研究目的
To propose a spectral–spatial method for classification of hyperspectral images (HSIs) by modifying traditional random walker (RW) to improve classification accuracy.
研究成果
The proposed spectral–spatial RW method significantly increases the classification accuracy of HSI by modifying the energy function and weighting function in the RW method, integrating spectral information, and reducing the impact of noisy edge weighting function.
研究不足
The main limitation is the computational cost associated with solving the system of linear equations in the RW method, although strategies to reduce this cost are suggested.
1：Experimental Design and Method Selection:
The proposed method modifies the traditional RW by introducing a low-frequency edge weighting function and a fusion of spectral and spatial Laplacian matrix.
2：Sample Selection and Data Sources:
Two hyperspectral datasets taken by AVIRIS sensor and ROSIS-3 hyperspectral sensor are used.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The method involves constructing a spatial RW from a proposed weighting function, constructing a spectral RW, and computing the classification map by the proposed energy function.
5：Data Analysis Methods:
The performance is evaluated using overall accuracy (OA) and Kappa coefficient (κ).
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
