Algorithm 2 Outline of adaptive projection window method.
Input: A 3.0 by 1.5 m image region I ∈ ℝ2N×N and TP ≪ 1 probability threshold, where N is the width of I in pixels.
Initialization: If the filter sliding window, S, of 128 × 64 pixels corresponds to less than 2.0 by 1.0 m, then, accordingly, scale down I.
Procedure:
while S corresponding dimension ≤ 3.0 by 1.5m do
n = 0;	▹ number of total pyramid windows
for i = 1; until S traverses all I vertically; i ← i + 8 do
for j = 1; until S traverses all I horizontally; j ← j + 8 do
Hn ← HoG descriptor of S(Iij)
Cn, Pn ← class and probability output of linear SVM classifier for input Hn
n ← n + 1
end for
end for
Scale down I by a constant scaling factor, F.
end while
Find k for which $P(X>k)=\left(\begin{array}{c}n\\ k\end{array}\right)p^k{(1-p)}^{n-k}\le 1-T_P$
if sum(Cn == +1) > k then
Cout ← +1
Pout ← max(Pn)
else
Cout ← −1
Pout ← min(Pn)
end if
Output: The output class, Cout, and probability, Pout