Related Experiment Videos
[The analysis of error sources for SAM and its improvement algorithms]
Hong Tang1, Pei-jun Du, Tao Fang
1The Institute of Image Processing and Pattern Recognition, Shanghai Jiaotong University, Shanghai 200030, China.
Guang Pu Xue Yu Guang Pu Fen Xi = Guang Pu
|December 7, 2005
Summary
This study identifies four error sources in spectral angle mapping (SAM) and introduces improved algorithms. These methods enhance spectral data processing by addressing common errors for accurate classification.
Area of Science:
- Remote Sensing
- Spectroscopy
- Data Analysis
Context:
- Spectral Angle Mapping (SAM) is a common technique for analyzing spectral data.
- Existing SAM methods are susceptible to various error sources.
- Accurate spectral classification is crucial in many scientific fields.
Purpose:
- To identify and analyze the primary error sources affecting Spectral Angle Mapping (SAM).
- To develop and present novel algorithms for improving SAM accuracy.
- To enhance the reliability of spectral data analysis.
Summary:
- Analysis revealed four key error sources in SAM: waveband location shifts, attribution ratio changes, random attribution variations, and overall waveband translation.
- Improved algorithms include grouping, normalization, and intersection methods.
- Grouping considers global and local spectral features to resolve pseudo-similarities. Normalization mitigates random data errors by standardizing spectral vectors. Intersection corrects errors from waveband translation, enabling accurate spectral class identification.
Impact:
- The proposed algorithms effectively address identified error sources in spectral data.
- These improvements enhance the accuracy and reliability of spectral classification.
- The methods offer a robust solution for processing spectrally erroneous data in remote sensing and other applications.