Abstract: In the survey of underwater microtopography, due to the influence of such factors as reverberation, noises disturbance, characteristics of bottom materials and gradient, there can be some exceptional altimetric values of microtopography in the computation result even if an accurate detection of echo signals is made by ultrasonic wave amplitude-related method, and the true landform is hard to characterize by digital microtopography reconstructed by the conventional least square method. Therefore, the algorithm of fuzzy support vector regression with the most interrelated coefficient of dispersion echo as accuracy factor was established, which, based on a strict mathematical theory, has great application potential in data regression, especially for small samples. The comparative experiments indicate that the algorithm's regression line can effectively eliminate the effect of exceptional values, so as to well reflect the exterior figure of actual microtopography.   

Key words: Microtopography, Ultrasonic wave, Least square method, Dispersion echo, Fuzzy support vector