A coastal salt marsh vegetation sample automatic generation method based on phenological characteristics and abnormal sample elimination

By constructing phenological feature rules and methods for removing abnormal samples, a high-quality automated sample set is generated, which solves the problems of sample acquisition difficulties and noise interference in the remote sensing classification of salt marsh vegetation, and realizes high-precision identification and dynamic monitoring of salt marsh vegetation.

CN122368680APending Publication Date: 2026-07-10HOHAI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HOHAI UNIV
Filing Date
2026-05-22
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

In remote sensing classification of vegetation in salt marshes, it is difficult to obtain high-quality training samples. Traditional methods rely on manual annotation, which is costly and time-consuming, and lacks an automatic sample construction mechanism, resulting in low classification accuracy. In particular, different vegetation types are easily confused in complex coastal environments, and noisy samples have a serious impact.

Method used

A method based on phenological characteristics and outlier removal is adopted. By constructing multi-time series remote sensing data, phenological characteristic rules are built using vegetation index time series curves. Combined with a collaborative representation model, pixel-by-pixel judgment and outlier removal are performed to generate a high-quality automated sample set and execute a hierarchical classification process.

Benefits of technology

It significantly improved the accuracy and robustness of salt marsh vegetation classification, with a significant increase in sample set purity and classification accuracy. The distinction between categories such as Spartina alterniflora, Suaeda salsa, and Phragmites communis was more accurate, with the overall accuracy increasing from 94.05% to 99.58% and the average accuracy increasing from 80.13% to 87.39%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure FT_1
    Figure FT_1
  • Figure FT_2
    Figure FT_2
  • Figure FT_3
    Figure FT_3
Patent Text Reader

Abstract

This invention discloses an automatic generation method for coastal salt marsh vegetation samples based on phenological characteristics and outlier removal. The method includes: acquiring a multi-temporal remote sensing image time-series dataset of the study area; calculating time-series curves of various vegetation indices after preprocessing; constructing phenological characteristic rules for the target salt marsh vegetation based on the time-series curves of the vegetation indices; performing pixel-by-pixel judgment on each pixel of the study area using the phenological characteristic rules to select pixels that conform to the phenological characteristics of various vegetation types, forming initial candidate sample areas for each type of vegetation; constraining the initial candidate sample areas using a priori spatial mask to obtain purified candidate sample areas; generating a pure automated sample set after traversing all candidate samples; and performing a hierarchical classification process based on the generated pure automated sample set. This invention can combine temporal remote sensing characteristics with a hierarchical classification strategy to achieve high-precision identification and long-term temporal change monitoring of coastal salt marsh vegetation.
Need to check novelty before this filing date? Find Prior Art