The Surface Roughness Analysis Based On Image Sequence

Sea surface roughness, as an important parameter in the research of marine physics, is a physical variable that measures the degree of roughness for micro scale sea surface, whose changing trend reflects on a large part the characteristics of energy exchange between the sea and the atmosphere. Micro wave remote sensing technology is mainly used in the study of sea surface roughness. Satellite altimeter and microwave Scatterometer are the devices for gathering information such as Sea surface wind field so as to calculate the sea surface roughness.In the dissertation, sea surface roughness is first defined from the definition of aerodynamics, followed by introduction to its research status. Then, image characteristic parameters are analyzed from the angle of fractal theory and texture feature to obtain parameters like fractional order and texture feature of see image. A total of six methods are used to calculate fractional order, and an improved fractal dimension algorithm is proposed to overcome the limitation of conventional one, making the result closer to true value. The six methods for calculation fractional order are compared and the best one, spectrum method, is chosen to compute fractal dimension of the sea.Gray level co-occurrence matrix (GLCM) is employed to derive texture characteristics of sea image. Parameters of the image such as second-order angular moment, moment of inertia, entroy and inverse difference moment are calculated to reflect texture feature of the image.For the date fitting, according to the relation between sea surface roughness and10m air speed presented by Xiong Kang in1990, sea surface roughness or ChangJiang rive roughness is computed based on measurement of air speed, and the corresponding fractal dimension and second order angular moment are also recorded. With fractal dimension and second order angular moment being the arguments, sea surface or river roughness being dependent variable, the optimal fitting model can be selected by feeding preciously recorded data into LS-SVM regression model, for analysis. From the model, the relation between sea surface roughness and fractal dimension, second order angular moment of texture is known.In conclusion, an approach that calculates sea surface roughness from texture parameters of image is proposed by studying the relationship between gray-level feature of image and ocean spatial information, providing reference and guide for computing characteristic parameters about the sea from the angle of image analysis.