Remote Sensing Image Segmentation Using Elevation-Weighted SLIC Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional remote sensing image segmentation methods that rely solely on spectral information are not accurate enough for ground objects with certain heights, leading to false segmentation results due to similar spectral characteristics despite elevation differences.
Innovation Solution
A method that fuses elevation information into the remote sensing image segmentation process using simple linear iterative clustering (SLIC) segmentation, introducing elevation information through weighting and setting thresholds, and performing region merging based on comprehensive similarity criteria that combine spectral and elevation information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional superpixel segmentation methods are used that rely solely on spectral information, then the segmentation process is simple and fast, but the segmentation accuracy is insufficient for ground objects with certain heights
Solution Approach 1:
The patent merges spectral information with elevation information to form a comprehensive feature vector for segmentation. By combining multiple information sources (spectral bands and elevation data), the method achieves more accurate segmentation of ground objects with different heights while maintaining a systematic processing framework.
Solution Approach 2:
The patent transitions from two-dimensional spectral information to three-dimensional information by incorporating elevation data. This dimensional expansion allows the segmentation to distinguish between ground objects at different heights that have similar spectral characteristics, thereby improving segmentation accuracy.
2Reliability
If traditional region merging methods are used that do not consider elevation information, then the processing is computationally efficient, but false determination results occur when spectral characteristics of ground objects with elevation differences are similar
Solution Approach 1:
The patent changes the parameters used for region merging from spectral-only criteria to composite criteria that include both spectral and elevation parameters. By modifying the determination parameters to incorporate elevation differences, the method eliminates false merging of regions with similar spectra but different heights, improving determination reliability.
3Loss of information
If only spectral information is used for segmentation, then the data processing is straightforward, but the segmentation results cannot accurately distinguish ground objects with different elevations
Solution Approach 1:
The patent creates a universal segmentation framework that can handle both two-dimensional spectral information and three-dimensional elevation information through a unified feature vector approach. This multi-functional system processes spectral and elevation data together, preventing information loss while maintaining systematic processing.
Data Source
AI summary
Disclosed in the present disclosure is a method for processing remote sensing images by fusing elevation information. The method includes the steps: acquiring an image data set, and preprocessing the image data set to obtain an image set; performing simple linear iterative clustering (SLIC) segmentation processing on the image set, introducing elevation information through weighting and setting a threshold in the SLIC segmentation processing process to obtain pre-segmentation results; updating a neighborhood relationship among clusters on the basis of the pre-segmentation results by using a data structure of a neighborhood array, and establishing a comprehensive similarity criterion of weighting combined with elevation features; and finally, for ground objects with complex elevation, setting determination conditions before merging according to elevation differences between adjacent clusters, and performing region merging by setting weight coefficients of different sizes in comprehensive similarity indexes according to determination results to complete image processing.