Sensor-Oriented Path Planning for Multiregion Surveillance with a Single Lightweight UAV SAR

. 2018 Feb 11;18(2):548. doi: 10.3390/s18020548

Algorithm 2 ROI Classification.

Input: A symbolic image with ROIs labeled by sequential integers.
Output: The category of each ROI in the map.
Begin:

1:
Detect the maximum integer in the symbolic image: $N_{R}$ ;
2:
Perform Canny edge detector to the image, then search the pixel coordinates on the edge of each ROI: $E_{i} = {[X_{i}, Y_{i}]}^{T};$
3:
for i = 1 to $N_{R}$ do
4:
for $θ_{k}$ = 0 to 90 do
5:
Calculate the coordinate of the edge pixels in the rotated Cartesian coordinate system: ${\hat{E}}_{i} (θ_{k}) = T (θ_{k}) \cdot E_{i} = {[{\hat{X}}_{i} (θ_{k}), {\hat{Y}}_{i} (θ_{k})]}^{T}$ ;
6:
Calculate the two-dimensional spans of the ROI: $Δ \hat{X} = \max [{\hat{X}}_{i} (θ_{k})] - \min [{\hat{X}}_{i} (θ_{k})]$ , $Δ \hat{Y} = \max [{\hat{Y}}_{i} (θ_{k})] - \min [{\hat{Y}}_{i} (θ_{k})]$ ;
7:
Initialize $R_{i}$ as a distributed ROI under $θ_{k}$ ;
8:
Calculate the side-lengths of the MBR: $a_{i}^{k} = \max [Δ \hat{X}, Δ \hat{Y}]$ , $b_{i}^{k} = \min [Δ \hat{X}, Δ \hat{Y}]$ ;
9:
if $b_{i}^{k} < l_{s w}$ then
10:
$R_{i}$ is a point ROI under $θ_{k}$ ;
11:
if $a_{i}^{k} > l_{s w}$ then
12:
$R_{i}$ is a quasi-point ROI under $θ_{k}$ ;
13:
end if
14:
end if
15:
end for
16:
Traverse all $θ_{k}$ and classify $R_{i}$ according to (16).
17:
end for