High resolution remote sensing image analysis tree automatic extraction method based on space consistency

Abstract

The invention discloses a high resolution remote sensing image analysis tree automatic extraction method based on space consistency and relates to the remote sensing image processing technology. The method includes the following steps: S1) building various complex target image representational sets; S2) adopting an over-segmentation algorithm to obtain different image objects; S3) conducting statistics on multi-dimensional characteristics of the objects such as color, grain, shape and the like; S4) acquiring class attribute of the objects combined with an implied probability semantic model; S5) selecting consistent object based on space consistency; and S6) calculating a similar matrix of the consistent objects and finally generating an analysis tree structure of the image scene through a split combination algorithm. The method highlights important target information in the scene by analyzing space geological arrangement, acquires an analysis tree of the scene, is favorable for improving semantic expressing accuracy and improves target recognition and scene classification robustness.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, and relates to a method for extracting a resolution tree in a high-resolution remote sensing image based on spatial consistency. Background technique [0002] Any image scene can be represented in the form of a parse tree. A parse tree is a comprehensive description of an image or a structural description of the target objects contained in the image. It includes: the attributes and names of the objects contained in the image, the relationship between each object, and the relationship between each object and the image scene formed by them, etc. The parse tree can not only use terminal nodes and non-terminal nodes to represent the hierarchical decomposition from scene to object, part, primitive and pixel, but also use horizontal connections to represent the spatial semantic association between nodes. [0003] In high-resolution remote sensing images, scenes and some comple...

AI Technical Summary

Problems solved by technology

[0004] In recent years, researchers have proposed many ideas for extracting parse trees: some methods mainly analyze factors such as the environment, scene angle, and geometric consistency of the target in a single image. The ability to describe images and targets in a single way, but is greatly affected by sample images and artificial prior assumptions; the other method is based on the image set, focusing on the analysis of dependencies between different parts of the target, this type of method relies too much on manual Image segmentation or labeling, due to the lack of consideration of the relative position between image segments or target parts, the stability of the parse tree is not good when the same target is deformed

Images