A method for extracting salient surface specific targets in remote sensing images

Abstract

The invention relates to a method for extracting a salient surface specific target in a remote sensing image, which is characterized in that it comprises the following steps: 1) segmenting the image; 2) selecting a small number of specific target samples, by estimating high-dimensional normal distribution parameters and enhancing transformation operations , to achieve threshold extraction of potential specific target area P; 3) Using the difference of feature values ​​of the object and its adjacent objects, the threshold extracts the salient area S; 4) The intersecting area K of P and S is used as the seed, and the combination of P and S The area G is the area to be grown, and the area growth is carried out with the object as the unit, and the preliminary target extraction result T is obtained. 0 ; 5) Through geometrical, morphological and other features, for T 0 The non-target area is further excluded to obtain the final planar specific target area. The method of the invention can quickly and effectively extract the seed object of a specific target area, and solves the problem of blurred edges when extracting prominent targets; it is suitable for automatic extraction of specific targets that emphasize area attributes, such as artificial cultivation areas, in remote sensing images.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, in particular to a method for extracting a salient surface specific target in a remote sensing image. Background technique [0002] Specific targets closely related to human activities, such as mining mines, ginseng cultivation areas, etc., present visually salient features in remote sensing images. In addition, this type of target is usually presented as a surface, and its area information has important reference value for ecological assessment, management, and related market assessment. Timely and accurate automatic acquisition of the area and spatial distribution of these prominent planar specific targets from remote sensing images is a very meaningful task compared to traditional field and market surveys with high cost and poor timeliness. [0003] Existing saliency-specific object extraction mainly includes supervised classification and pixel-based visual attention models. ...

AI Technical Summary

Problems solved by technology

Its disadvantage is that in sample selection, not only positive samples belonging to a specific target need to be included, but also negative samples that do not belong to a specific target need to be included. In practice, the types of negative samples are very complicated, which leads to the cost of this type of method in practical applications. high

This type of processing method has two main problems: (1) There is no distinction for salient objects, that is, any visually salient object will be extracted; (2) The limited pixel-based salient feature map calculation method leads to blurred edges , making the area less accurate

Retrieval of Chinese and foreign literature, this method has not been used in the prior art to extract the specific target of the salient surface

Images