研究目的
To propose a novel hybrid parametric-nonparametric target detector for hyperspectral images that estimates conditional probability density functions nonparametrically and target strength parametrically, aiming to detect rare targets in complex backgrounds.
研究成果
The proposed hybrid detector effectively detects targets in complex backgrounds with reasonable target strength estimates, showing better performance than the parametric AMF in terms of lower false alarm rates, and warrants further investigation for broader applications.
研究不足
The study uses a specific hyperspectral dataset and assumes an additive target model, which may not capture all complexities; performance could be affected by background variability and the choice of the k parameter in kernel density estimation.
1：Experimental Design and Method Selection:
The study employs a hybrid detector combining nonparametric kernel density estimation for background pdf and parametric Maximum Likelihood estimation for target strength, based on an additive target model.
2：Sample Selection and Data Sources:
Real hyperspectral data acquired by the SIM.GA sensor over a suburban area in Viareggio, Italy, in May 2013, with ground truth from a FieldSpec spectroradiometer.
3：List of Experimental Equipment and Materials:
Hyperspectral sensor (SIM.GA), spectroradiometer (FieldSpec), calibration tarps for Empirical Line Method.
4：Experimental Procedures and Operational Workflow:
Data pre-processing (de-striping, spectral binning), radiometric calibration using ELM, application of the detector to image pixels, comparison with Adaptive Matched Filter (AMF).
5：Data Analysis Methods:
Evaluation of false alarm rates and target strength estimates, with performance metrics based on false alarm pixels and rates.
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
