A Structured Approach to the Analysis of Remote Sensing Images

Donghui Yan,Congcong Li,Na Cong,Le Yu,Peng Gong

DOI: https://doi.org/10.1080/01431161.2019.1607611

2020-01-05

Abstract:The number of studies for the analysis of remote sensing images has been growing exponentially in the last decades. Many studies, however, only report results---in the form of certain performance metrics---by a few selected algorithms on a training and testing sample. While this often yields valuable insights, it tells little about some important aspects. For example, one might be interested in understanding the nature of a study by the interaction of algorithm, features, and the sample as these collectively contribute to the outcome; among these three, which would be a more productive direction in improving a study; how to assess the sample quality or the value of a set of features etc. With a focus on land-use classification, we advocate the use of a structured analysis. The output of a study is viewed as the result of the interplay among three input dimensions: feature, sample, and algorithm. Similarly, another dimension, the error, can be decomposed into error along each input dimension. Such a structural decomposition of the inputs or error could help better understand the nature of the problem and potentially suggest directions for improvement. We use the analysis of a remote sensing image at a study site in Guangzhou, China, to demonstrate how such a structured analysis could be carried out and what insights it generates. The structured analysis could be applied to a new study, or as a diagnosis to an existing one. We expect this will inform practice in the analysis of remote sensing images, and help advance the state-of-the-art of land-use classification.

Applications

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to better understand the key factors in remote sensing image analysis research and their interactions through a structured analysis method, thereby providing directions for improving land - use classification. Specifically, the paper focuses on:

1. **Understanding the essence of the research**: Understanding the essence of a research through the interaction among algorithms, features and samples, which jointly determine the research results.
2. **Determining the direction of improvement**: When improving existing research, judging whether more efforts should be made to optimize algorithms, find better features, or simply increase the sample size.
3. **Evaluating sample quality and feature value**: How to evaluate the quality of samples or the value of a set of features, especially in the case of a large number of existing samples, whether it is worthwhile to further collect data.
4. **Diagnosing existing research**: Providing a framework that can be applied not only to new research, but also as a diagnostic tool for existing research to help identify potential improvement points.

To achieve the above - mentioned goals, the paper proposes a structured analysis method, regarding the output of the research as the result of the interaction of the three input dimensions of algorithms, features and samples, and decomposing the error into the error along each input dimension. This method helps to better understand the nature of the problem and may point out the direction of improvement. The paper shows how to conduct this structured analysis and the insights it generates by analyzing the data of a remote sensing image research site in Guangzhou, China.