Smart Shelf System for Customer Behavior Tracking in Supermarkets

. 2024 Jan 8;24(2):367. doi: 10.3390/s24020367

Algorithm 2: Multiple Particle Filter Algorithm
1:	Input: $(M_{τ})$ $containing matched vision attributes (V_{τ}$ $) : customer track ID (V_{τ T_I D}$ $), bounding box coordinates (V_{τ b b o x}$ $), Keypoint 9 coordinates (V_{τ K 9}$ $), Keypoint 10 coordinates (V_{τ K 10}$ )
2:	Output $: Bounding box particles (P_{b b o x}$ $), Keypoint 9 particles (P_{K 9}$ $), Keypoint 10 particles (P_{K 10}$ $) of each track ID (V_{τ T_I D}$ )
3:	For $len (M_{τ}$ ) iterations, follow the following procedure:
4:	$Extract the vision attributes inside M_{τ}$
5:	If ${V_{τ}}^{1}$ , then
6:	$Initialize random particles P$ $with coordinates (P_{x}, P_{y}$ ) based on the $initial coordinates of {V_{τ}}^{1}$
7:	$Compute the distance between the landmark and initial coordinates {V_{τ}}^{1}$
8:	Move the particles diagonally by adding a diagonal distance to the particles
9:	Calculate for the weights of each particle
10:	Apply systemic resampling
11:	$Store particles (P$ ) to global particles repository
12:	Otherwise
13:	$Extract global particles (P$ ) $containing current P_{b b o x}$ $,_{} P_{K 9}$ $, P_{K 10}$
14:	Compute the distance between the landmark and the current $coordinates {V_{τ}}^{n}$
15:	Move the particles diagonally by adding a diagonal distance to the particles
16:	Calculate the weights of each particle
17:	Apply systemic resampling
18:	Store particles (P) to global particles repository
19:	End
20:	$Collect final particles (P$ ) $containing final P_{b b o x}$ $,_{} P_{K 9}$ $, P_{K 10}$