研究目的
To overcome the computational complexity and the issue of extracting high brightness points, outliers, and isolated noise points as endmembers in the simplex maximum distance (SMD) algorithm by presenting an improved simplex maximum distance (ISMD) algorithm.
研究成果
The ISMD algorithm simplifies the computational procedure of the SMD algorithm and reduces the impact of noise pixels and outliers on endmember extraction. It extracts more accurate endmembers that show a smaller difference with real mineral spectral curves, proving to be more reliable than SMD.
研究不足
The number of endmembers that can be extracted is limited by the number of bands in the hyperspectral image. The algorithm may not perform well if the number of endmembers exceeds the number of bands.
1：Experimental Design and Method Selection:
The ISMD algorithm simplifies the computation procedure by defining the distance from a pixel to a simplex as the ratio of volumes of parallel polyhedrons with adjacent dimensions. It selects a set of similar pixels from multiple pixels with a larger distance according to the spectral angle and averages them to be the new endmember.
2：Sample Selection and Data Sources:
Both simulated and real AVIRIS images were used to assess the ISMD algorithm. The simulated image was generated using five types of mineral spectra from the USGS spectral library.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The ISMD algorithm was implemented by computing distances from pixels to simplexes, selecting similar pixels based on spectral angles, and averaging them to form new endmembers.
5：Data Analysis Methods:
The performance of ISMD was compared with SMD using spectral angles between extracted endmembers and corresponding minerals from the USGS spectral library.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
