An Unsupervised Deep Hyperspectral Anomaly Detector

. 2018 Feb 26;18(3):693. doi: 10.3390/s18030693

Algorithm 1 Adaptive Weight DBN Based HSI Anomaly Detection

Input:
X HSI image dataset

$N_{X}$ the number of pixels in X

$N_{n}$ the number of selected neighboring pixels
Output:
$δ$ Detection results that anomaly score

1: function Anomaly Detection(X)
2:

M_{D B N}

← Training via gradient descent with X
3: (C, R) ← EncodeDecode(

M_{D B N}, X

)
4: for j = 1 to

N_{X}

do
5:

c_{n}

← from C following Section 3.3
6:

r_{n}

← from R following Section 3.3
7: for

i = 1

N_{n}

do
8:

w t_{n} [i]

← Equation (9) and Equation (10) with

r_{n}

9: i ←

i + 1

10: end for
11:

δ [j]

← Equation (7) by using

w t_{n}

and

c_{n}

12: j ←

j + 1

13: end for
14: return

δ

15: end function
16:
17: function EncodeDecode(

M_{D B N}, X

)
18: Initialize the

auto_encode

from

M_{D B N}

19: for j = 1 to

N_{X}

do
20: x ← one pixel from X
21: Encode x with

auto_encode

22:

C [j]

← output of middle layer of

auto_encode

23:

\hat{Y} [j]

← decode

C [j]

with

auto_encode

24:

R [j]

← Equation (5) with

h [j]

and x
25: j ←

j + 1

26: end for
27: return R and C
28: end function