A model-driven approach to estimate atmospheric visibility with ordinary cameras

Abstract

Atmospheric visibility is an important input for road and air transportation safety, as well as a good proxy to estimate the air quality. A model-driven approach is presented to monitor the meteorological visibility distance through use of ordinary outdoor cameras. Unlike in previous data-driven approaches, a physics-based model is proposed which describes the mapping function between the contrast in the image and the atmospheric visibility. The model is non-linear, which allows encompassing a large spectrum of applications. The model assumes a continuous distribution of objects with respect to the distance in the scene and is estimated by a novel process. It is more robust to illumination variations by selecting the Lambertian surfaces in the scene. To evaluate the relevance of the approach, a publicly available database is used. When the model is fitted to short range data, the proposed method is shown to be effective and to improve on existing methods. In particular, it allows envisioning an easier deployment of these camera-based techniques on multiple observation sites.

Computer vision for the remote sensing of atmospheric visibility

“…of there being an object at the distance d in the scene. To compute m, a realistic expression for the density of objects ψ in the scene is needed, as well as a map of the albedo of the Lambertian surfaces [2].…”

“…In the field of atmospheric environment, a linear fitting of data is generally performed, but such approaches are not valid anymore in case of shorter visibility ranges. Recently, a model-driven approach has been proposed [2]. The calibration of the method is made easier by introducing a physics-based model and a simplified depth distribution model of the scene.…”

“…First, Koschmieder's theory on the visual appearance of objects in the atmosphere is recalled as well as the model-driven approach proposed in [2] to estimate the visibility by camera. Second, our method of Lambertian surfaces selection is presented and justified.…”

Meteorological Visibility Evaluation on Webcam Weather Image Using Deep Learning Features

“…In this method, features related to the extinction coefficient or visibility distance are extracted from the image to evaluate the visibility based on the meteorological law. For example, Babari et al [7]- [9] proposed a physics-based model to express the mapping between the contrast and the atmospheric visibility distance with the consideration of Lambertian surfaces of the scene. Carretas et al [10] used two tree sets located at different but known distances to calculate the extinction coefficient based on the different wavelengths associated with R,G,B channels of the image to get the visibility.…”

“…Enough targets are needed for the measurement range and accuracy in this method. The third image method is according to the law in meteorology, such as Koschmieder equation or Lambert-beer law [7]- [15]. In this method, features related to the extinction coefficient or visibility distance are extracted from the image to evaluate the visibility based on the meteorological law.…”