一种遥感影像要素提取过程解释的方法和系统
By collecting and visualizing the features of the intermediate layers of a deep learning model, and using principal component analysis to explain the feature extraction process in remote sensing image processing, the problem of poor interpretability of deep learning models in remote sensing image processing is solved, and the clear definition of features and their importance is achieved.
CN116740556BActive Publication Date: 2026-07-17CHINESE ACAD OF SURVEYING & MAPPING
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINESE ACAD OF SURVEYING & MAPPING
- Filing Date
- 2023-04-10
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Deep learning models have poor interpretability in the feature extraction process in remote sensing image processing, making it difficult to determine the extracted features and their importance.
Method used
By collecting the output of each intermediate layer in the deep learning model, the features are saved in matrix form using the get_layer() function of TensorFlow, and the principal component information is calculated using IncrementalPCA. The results are then visualized using matplotlib to explain the main features of each intermediate layer.
Benefits of technology
This improved the interpretability of deep learning models in the process of interpreting remote sensing images and clarified the extracted elements and their importance.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116740556B_ABST
Abstract
本发明提供一种遥感影像要素提取过程解释的方法和系统，涉及遥感影像图像处理技术领域，包括如下步骤：S1、采集深度学习模型中每个中间层的输出，调用TensorFlow的get_layer()函数将每个中间层计算出的特征以矩阵形式输出并保存；S2、聚合每个中间层的遥感影像数据的多个维度的特征，并计算每个中间层的多个维度的特征的主成分信息，将主成分信息还原为与主成分信息对应的主要特征，将主要特征进行可视化展示；S3、基于主成分信息和可视化展示的主要特征解释每个中间层的主要特征；解决了深度学习模型不能展示提取了具体的要素和每个要素的重要程度，提高了深度学习模型在遥感影像解译过程中的解释性。
Need to check novelty before this filing date? Find Prior Art