A variational approach to atmospheric visibility estimation in the weather of fog and haze. (May 2018)

• Record Type: Journal Article
• Authors: Cheng, Xiaogang; Yang, Bin; Liu, Guoqing; Olofsson, Thomas; Li, Haibo
• Is Part Of: Sustainable cities and society. Volume 39(2018)
• Page Start: 215
• Page End: 224
• Publication Date: 2018-05
• Subjects: Atmospheric visibility estimation -- Variational approach -- Piecewise stationary time series -- Computer vision -- Fog and haze; Sustainable urban development -- Periodicals; Sustainable buildings -- Periodicals; Urban ecology (Sociology) -- Periodicals; 307.76
• DOI: 10.1016/j.scs.2018.02.001
• Languages: English
• ISSNs: 2210-6707

Abstract

Graphical abstract: Highlights: Foggy and hazy visibility was studied and the surveillance video was collected. Variation approach and piecewise stationary time series were used. The extinction coefficient ( k ) was treated as a time dependent function ( k(t) ). Koschmieder's formula and corresponding luminance curve model were validated with a big video dataset. Abstract: Real-time atmospheric visibility estimation in foggy and hazy weather plays a crucial role in ensuring traffic safety. Overcoming the inherent drawbacks with traditional optical estimation methods, like limited sampling volume and high cost, vision-based approaches have received much more attention in recent research on atmospheric visibility estimation. Based on the classical Koschmieder's formula, atmospheric visibility estimation is carried out by extracting an inherent extinction coefficient. In this paper we present a variational framework to handle the nature of time-varying extinction coefficient and develop a novel algorithm of extracting the extinction coefficient through a piecewise functional fitting of observed luminance curves. The developed algorithm is validated and evaluated with a big database of road traffic video from Tongqi expressway (in China). The test results are very encouraging and show that the proposed algorithm could achieve an estimation error rate of 10%. More significantly, it is the first time that the effectiveness of Koschmieder's formula in atmospheric visibility estimation was validated with a big dataset, which contains more than two million luminance curves extracted from real-world traffic video surveillance data.