Extraction of raft cultivation of remotely sensed high-spatial-resolution images based on LBV and wavelet transforms

Abstract

Abstract A method to extract information on raft cultivation based on the LBV (The general radiance level L, visible-infrared radiation balance B, and band radiance variation vector V) transform and wavelet multi-scale decomposition was proposed. This paper selects the spectral information of typical objects and uses regression analysis to solve for the coefficients of regression. The LBV transformation of remotely sensed high-spatial-resolution imagery was derived according to the spatial characteristics of the object space, color space, and LBV variable space to obtain the B band that can be used to identify maritime targets. Multi-scale wavelet decomposition was then used to calculate the eigenvalues of the components of each sub-image by using different windows to analyze the texture features of the maritime objects, and to explore the rules of texture extraction of raft cultivation from high-spatial-resolution images. Experiments on remotely sensed ZY-3 images from Liandao Island show the following: (1) compared with the original image, the LBV-transformed image was clearer and finer details were visible in it. They helped identify and extract ground objects. (2) The eigenvalues extracted by wavelet decomposition were beneficial for the segmentation of the image. As the dimensionality of the eigenvalue increased, the accuracy of raft cultivation extraction showed an upward trend. (3) The proposed method is applicable to texture feature extraction owing to its clear rules and directionality.