Granted Patent B2

US 12,450,754 · App. 17/769,374 · Granted Oct 21, 2025

Inventors: In Sik Shin (Seoul, KR); Jun Hyuk Park (Seoul, KR); Jin Woo Park (Incheon, KR); Jae Hyun Kim (Seoul, KR); Yeo Hyun Park (Wonju-si, KR)

Assignee: MAY-I INC.

G06T7/246G06Q30/0202G06Q30/0205G06T7/73G06V10/761G06V10/764G06V40/10G06V40/20G06T2207/10016G06T2207/30196G06T2207/30241

Abstract

Claims (25)

1. A visitor information acquisition method performed by a computing apparatus, comprising the steps of:

detecting one or more visitor objects from a video;

estimating attribute information of a first visitor using feature data of a first visitor object;

acquiring a perspective model of the first visitor on the basis of a travel route of the first visitor object; and

acquiring interaction information of the first visitor using the perspective model of the first visitor,

wherein in the step of detecting the one or more visitor objects from the video, in response to a plurality of visitor objects being detected from the video, it is determined whether the plurality of visitor objects are simultaneously detected from a single frame of the video, and according to a result of the determination, it is determined whether to synchronize identification values of the plurality of visitor objects, and

wherein in the step of acquiring the perspective model, a trend plane for each of the one or more visitor objects is acquired using a bounding box corresponding to a head region of each of the one or more visitor objects, and the perspective model of the first visitor is acquired by applying a constant corresponding to a height of the first visitor to a straight line representing a line of intersection between the trend plane for each of the one or more visitor objects and a xy-plane.

2. The visitor information acquisition method of claim 1 , wherein the step of detecting the one or more visitor objects comprises the steps of:

detecting the first visitor object and a second visitor object; and

changing an identification value of a second visitor to an identification value of the first visitor when similarity between the feature data of the first visitor object and feature data of the second visitor object is not less than a reference value.

3. The visitor information acquisition method of claim 2 , wherein the step of changing the identification value of the second visitor to the identification value of the first visitor comprises the step of:

not changing the identification value of the second visitor to the identification value of the first visitor when the first visitor object and the second visitor object are detected from a single frame.

4. The visitor information acquisition method of claim 1 , wherein the perspective model is a linear model for transforming a screen coordinate system of the video to a real-world coordinate system, and

wherein the step of acquiring the interaction information of the first visitor comprises the steps of:

acquiring a travel route of the first visitor object using the perspective model of the first visitor;

converting a travel route of the first visitor object in the screen coordinate system into a travel route of the first visitor in the real-world coordinate system on the basis of the perspective model of the first visitor; and

acquiring the interaction information of the first visitor using the travel route of the first visitor in the real-world coordinate system.

5. The visitor information acquisition method of claim 4 , wherein the step of estimating the attribute information of the first visitor comprises the step of:

acquiring pose data of the first visitor using the feature data of the first visitor object, and

wherein the step of acquiring the interaction information of the first visitor comprises the step of:

using information on a product corresponding to the travel route of the first visitor in the real-world coordinate system and the pose data of the first visitor to acquire information on the first visitor's interaction with the product.

6. The visitor information acquisition method of claim 4 , wherein the step of acquiring the interaction information of the first visitor comprises the steps of:

acquiring a region-specific interest index of the first visitor on the basis of the travel route of the first visitor in the real-world coordinate system and a travel time; and

designating a category of the first visitor using the region-specific interest index of the first visitor and the interaction information of the first visitor, and

wherein the category of the first visitor includes a visiting customer, an interested customer, an intrigued customer, and a purchasing customer.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (11)