LS-SVM algorithm sounding training sample thinning method based on sample Euclidean distance

The invention relates to an LS-SVM algorithm sounding training sample thinning method based on a sample Euclidean distance. The method comprises the following technical characteristics: selecting a sounding training sample used for constructing a seabed trend surface function; determining the sample center of the sounding training sample, and calculating the Euclidean distance from the sounding training sample to the sample center one by one; adopting different Euclidean distances to carry out first-time and second-time thinning on the sounding training sample, comparing mean square errors of the first-time thinning and the second-time thinning to judge the precision index of the seabed trend surface function model, and determining the threshold value of the Euclidean distance on the premise that the precision of the seabed trend surface function model is ensured; and thinning a sounding training sample by using an Euclidean distance threshold, and adjusting combined parameters of an LS-SVM algorithm to const