Improved CNN-Based Indoor Localization by Using RGB Images and DBSCAN Algorithm

. 2022 Dec 6;22(23):9531. doi: 10.3390/s22239531

Algorithm 1 Different DBSCAN-based K value selections.

Input :
P from CNNs, $P \in R^{1 \times Q}$ , epsilon, minPts, D.
1:
Sort P in descending order of size, and go from one dimension to two dimensions;
2:
$P_{m a x} \leftarrow P$ , epsilon = 2, minPts = 1, D = 0, K = 0;
3:
while $j = 1$ to Q do
4:
$D = \sqrt{{[p_{max} (x) - p_{j} (x)]}^{2} + {[p_{max} (y) - p_{j} (y)]}^{2}}$
5:
if $D < e p s i l o n$ then
6:
$K + +$ ;
7:
else
8:
K;
9:
end if
10:
end while
Output :
K, $K \in (1, 2, 3)$