle speed, vehicle path, flow rates, vehicle density, vehicle dimension, vehicle class and vehicle identity via the number plate are all example of useful traffic parameters. Various kinds of traffic control systems, for example, law enforcement, automatic tolls, congestion and incident detection, and increasing road capacity via automatic routing and variable speed limit, can be implemented with these traffic parameters.

Magnetic loop detector is most common method of traffic parameter extraction for traffic monitoring and control. Magnetic loops are very inexpensive and provide traffic parameters such as average speed, vehicle flow, and vehicle density. But, they have some problems as follows. First, they are very inflexible and modifications or additions require digging grooves in the road, thus producing traffic disturbances. Second, they cannot be used for more sophisticated tasks such as queue length measurement, tracking, etc. [8]. In recent years, an image-based traffic monitoring system is a remarkable alternative for magnetic loop detectors.

Current interest in the image-based traffic monitoring system is due to its ability to solve magnetic loop detector's problems described above. Also, image-based traffic monitoring systems offer a number of advantages. In addition to vehicle counts, a much larger set of traffic parameters, such as queue length, vehicle speed, vehicle class, vehicle path, etc. can be extracted. Besides, image-based traffic monitoring systems are much less disruptive to install than magnetic loops, thus they don't produce serious traffic disturbances [5]. In spite of these advantages, the vehicle detection error due to variation of ambient lighting condition, shadows and different shape or size of vehicles makes serious difficulties to extract traffic parameters [3], [4].

In this paper, we propose a neural-edge-based vehicle detection method for improvement of vehicle detection and vehicle classification accuracy. This paper starts by describing an overview of related work in Section 2 and the object region extraction method based on background subtraction is presented in Section 3. In 4 Neural-edge-based vehicle detection, 5 Traffic parameter extraction via vehicle tracking, the neural-edge-based vehicle detection method and the traffic parameter extraction via vehicle tracking is described. Experimental result and conclusion are given in 6 Experimental result, 7 Conclusion.