Limits...
A Weighted Spatial-Spectral Kernel RX Algorithm and Efficient Implementation on GPUs

View Article: PubMed Central - PubMed

ABSTRACT

The kernel RX (KRX) detector proposed by Kwon and Nasrabadi exploits a kernel function to obtain a better detection performance. However, it still has two limits that can be improved. On the one hand, reasonable integration of spatial-spectral information can be used to further improve its detection accuracy. On the other hand, parallel computing can be used to reduce the processing time in available KRX detectors. Accordingly, this paper presents a novel weighted spatial-spectral kernel RX (WSSKRX) detector and its parallel implementation on graphics processing units (GPUs). The WSSKRX utilizes the spatial neighborhood resources to reconstruct the testing pixels by introducing a spectral factor and a spatial window, thereby effectively reducing the interference of background noise. Then, the kernel function is redesigned as a mapping trick in a KRX detector to implement the anomaly detection. In addition, a powerful architecture based on the GPU technique is designed to accelerate WSSKRX. To substantiate the performance of the proposed algorithm, both synthetic and real data are conducted for experiments.

No MeSH data available.


(a) HyMap hyperspectral image; (b) SanDiego Airport image; (c) synthetic image; (d) ground truth of synthetic image.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5375727&req=5

sensors-17-00441-f008: (a) HyMap hyperspectral image; (b) SanDiego Airport image; (c) synthetic image; (d) ground truth of synthetic image.

Mentions: In Figure 8a, The HyMap image was acquired by HyMap hyperspectral remote sensor in the Cook City, MT, USA. There are 126 bands ranging from 0.4 to 2.5 μm with a size of 280 × 800 pixels. The background of synthetic dataset is set as the real ground which cut out the size of 90 × 90 pixels from the HyMap image (white box in Figure 8a). And the targets of synthetic dataset are designed by an airborne hyperspectral image collected by the AVIRIS imaging sensor from San Diego airport. The four targeted spectral signatures (G, H, T, P), marked by circles in Figure 8b, are used to form the synthetic targets. The four spectral signatures are used to simulate 16 targets shown in Figure 8c with 4 targets in each row simulated by same spectral signature. The sizes of panels from left column to right column are 4 × 4, 3 × 3, 2 × 2, 1 × 1, respectively. The panels in the first column are truncated from the original image, and others in 2–4 column for each row are superposed by a different proportion of background interference. And the ground truth of synthetic dataset is given by Figure 8d.


A Weighted Spatial-Spectral Kernel RX Algorithm and Efficient Implementation on GPUs
(a) HyMap hyperspectral image; (b) SanDiego Airport image; (c) synthetic image; (d) ground truth of synthetic image.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5375727&req=5

sensors-17-00441-f008: (a) HyMap hyperspectral image; (b) SanDiego Airport image; (c) synthetic image; (d) ground truth of synthetic image.
Mentions: In Figure 8a, The HyMap image was acquired by HyMap hyperspectral remote sensor in the Cook City, MT, USA. There are 126 bands ranging from 0.4 to 2.5 μm with a size of 280 × 800 pixels. The background of synthetic dataset is set as the real ground which cut out the size of 90 × 90 pixels from the HyMap image (white box in Figure 8a). And the targets of synthetic dataset are designed by an airborne hyperspectral image collected by the AVIRIS imaging sensor from San Diego airport. The four targeted spectral signatures (G, H, T, P), marked by circles in Figure 8b, are used to form the synthetic targets. The four spectral signatures are used to simulate 16 targets shown in Figure 8c with 4 targets in each row simulated by same spectral signature. The sizes of panels from left column to right column are 4 × 4, 3 × 3, 2 × 2, 1 × 1, respectively. The panels in the first column are truncated from the original image, and others in 2–4 column for each row are superposed by a different proportion of background interference. And the ground truth of synthetic dataset is given by Figure 8d.

View Article: PubMed Central - PubMed

ABSTRACT

The kernel RX (KRX) detector proposed by Kwon and Nasrabadi exploits a kernel function to obtain a better detection performance. However, it still has two limits that can be improved. On the one hand, reasonable integration of spatial-spectral information can be used to further improve its detection accuracy. On the other hand, parallel computing can be used to reduce the processing time in available KRX detectors. Accordingly, this paper presents a novel weighted spatial-spectral kernel RX (WSSKRX) detector and its parallel implementation on graphics processing units (GPUs). The WSSKRX utilizes the spatial neighborhood resources to reconstruct the testing pixels by introducing a spectral factor and a spatial window, thereby effectively reducing the interference of background noise. Then, the kernel function is redesigned as a mapping trick in a KRX detector to implement the anomaly detection. In addition, a powerful architecture based on the GPU technique is designed to accelerate WSSKRX. To substantiate the performance of the proposed algorithm, both synthetic and real data are conducted for experiments.

No MeSH data available.