Clustering learning model of CCTV image pattern for producing road hazard meteorological information

Abstract

The method of real-time estimation of weather, especially the amount of rainfall, by analyzing CCTV images is much cheaper than one using the existing expensive weather observation equipment. In this paper, we propose a method to find an estimation model function which has its input as CCTV images and output as the amount of rainfall. From the CCTV images, we propose an algorithm for selecting the number and size of the region of interest optimized for rainfall estimation, generating a data pattern graph showing a clear distinction from the number of region of interest, clustering the pattern data graphs, and estimating the amount of rainfall. Experiments using real CCTV images show that the estimation accuracy is over 80%.

Understanding the potential of emerging digital technologies for improving road safety

“…The weather-related studies involve a wider range of technologies compared to other categories, mainly because of the relatively higher interdependency with other contributing factors. These include weather monitoring [99][100][101][102], travelling condition [103], modelling and projection of road condition related to the weather [104,105], rainfall estimation using CCTV images [106], impacts of weather on the operation of autonomous vehicles [107], driver warning systems [108,109], public warning systems [110] and enhancing road weather management during extreme weather events [111]. A study of particular interest that has potential for immediate uptake is the approach developed by Lee, Hong [106] employing image processing and machine learning to detect and estimate rainfall from CCTV images in real-time and cluster the regions.…”

Evaluation of Clustering Algorithms on HPC Platforms

“…Individuals from a cluster are closer to its centre than to any other given cluster centre [13]. Clustering algorithms are critical to analyse datasets in many fields, including image processing [14], smart cities [15,16] or bioinformatics [17]. The most relevant clustering algorithms have the following characteristics: (1) scalability, i.e., theability to handle an increasing amount of objects and attributes; (2) stability to determine clusters of different shape or size; (3) autonomy or self-driven, i.e., it should require minimal knowledge of the problem domain (e.g., number of clusters, thresholds and stop criteria); (4) stability, as it must remain stable in the presence of noise and outliers; and, finally, (5) data independency to be independent of the way objects are organised in the dataset.…”

Mobile phone data statistics as a dynamic proxy indicator in assessing regional economic activity and human commuting patterns

“…Mobile phone data have been used for transport planning (Elias, Nadler, Stehno, Krosche, & Lindorfer, ; Liu et al, ), traffic measurement (Dong et al, ; Hongyan & Fasheng, ; Steenbruggen, Tranos, & Rietveld, ) and modelling (Oliveira, Viana, Naveen, & Sarraute, ), trajectory evaluation (Bonnel, Hombourger, Olteanu‐Raimond, & Smoreda, ; Chen, Bian, & Ma, ), and travel time predictions (Woodard et al, ). The general problems in urban planning (Jonge, Pelt, & Roos, ; Ricciato, Widhalm, Pantisano, & Craglia, ) and analysis (Lee et al, ), land use (Ríos & Muñoz, ), and smart city development (Steenbruggen, Tranos, & Nijkamp, ) can all be addressed with the help of mobile phone data analysis.…”