Optical remote sensing image cloud judging result correcting method based on context

Abstract

The invention discloses an optical remote sensing image cloud judging result correcting method based on context. No probabilistic model is need building, the design is carried out aiming at cloud judging problems, and judging results and background information of adjacent modules are used at the same time to correct an original judging result. The optical remote sensing image cloud judging result correcting method based on the context includes the steps: firstly, reading in related data to be processed, wherein the related data to be processed comprises a partitioned index matrix, a partitioned gray mean matrix and background information of an original image; secondly, generating a judging value of Block by using a judging value of Tile, composing a Block index matrix, and constructing a Block gray mean matrix according to the partitioned gray mean matrix; thirdly, utilizing the Block with the judging value of pretty sure cloud and pretty sure ground features to correct the Block judging value which is uncertainty; fourthly, using target context and scene context to judge background of a processed object, to reject false-alarm at the junction of ground features, to reject false-alarm of cloudlet and adjust an adjacent target of the object which is judged to be certainty, and finally whether the Block being cloud or not is judged out.

Description

technical field [0001] The invention relates to a context-based correction method for cloud judgment results of optical remote sensing images, and belongs to the technical field of pattern recognition. Background technique [0002] At present, the most widely used model in computer vision is Mary's visual computing theory. Although it has greatly advanced the development of computer vision research, it is largely considered to be incomplete, mainly because it does not include the selective and holistic aspects of human vision. The so-called selectivity refers to the top-down processing method under the guidance of the context, while the integrity refers to the overall priority strategy. [0003] At present, there have been many studies drawing on some viewpoints of human vision to propose context-based target detection and recognition methods. [0004] Among them, context can be defined as all information related to the target but not the apparent description of the target...

AI Technical Summary

Problems solved by technology

First of all, this increases the difficulty of the algorithm in terms of hardware implementation; secondly, it does not make full use of the very important spatial domain information in the image; finally, these are general-purpose algorithms, and there is no particularity for cloud judgment problems

Images