| abstract |
The invention relates to a dimensionality reduction method of a hyperspectral image based on semi-supervised learning, the method comprising the following steps: constructing an improved semi-supervised similarity weight matrix Q; calculating the diagonal matrix D * of the similarity weight matrix and Laplacian Matrix L * ; Construct an improved objective function according to the semi-supervised similarity weight matrix; Solve the generalized characteristic equation according to the objective function; Solve the transformation matrix A=(a 1 , a 2 ,..., a d ) and the low-dimensional subspace Y=A T X = {y 1 ,y 2 ,...,y N }. Its advantages are as follows: using the class label information of samples and considering the neighborhood information between sample points, it can minimize the distance between samples of the same type and maximize the distance between samples of different types, thereby improving the classification accuracy of samples. |