Abstract
This article presents a label similarity probability filter (LSPF) to make hyperspectral image postclassification. The LSPF is inspired by the first law of geography and proposes a class label probability function to quantify the probability of both centered and its neighboring pixels belonging to the same class. It first classifies the hyperspectral data using the regular support vector machine classifier. Then, it binarizes the posterior classification result to obtain the binary label maps of each class. After that, it traverses all spatial windows centered by each pixel and calculates the cumulative probability of all pixels in each class. Finally, the cumulative probabilities are used to make reclassification to obtain the refined classification map. The experiments on Indian Pines, Pavia University, and ZY1-02D Yellow River Estuary data show that LSPF greatly improves the classification accuracy of spectral signatures and outperforms other state-of-the-art spectral-spatial methods.
Highlights
Hyperspectral imaging sensor, as an advanced earthobservation technique, can simultaneously capture hundreds of narrow spectral bands that often range from visible to shortwave infrared wavelength [1, 2]
When having the pre-classification result from regular classifiers like support vector machine (SVM), the classification map can be transformed into a series of binary classification maps
Label Similarity Probability Filter (LSPF) 91.71±4.43 96.37±0.82 96.87±2.23 98.78±0.79 96.92±1.18 98.81±0.50 94.40±10.43 99.95±0.10 82.22±20.18 96.05±1.40 96.78±1.17 97.12±2.48 98.70±0.73 98.89±0.79 98.90±0.94 95.24±3.26. This proposed LSPF is compared with five state-of-the-art methods, including spectral characteristics-based methods: only spectral features (SF) are used for classification by SVM (SVM-SF) [40], preprocessing-based methods: Extend Attribute Profile (EAP) [41] and Simple Linear Iterative Clustering (SLIC) [42], integrated-based methods: SVM-CK [43], and postprocessing-based methods: bilateral filter (BF) [44]
Summary
Hyperspectral imaging sensor, as an advanced earthobservation technique, can simultaneously capture hundreds of narrow spectral bands that often range from visible to shortwave infrared wavelength [1, 2]. The LSPF is inspired from the first law of geography and assumes that two spatially closer pixels within a neighborhood have higher probability in taking the same class labels when compared against other pixels that are further away It defines the label similarity probability (LSP) to quantify the correlations between the center pixel and its neighbors within the same spatial window. It filters the binary label maps using the LSPF to calculate the cumulative probabilities of each pixel in all classes to refine the pre-obtained class label. The LSPF estimates the label probability of all pixels in all the classes by using the spatial correlations with surrounding pixels and implements the probability data into the classifier to refine the pre-obtained classification map. Using the regular SVM classifier, we reclassify the accumulative probability data of all pixels into different classes and obtained the refined classification map
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
