Remote sensing image sample migration method

Abstract

The invention discloses a remote sensing image sample migration method. The method comprises the following steps: acquiring an initial sample set and a target domain sample set; screening the initial sample set to obtain a source domain sample set X; extracting a generation factor set Zs of the source domain sample set and a generation factor set Zt of the target domain sample set, and obtaining a common generation factor set Zc of the source domain sample set and the target domain sample set, a non-common generation factor set Zs0 of the source domain sample set and a non-common generation factor set Zt0 of the target domain sample set according to the generation factor sets Zs and Zt; replacing Zs0 with Zt0 to obtain a source domain sample generation factor set Zs1 after feature migration; and generating a migration sample of a fitting target domain based on the Zs1. The target domain sample data is generated based on the low-error source domain sample data, sample migration is achieved, the utilization rate of the source domain samples is increased, and the problem that the target domain samples are insufficient is solved.

Description

Technical field [0001] The present invention relates to the field of remote sensing image processing, and more particularly to a remote sensing image migration method. Background technique [0002] With the rapid progress and development of satellite technology, multiple remote sensing satellite successively launched. These remote sensing satellite all day offers a large number of remote sensing image data to meet the needs of various information applications on massive data, and promote the overall development of remote sensing technology. However, how to effectively use a massive remote sensing image data is still the topic that needs to be solved urgently. [0003] In remote sensing imaging classification, the label of remote sensing image samples is the premise of efficient use of remote sensing image data. Remote sensing imaging samples are also inevitable in sample data of manual operation, high engineering, high cost, difficult to implement, artificial or migratory sample ...

AI Technical Summary

Problems solved by technology

The labeling of remote sensing image samples relies on manual operation, which is a large amount of engineering, high cost, and difficult to implement. It is inevitable that wrongly labeled data will appear in the sample data labeled manually or by machine.

Remote sensing image sample data from different regions, different time periods, and different sources cannot be directly reused because of their different characteristics and distributions. At present, there is no effective strategy for reusing historical samples. These problems have led to the limitations of existing remote sensing image samples. Large and underutilized

Images