Powered by VANTIQ

Vantiq Solution Center

AI-Enabled Monitoring of Factory Workers

Improving Worker Safety and Food Quality Assurance with Camera-Based AI Detection System in a Food Manufacturing Factory

by BroaderBiz, Inc

Overview Details Benefits Contact

Using Edge devices, IP cameras and AI Box, the solution observes workers' movements within a factory and detects laborers who have collapsed or are injured or ill. The solution monitors suspicious acts such as unauthorized entry or work interruption. Once the status is judged abnormal, data is sent to VANTIQ and a supervisor is alerted.

Highlights

The workers' movement data is gathered through edge devices, IP cameras, and AI Box (Jetson TX2).
Posture identification (Supervised Learning, In-house Teaching data is used) alerts when positions that do not occur in usual work are detected, such as a situation when a worker has fallen or is injured.
Detection of abnormalities in position, posture, number of workers, and moving area. Entry to the backside of the manufacturing line, higher or lower moving speed of workers, and detection of dollies or containers left out in working areas.
Detection of an abnormality such as unsupervised staff. By learning patterns of stored data over months, one can use pre-defined posture types and coordinate values for the detection of new data as an abnormality.
Development of notification rules. Defining rules for each abnormality creates an immediate notification of such abnormality. One can use appropriate communication channels such as email, LINE, SMS, and smartphone apps for each notification.
Data Accumulations Data for all abnormalities, including predefined HIARI HATTO, means scary incidents nearly missed in Japanese. Enabling check records of anomaly detection on a dashboard.

Solution Video

Gallery

Resources

Brochure-AI_watching_system

A print material about this solution

In manufacturing plants, a review of the labor environment is as important as the efficiency of the manufacturing process. In the food industry, a high level of labor motivation is as important as preventing contamination with foreign matters and with drugs by malicious persons. In such circumstances recognizing individual persons is very difficult because the uniforms, caps, masks are worn.

This system is able to identify individual persons by ID number given by the system to all the laborers at their entry to the plant. OpenPose, which works on edge devices, like IP cameras and an AI Box set up near production lines, detects a worker’s skeleton and posture in a factory. Through skeleton data, OpenPose captures and TensorFlow judges if a posture is normal or abnormal at the primary level. Data captured by edge devices is sent to MQTT broker in an edge server, then MQTT data is registered in Elasticsearch. Unusual skeleton data detected through unsupervised machine learning is sent to VANTIQ, using Webhook. All data gathered to VANTIQ Edge are checked again to determine if they’re normal or not. If the system detects an abnormality in monitored labor notification is sent to an appropriate factory manager according to predefined rules. Data are recorded along with captured images in VANTIQ Cloud and used for anomaly detection’s administration and notification. A dashboard is provided as well, on which recorded data can be seen.

Overview
https://www.youtube.com/watch?v=I9sex7uJZx4&feature=youtu.be

Technology
https://www.youtube.com/watch?v=hbElTIRwTFU&feature=youtu.be

Detect unusual situation in a factory immediately, in case a worker collapses or unauthorized entry or work interruption takes place.
Find causes for repetitive drops of products.
Monitor the production process for making continuous improvement (KAIZEN) change and positioning of the process.
Data acquisition for replacing with robotics.
Rapid detection of contamination or failure by checking for workers' unusual behaviors.

This form may not appear if your browser is running in private mode.

Website

Location

14th floor, Telecom-Center bldg. , 2-5-10 Aomi, Koto-ku, Tokyo 1350064

English