A method for joint classification of hyperspectral images and lidar data

By decoupling the three-branch encoder, feature fusion module, and structural consistency perception discriminator, the problems of modal feature entanglement and training stability of hyperspectral and lidar data are solved, and high-precision multimodal feature fusion and classification are achieved.

CN121305244BActive Publication Date: 2026-03-03QILU UNIVERSITY OF TECHNOLOGY (SHANDONG ACADEMY OF SCIENCES)
2 Cites 0 Cited by

Patent Information

Application Number
CN202511852636.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-12-10
Publication Date
2026-03-03
Estimated Expiration
2045-12-10

AI Technical Summary

Technical Problem

Existing technologies for classifying hyperspectral and lidar data suffer from problems such as modal feature entanglement and independence, imprecise cross-modal guidance, insufficient stability during adversarial training, and inadequate semantic preservation, all of which affect classification accuracy.

Method used

A decoupled three-branch encoder is used for multi-scale feature extraction, a feature fusion module is used for cross-modal fusion, data is reconstructed through a weight-shared decoder, and a structure consistency-aware discriminator is used for identification. Finally, a multi-level feature fusion classifier is used for classification.

Benefits of technology

It improves the adaptive fusion and enhancement of multimodal features, increases classification accuracy, and realizes joint classification of hyperspectral images and lidar data.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

This invention relates to the field of deep learning technology, and in particular provides a joint classification method for hyperspectral images and LiDAR data. The method first extracts multi-scale features from multimodal data by decoupling a three-branch encoder. Then, a self-designed feature fusion module performs attention-guided cross-modal fusion on three initial features at the same scale, resulting in three multimodal fused features at different scales. The deepest fused feature is input to a weight-shared decoder, where pseudo-data is obtained through a reconstruction task. A structure consistency-aware discriminator identifies real data from pseudo-data and returns a discrimination loss, thereby updating the encoder parameters and iterating until optimal. Finally, the fused features at the three scales obtained based on the optimal parameters are input into a multi-level feature fusion classifier to complete the classification. This method effectively improves classification accuracy and achieves adaptive fusion and enhancement of multimodal features.
Need to check novelty before this filing date? Find Prior Art