A training sample determination method for high-resolution remote sensing image intelligent classification

By quantifying the spectral features and spatial structural complexity of image patches, a two-dimensional spectral-spatial complexity hierarchy is constructed. A highly representative sample set is selected for deep learning model training, which solves the problem of insufficient representativeness of training sample sets in high-resolution remote sensing images and improves the model's interpretation accuracy and generalization ability in complex terrain areas.

CN122176558APending Publication Date: 2026-06-09INST OF GEOGRAPHICAL SCI & NATURAL RESOURCE RES CAS
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
INST OF GEOGRAPHICAL SCI & NATURAL RESOURCE RES CAS
Filing Date
2026-04-30
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing remote sensing image classification methods struggle to construct highly representative training sample sets in high-resolution remote sensing images, resulting in insufficient generalization performance of the models in complex terrain areas. Furthermore, existing sampling methods fail to effectively consider the complexity differences between image patches, leading to insufficient training set coverage.

Method used

By quantifying the spectral feature complexity and spatial structure complexity of image patches, a two-dimensional spectral-spatial complexity hierarchy is constructed. A hierarchical decision criterion is used to select samples from the image patch set to form a highly representative training sample set for training deep learning models.

Benefits of technology

It improves the accuracy and generalization ability of semantic segmentation models in remote sensing image interpretation tasks, and ensures coverage of complex terrain features and model performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122176558A_ABST
    Figure CN122176558A_ABST
Patent Text Reader

Abstract

This application discloses a method for determining training samples for intelligent classification of high-resolution remote sensing images. The method includes: cropping each remote sensing image into at least two image blocks, forming an image block set based on these blocks; determining the spectral feature complexity and spatial structure complexity of each image block in the image block set; determining hierarchical decision criteria corresponding to at least two spectral-spatial two-dimensional complexity levels; based on the hierarchical decision criteria, spectral feature complexity, and spatial structure complexity corresponding to each spectral-spatial two-dimensional complexity level, determining a subset of image blocks belonging to each spectral-spatial two-dimensional complexity level from the image block set; selecting a predetermined number of image block samples from each image block subset, and constructing an image block training sample set for a semantic segmentation model based on the selected image block samples from each image block subset. This scheme can construct a training sample set for a highly representative semantic segmentation model.
Need to check novelty before this filing date? Find Prior Art