Image-Based Learning to Measure the Space Mean Speed on a Stretch of Road without the Need to Tag Images with Labels.

Jincheol Lee,Johyun Shin,Seungbin Roh,Keemin Sohn

doi:10.3390/s19051227

Abstract

Space mean speed cannot be directly measured in the field, although it is a basic parameter that is used to evaluate traffic conditions. An end-to-end convolutional neural network (CNN) was adopted to measure the space mean speed based solely on two consecutive road images. However, tagging images with labels (=true space mean speeds) by manually positioning and tracking every vehicle on road images is a formidable task. The present study was focused on naïve animation images provided by a traffic simulator, because these contain perfect information concerning vehicle movement to attain labels. The animation images, however, seem far-removed from actual photos taken in the field. A cycle-consistent adversarial network (CycleGAN) bridged the reality gap by mapping the animation images into seemingly realistic images that could not be distinguished from real photos. A CNN model trained on the synthesized images was tested on real photos that had been manually labeled. The test performance was comparable to those of state-of-the-art motion-capture technologies. The proposed method showed that deep-learning models to measure the space mean speed could be trained without the need for time-consuming manual annotation.

Highlights

It is difficult to use existing traffic surveillance systems to directly measure the traffic density and space mean speed
Chung et al [1] introduced a convolutional neural network (CNN) that can be used to measure traffic density based solely on a road image. This shows that the space mean speed could be measured from two consecutive photos taken over a short time interval
A YOLO model recorded the second worst performance because it failed to detect crowded vehicles. This does not mean that a YOLO is an unacceptable model for every case

Summary

Introduction

It is difficult to use existing traffic surveillance systems to directly measure the traffic density and space mean speed. Chung et al [1] introduced a convolutional neural network (CNN) that can be used to measure traffic density based solely on a road image. This shows that the space mean speed could be measured from two consecutive photos taken over a short time interval. For a given stretch of road the most definite method to measure the space mean speed for a given moment is to average the instantaneous speeds of all the vehicles (see Figure 1) If this measurement is possible, a profile of the space mean speeds along a time axis could be obtained, which is the most accurate indicator of the traffic dynamics of a road segment. There currently is no direct method to observe such a speed profile based on spot detectors that prevail in traffic surveillance

Objectives

Methods

Results

Conclusion