Content area
Full Text
Abstract-Network attached cameras, or webcams, provide a relatively inexpensive source of visual information for Internet users. In recent years, live webcams have been used to monitor active regions of cities to allow users to view sites of interest. Live traffic images can be used for a more detailed analysis as well as for creating useful decision support systems using statistical learning techniques. The data set size produced by a streaming webcam grows rapidly, and methods for effectively summarizing the key traits of traffic images becomes increasingly important for reducing the overall data set size. This paper describes a feature extraction system for traffic webcams images, that represents an individual image using attributes based on three criteria that are important to drivers: weather, lighting, and traffic conditions. The processed data set is summarized by the use of unsupervised clustering to demonstrate the ability of the proposed feature extraction step.
Keywords-feature extraction, activity detection, image analysis
(ProQuest: ... denotes formulae omitted.)
I. INTRODUCTION
The image stream provided by a traffic webcam a good source for data mining and predictive analytics. This paper uses the webcam traffic images from MacKay bridge of Nova Scotia, Canada. The MacKay bridge connects major metropolitan areas within the city of Halifax, and it is one of the most heaviest traffic areas in the region where thousands of vehicles cross the bridge each day. The bridge is an important determinant of the travel time for a large number of commuters. However, while a raw visual depiction of the traffic image is more meaningful to a user than textual or numeric summary, browsing through such a large sequence of images for discovering any meaningful patterns is a difficult task for an individual person or a computer application. In order to develop useful tools based on the images obtained from such a webcam, methods for summarizing the key traits from a traffic image is important to quantify the weather, traffic, and lighting conditions.
Therefore, it is essential to process the images and extract relevant information in order to summarize traffic data sets for easier browsing and further analysis. A number of researchers have proposed image processing techniques that make it possible to describe various features of these images [1]. This paper describes the...