研究目的
To address the problem of ignoring edge information in hyperspectral image classification by utilizing spatial and edge features through deep learning to achieve high accuracy classification.
研究成果
The proposed MNFB-CNN method effectively combines spatial and edge information for hyperspectral image classification, achieving higher accuracy compared to other methods. The integration of MNF for spatial information extraction and bilateral filtering for edge information, followed by classification using CNN, demonstrates significant improvement in classification accuracy. The method's success is attributed to the preservation of useful information and effective noise reduction.
研究不足
The study acknowledges the influence of experimental equipment, environment, and other factors on the size of the training sample and the classification results. The method's effectiveness is demonstrated on specific datasets, and its generalizability to other datasets or real-world applications may require further validation.
1：Experimental Design and Method Selection:
The methodology involves extracting spatial information using Minimum Noise Fraction (MNF) and edge information using bilateral filtering. These features are then fused and fed into a Convolutional Neural Network (CNN) for classification. An additional edge-preserving filter is used to amend the final classification results.
2：Sample Selection and Data Sources:
Three public datasets were used: the University of Pavia, the Salinas, and the Indian Pines.
3：List of Experimental Equipment and Materials:
The study utilized hyperspectral image data collected by ROSIS sensor (ROSIS-3) and AVIRIS sensor.
4：Experimental Procedures and Operational Workflow:
The process includes preprocessing the hyperspectral image to extract spatial and edge information, fusing these features, and then classifying them using CNN. The classification results are further refined using an edge-preserving filter.
5：Data Analysis Methods:
The performance of the proposed method was evaluated using overall accuracy and kappa coefficient, comparing it with other methods like KNN, RAW-SVM, LFDA-SVM, CNN, and PCA-CNN.
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
