A Precision Evaluation Index System for Remote Sensing Data Sampling Based on Hexagonal Discrete Grids

Li Guoqing, Yue Ma, Guoqing Li, Xiaochuang Yao, Qianqian Cao, Long Zhao, Shuang Wang, Lianchong ZhangIsprs Int. J. geo inf 2021 · DOI

TL;DR

This paper introduces a novel evaluation index system to assess the accuracy of remote sensing data represented on hexagonal grids. By developing and applying a suite of metrics, the authors demonstrate that hexagonal grids can effectively preserve the information and geometric characteristics of original raster data, paving the way for more efficient spatial data management and analysis.

What Problem Does This Paper Address?

The explosion of spatial data from earth observation, navigation, and communication technologies presents significant challenges for efficient storage, management, and analysis. Global discrete grid systems offer a promising solution for organizing this data, with hexagonal grids being particularly attractive due to their topological properties and sampling efficiency. However, a critical gap exists in evaluating the accuracy of remote sensing data when converted to these hexagonal grid formats. This paper aims to bridge this gap by proposing a comprehensive system to quantify the information loss and geometric distortions introduced by this conversion process.

Methods

The authors propose a multi-faceted approach to evaluate remote sensing data on hexagonal grids:

  • Hexagonal Grid Conversion: They detail a method for converting original raster remote sensing images into hexagonal grid data. This involves determining the appropriate hexagonal grid level based on the spatial resolution of the input image and then resampling the image data to fit the hexagonal cells. Various resampling methods (nearest-neighbor, bilinear, bicubic interpolation) are explored.
  • Accuracy Evaluation Index System: A novel evaluation system is developed, comprising three categories:
  • Basic Evaluation Indicators: These include information entropy, mutual information, deviation index of spectral information, and structural similarity index (SSIM). These metrics assess the overall information content and similarity between the original and converted images.
  • Evaluation Index Based on Image Construction: This category focuses on edge information. The paper introduces hexagonal Sobel operators to analyze edge retention and proposes a method to calculate the relative edge strength and direction.
  • Geometric Evaluation Index: This involves using a checkpoint method to calculate the Root Mean Square Error (RMSE) of pixel values at selected feature points. Additionally, an error band calculation is used to assess the accuracy of linear feature extraction and retention.
  • Analytic Hierarchy Process (AHP): To combine the various evaluation indicators into a single, comprehensive score, AHP is employed to determine the weights of each indicator based on their relative importance.

Datasets

The study utilizes the following remote sensing datasets:

  • Landsat 8: Used for its 30m spatial resolution, particularly focusing on the Yellow River mouth area.
  • GF-1: Used for its higher spatial resolution (2m panchromatic/8m multispectral), with data from Hainan.

The paper does not introduce new benchmark datasets but rather uses publicly available satellite imagery to validate its proposed evaluation system.

Key Insights and Ideas

  • Hexagonal Grids Preserve Information: The study demonstrates that hexagonal grids, when appropriately implemented, can effectively represent remote sensing data while retaining a high degree of information and structural features compared to the original raster data.
  • Bilinear Interpolation is Optimal for Conversion: Among the tested resampling methods, bilinear interpolation consistently yielded the best results in terms of information preservation and accuracy, striking a balance between smoothing and retaining original data characteristics.
  • Comprehensive Evaluation is Crucial: A single metric is insufficient. The proposed multi-indicator system, combining spectral, structural, and geometric evaluations, provides a more robust assessment of data quality on hexagonal grids.
  • Hexagonal Grids Excel in Geometric Feature Representation: The hexagonal structure offers advantages in representing directional and curved features, showing better performance in linear element extraction and refinement compared to traditional quadrilateral grids.

Key Results

  • The comprehensive similarity coefficient between original and converted images using the proposed evaluation method exceeded 98%.
  • Bilinear interpolation resulted in the highest similarity scores across various evaluation metrics for both Landsat 8 and GF-1 data.
  • The RMSE of pixel values at checkpoints was minimized with bilinear interpolation, indicating higher accuracy.
  • The error band width between hexagonal grid data and vector data was smaller than that between original raster data and vector data, highlighting the hexagonal grid's superior ability to maintain curve characteristics.
  • The proposed evaluation index system demonstrated efficient calculation times, generally within 3 seconds per scene image, meeting practical requirements.

Limitations and Future Work

  • Weight Determination: The paper acknowledges that the weights assigned to evaluation indicators through AHP are not absolute and can be adjusted based on specific application scenarios and priorities.
  • Computational Efficiency of Cubic Interpolation: While cubic convolution interpolation can capture more information, its low calculation efficiency makes it unsuitable for large-scale applications.
  • Future Work: The authors suggest that the hexagonal grid's ability to be infinitely divided makes it suitable for storing unstructured emerging spatial big data, including point, line, and human activity data. Further research could explore these applications and refine the evaluation system for different data types and scales.

Summary updated

BibTeX
@article{guoqing2021precision,
  author = {Li Guoqing and Yue Ma and Guoqing Li and Xiaochuang Yao and Qianqian Cao and Long Zhao and Shuang Wang and Lianchong Zhang},
  title = {{A Precision Evaluation Index System for Remote Sensing Data Sampling Based on Hexagonal Discrete Grids}},
  journal = {Isprs Int. J. geo inf},
  year = {2021},
  doi = {10.3390/ijgi10030194},
}
Figure