Image - Based Sea Surface Roughness Analysis

Sea surface roughness is a physical parameter about surface roughness of the sea, which mainly describes the heave of sea on the small scale, the degree of sea waves determines the wave height, and thus affects the ocean wave energy, so the change of the surface roughness represents the process of energy transfer between ocean and atmosphere in a large extent. The study of the parameter extraction has very important significance to physical oceanographic research. Many calculation methods of the surface roughness have been achieved, but the existing research is mainly based on marine microwave remote sensing technology using the data obtained by microwave scatterometer, satellite altimeter or some other measurement equipment to calculate the surface roughness.First of all, this paper introduces the basic concept of image texture, and then on this basis, it does some brief analysis about the mainstream techniques of image analysis including structure analysis, statistics method, model method, signal processing method, application field. So, it is clear in this paper that the method based on statistics is more suitable for processing the surface images.Then, the paper selects five different statistical methods to extract the surface roughness of the sea image, including gray gradient co-occurrence matrix, autocorrelation function, image autocorrelation method based on FBM’s brightness difference, edge frequency method based on the distance, and Tamura texture feature, respectively calculating the energy moment, second moment and variance estimation index, maximum fringe frequencies, averages of the optimal size. These parameters can reflect the roughness of the image.Finally, the paper calculates the surface roughness of sea image according to empirical formulas between the sea surface roughness and the wind speed proposed by Xiong Kang. This is a dependent variable and the parameters representing image roughness in each method are treated as independent variables. And then the least squares support vector machine (LSSVM) is used to do some regression and fitting analysis, so we can get the regression model of each method.All in all, from the perspective of offshore video image, this paper studies the precise relationship between the roughness of air-sea dynamic and video images, and establish a new type of surface roughness parameterization scheme on the basis of image texture analysis and feature extraction technique.