TRASCOPE-AI

[Marubeni Network Solutions Inc.]
TRASCOPE-AI
(Commercial DX: People flow marketing solution)

Videos captured by a camera (which may be an existing one) installed in a commercial facility are processed for traffic line analysis and attribute analysis performed by a human behavior analysis algorithm, and the analysis results are displayed on the dashboard.

Issues

In recent years, store marketing of large commercial facilities is analyzing activities of individual consumers and introducing strategies that match the placement of stores and commodities with people's flow. However, it is not easy to analyze human flow at multiple distant locations.

There were the following issues:

Issue 1: Face authentication The face of an individual cannot be captured by all cameras, so the flow of the person cannot be tracked. Also, there are issues of privacy and lowered recognition performance due to wearing of masks.

Issue 2: Gait authentication It is difficult to install cameras at positions from which gaits of individuals can be identified.

Issue 3: Estimation of people flow Estimation of people flow is difficult to realize due to the cost problem, as well as the doubt of accuracy although estimation is made based on various sensors.

Solution overview

To solve these problems, Marubeni Network Solutions furnished its AI video monitoring service, "TRASCOPE-AI", with a "people flow marketing" function developed using Imaging AI*, which is a FORXAI core technology that recognizes a person from the person's total image including clothes and personal effects. This service contributes to the planning of an efficient marketing strategy and sales increase by automatically obtaining marketing data of visitors to commercial facilities and retail stores.

This service can recognize and track a person even if the face is hidden by a mask or hat. The service does not determine who the recognized person is, so it does not involve much privacy concern of customers. In addition, the FORXAI image integrity coordination technology solved the problem of integrity between multiple cameras at different positions, which has been difficult to solve. These AI processes are executed in edge devices installed at stores, so that captured videos are not sent to external destinations such as those in the cloud, reducing the risk of information leak.
Results of these AI processes are sent as anonymous data to "TRASCOPE-AI" for analysis, to provide data on the dashboard mainly about the following three features that customers want:

(1) Flow of visitors
(2) Relationship between stores in the facility and visitors as well as between different stores (in the quest for synergistic effect)
(3) Attributes (number of visitors, age, clothes, etc.) of visitors in different hours

* The figure only illustrates the concept. Actually, there is not a function to store videos of captured people.

"People flow marketing" has only been possible after the identification of individual people. The advantage of this solution is that it makes the analysis of the flow of individuals possible with existing cameras. This solution visualizes the flow of people, with consideration for consumers' privacy, by automatically obtaining marketing data of facility store visitors.

*FORXAI Imaging AI
One of the features of FORXAI imaging AI technology is to recognize different images of persons as an identical person by recognizing and comparing multiple properties of the person such as gender, age group, top/bottom clothing, wearing of a hat, and personal effects. This feature does not need face recognition, and can recognize people from images captured in different angles. The ability of clothes color recognition could be used for fashion trend analysis.
* Attribute recognition is an estimation based on statistics and machine learning obtained depending appearance, and does not make an accurate judgment.

Reason for using FORXAI

There are other solutions for tracking the flow of people by recognizing individuals based on various attributes. The reason for using FORXAI is that it overcame the challenge of recognizing clothes colors by calibrating fluctuation caused by lighting (based on high-accuracy color calibration technology).

Future prospect of the solution

Traffic DX that Konica Minolta engages in commercial facilities: We are considering the use of this solution in combination with traffic jam forecast, as well as in other industrial fields.

FORXAI technology being used