研究目的
To improve change detection performance in multispectral and hyperspectral images by effectively combining magnitude and angle measures through a novel fuzzy inference combination strategy.
研究成果
The proposed Fuzzy CVA (FuzCVA) approach effectively combines the distance and angle measures of the change vector for enhanced accuracy in change detection. It provides a performance as good as current state-of-art with the advantage of facilitating simple and fast pixel-by-pixel processing. The FIS system can be tuned according to data or imaging system for improved accuracy.
研究不足
The generic FIS with membership functions and parameters may not be optimized for all data or applications, potentially limiting performance in some cases.
1：Experimental Design and Method Selection:
The study proposes a Fuzzy CVA (FuzCVA) approach combining the distance and angle measures of the change vector using a Fuzzy Inference System (FIS).
2：Sample Selection and Data Sources:
Synthetic change data constructed from the Indian Pines Hyperspectral data set and real-world Kulis data captured by a visible hyperspectral camera.
3：List of Experimental Equipment and Materials:
Visible hyperspectral camera in the range of 400-900 nm with 270 bands.
4：Experimental Procedures and Operational Workflow:
The FuzCVA approach is applied pixel-by-pixel, with change decisions made by comparing the output change value against a hard threshold.
5：Data Analysis Methods:
Performance evaluated using receiver operating characteristic (ROC) curves, true-positive (TP) and false-positive (FP) detection results, F1 score, and Cohen’s Kappa coefficients.
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
