Bangkok CCTV Image through a Road Environment Extraction System Using Multi-Label Convolutional Neural Network Classification

Chairath Sirirattanapol,Apichon Witayangkurn,Masahiko Nagai,Mongkol Ekpanyapong,Surachet Pravinvongvuth

doi:10.3390/ijgi8030128

Abstract

Information regarding the conditions of roads is a safety concern when driving. In Bangkok, public weather sensors such as weather stations and rain sensors are insufficiently available to provide such information. On the other hand, a number of existing CCTV cameras have been deployed recently in various places for surveillance and traffic monitoring. Instead of deploying new sensors designed specifically for monitoring road conditions, images and location information from existing cameras can be used to obtain precise environmental information. Therefore, we propose a road environment extraction framework that covers different situations, such as raining and non-raining scenes, daylight and night-time scenes, crowded and non-crowded traffic, and wet and dry roads. The framework is based on CCTV images from a Bangkok metropolitan dataset, provided by the Bangkok Metropolitan Administration. To obtain information from CCTV image sequences, multi-label classification was considered by applying a convolutional neural network. We also compared various models, including transfer learning techniques, and developed new models in order to obtain optimum results in terms of performance and efficiency. By adding dropout and batch normalization techniques, our model could acceptably perform classification with only a few convolutional layers. Our evaluation showed a Hamming loss and exact match ratio of 0.039 and 0.84, respectively. Finally, a road environment monitoring system was implemented to test the proposed framework.

Highlights

We propose an application framework to test our model, and we expect that it can be implemented in many different locations, including developing countries, which have a large number of circuit television (CCTV) images but lack specific sensors
Insofar as existing CCTV cameras can be utilized to extract road environment information such as weather data, traffic, and road surface conditions, all of which are vital to drivers, we proposed a framework to extract this information by applying multi-label convolutional neural network (CNN) classification with a Bangkok Metropolitan Administration (BMA) dataset
Our results indicated that the former strategy performed slightly better, grouping opposite classes was less time consuming when preparing the dataset. Evaluation indices such as the Hamming loss, exact match accuracy, and mean average precision (MAP) were measured in addition to the training and prediction times

Summary

Introduction

Weather conditions can cause low visibility, reduce pavement friction, and impact driver behavior and performance [1]. Such events could be acceptably detected using embedded and non-invasive sensors [2] to help prevent accidents by providing an early warning system to drivers. Such sensors typically incur high installation costs and, with embedded sensors, suitable locations for placement. Closed circuit television (CCTV) is common in many places. The common goals of such systems include crime prevention and detection by tracking and observation

Methods

Results

Conclusion