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.


The local sliding window Ω(ri).
© Copyright Policy - open-access
Related In: Results  -  Collection

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

sensors-17-00441-f002: The local sliding window Ω(ri).

Mentions: For any pixel ri = (ri1,ri2,…,riL)T where L is the total number of spectral bands, its neighborhood information is spectral correlative and spatial correlative. Assume that Ω(ri) is a local sliding window with ri being the centered pixel and the size of window being w2 = w × w where w is an odd number and positive integer (as shown in Figure 2). The representation form of Ω(ri) is given by:(15)Ω(ri)={rp/p∈[i−a,i+a]}where rp is a pixel within the neighborhood window, and a = (w2 − 1)/2 is a constant. In order to better capture the spatial information, the centered pixel is reconstructed according to the weighted spatial-spectral information in this paper. The reconstructed pixel is specified by:(16)r^i=∑rp∈Ω(ri)ωprp∑rp∈Ω(ri)ωpwhere is the weight of any pixel rp to center pixel ri in the spatial neighborhood , denotes two norm operation. 0 is a spectral factor, indicating the degree of interaction effects between different pixels in the same neighboring space. Accordingly, we can get the kernel function between the reconstructed pixels as:(17)kSS(r^i,r^j)=<Φ(r^i),Φ(r^j)>   =〈∑rp∈Ω(ri)ωprp∑rp∈Ω(ri)ωp,∑rq∈Ω(rj)ωqrq∑rq∈Ω(rj)ωq〉   =∑rp∈Ω(ri)∑rq∈Ω(rj)ωpωqk(rp,rq)∑rp∈Ω(ri)ωp∑rq∈Ω(rj)ωq


A Weighted Spatial-Spectral Kernel RX Algorithm and Efficient Implementation on GPUs
The local sliding window Ω(ri).
© Copyright Policy - open-access
Related In: Results  -  Collection

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

sensors-17-00441-f002: The local sliding window Ω(ri).
Mentions: For any pixel ri = (ri1,ri2,…,riL)T where L is the total number of spectral bands, its neighborhood information is spectral correlative and spatial correlative. Assume that Ω(ri) is a local sliding window with ri being the centered pixel and the size of window being w2 = w × w where w is an odd number and positive integer (as shown in Figure 2). The representation form of Ω(ri) is given by:(15)Ω(ri)={rp/p∈[i−a,i+a]}where rp is a pixel within the neighborhood window, and a = (w2 − 1)/2 is a constant. In order to better capture the spatial information, the centered pixel is reconstructed according to the weighted spatial-spectral information in this paper. The reconstructed pixel is specified by:(16)r^i=∑rp∈Ω(ri)ωprp∑rp∈Ω(ri)ωpwhere is the weight of any pixel rp to center pixel ri in the spatial neighborhood , denotes two norm operation. 0 is a spectral factor, indicating the degree of interaction effects between different pixels in the same neighboring space. Accordingly, we can get the kernel function between the reconstructed pixels as:(17)kSS(r^i,r^j)=<Φ(r^i),Φ(r^j)>   =〈∑rp∈Ω(ri)ωprp∑rp∈Ω(ri)ωp,∑rq∈Ω(rj)ωqrq∑rq∈Ω(rj)ωq〉   =∑rp∈Ω(ri)∑rq∈Ω(rj)ωpωqk(rp,rq)∑rp∈Ω(ri)ωp∑rq∈Ω(rj)ωq

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.