研究目的
To overcome the 'small sample size problem' in hyper spectral classification methods by using Maximum Scatter Discriminant analysis.
研究成果
The proposed Maximum Scatter Discriminant analysis method effectively overcomes the 'small sample size problem' in hyper spectral image classification, demonstrating better recognition accuracy and robustness compared to PCA and LDA.
研究不足
The study does not discuss the computational complexity or the scalability of the proposed method to larger datasets or different types of hyper spectral images.
1：Experimental Design and Method Selection:
The study uses Maximum Scatter Discriminant (MSD) analysis to classify hyper spectral images by maximizing the difference between between-class scatter and within-class scatter matrices.
2：Sample Selection and Data Sources:
The Indian Pines HSI data set, collected by the AVIRIS sensor, is used for experiments.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The method involves PCA projection for dimensionality reduction, constructing scatter matrices, MSD projection, and classification using a nearest neighbor classifier.
5：Data Analysis Methods:
Recognition accuracy is compared between the proposed method, PCA, and LDA.
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
