top of page

Linker Vision harnesses the power of AI to secure worker safety

Two people in a factory observing machinery

Employee safety is a critical priority for manufacturers around the world. But for many, the process has remained highly labor-intensive. A limited number of personnel are often responsible for monitoring the well-being of an entire factory's workforce, which can lead to oversights and risks. As it has in many industries, AI is helping to revolutionize how companies operate, and, in this case, protect employees.


Founded in 2013, Microsoft partner Linker Vision is a United States and Taiwanese company paving the way for improved workplace safety and precision with their flagship AI-powered software product, Observ. With a large vision model as a foundation, Observ is built on the VisionAI platform, integrated with Microsoft Azure AI, and enables real-time streaming for customers. Linker Vision's solution can be used across multiple industries and is designed for a variety of real-world applications from traffic management to retail.


"As long as you connect to the Real-Time Streaming Protocol, you can start doing all kinds of AI analytics behind it," said Wendy Lee, Business Development Manager, Global Partnership at Linker Vision, "and all you have to do is connect it with a camera."


A time-consuming, error-prone process

A global semiconductor manufacturer in Asia needed to modernize their monitoring approach for improved workplace safety. They produce more than 12,000 products for various applications covering a variety of end markets including high-performance computing, smartphones, the Internet of Things (IoT), automotive, and digital consumer electronics. The organization's monitoring needs included everything from personal protective equipment (PPE) compliance to restricted area detection for prevention of accidents or injuries.


While robust, the customer's safety protocols needed updating to address dynamic conditions and evolving safety requirements. The integral function of safety on the production line had historically been handled manually with team members on the ground. The company had just five employees working to monitor 1,700 cameras plus infrared sensors across the perimeter, making real-time reporting a major hurdle.


The verifying, assessing, and reporting of various incidents to management was a tedious process. "The real-time issue was very painful. When something happened, the customer needed time to dig in and analyze it—it took 30 minutes," according to Twinkle Cheng, Product Manager Lead at Linker Vision. "It costs a lot of time and effort to do things one by one. Not only to monitor if something happens but also when it does happen, to make sure if it's correct or not," Cheng said.


The lack of real-time monitoring capabilities meant that potential safety risks could go undetected until after an event occurred. To make matters worse, the sensors could raise false alarms, adding more work to an already overloaded security team. The company realized it was time to seek out an innovative AI-powered safety solution.


Three people in hardhats in an outside construction site



"The real-time issue was very painful. When something happened, the customer needed time to dig in and analyze it—it took 30 minutes." —Twinkle Cheng, Product Manager Lead, Linker Vision

Modernization meets the factory

As manufacturing modernizes and automates, the company needed something that would provide real-time monitoring, accurately identify safety risks, and integrate seamlessly into their existing framework. The goal was not only compliance but also a proactive safety culture that could prevent incidents and optimize operational efficiency. The customer decided to take that technological leap with Linker Vision and their Observ platform.


As a first action, Linker Vision collaborated with the semiconductor manufacturing leader to understand their specific needs, pain points, and user journey. "We had interviews with the users to understand what their requirements were for the AI solution, their expectations, and their goal," said Cheng. This insight helped the company understand the best way to deploy Observ and deliver real-time video analytics on the edge.


Linker Vision explored several tools while researching what might work for their customer, but because their platform is built on Azure, it's a simple process to integrate other Azure technologies. From App Service, IoT Hub, and IoT Edge to Azure Machine Learning, they can seamlessly implement any number of tools into a customer's environment. For the manufacturer, Linker Vision also introduced Azure Active Directory to connect the account permission with the enterprise account.


Azure provided a scalable infrastructure for Observ that could process large volumes of data and run AI video analytics efficiently. "We don't need to own the data center construction," noted Lee. "We can focus on providing AI software solutions to our customers and leave all the infrastructure work to Azure."


Even though Linker Vision is an expert in their field, bringing AI to a factory floor is easier said than done. One bad camera angle can derail the results. "Once a model is deployed to one specific environment, it tends not to work well if you put it in another environment with a change of camera angle," said Lee. "That's the pain point currently in terms of deploying VisionAI models to the real world."


But Observ offers a continuous learning platform and data-centric approach that accounts for environmental context and diversity. The semiconductor manufacturer using Observ is tapping into its capabilities for their standard operating procedure and PPE detection, restricted area detection, camera screen availability detection, and people/vehicle counting. Meanwhile, the tool facilitates drift data collection, feeding valuable insights back into its data engine for continuous model improvement, retraining, and accuracy.


This intuitive, vision-based AI platform is helping customers stay ahead of safety issues—both increasing accuracy and reducing time spent managing and monitoring.


A person using a laptop at a desk in a factory

"We don't need to own the data center construction. We can focus on providing AI software solutions to our customers and leave all the infrastructure work to Azure." —Wendy Lee, Business Development Manager, Global Partnership, Linker Vision

Smart surveillance, happy customer

Peace of mind is priceless for managers and workers in a factory environment. Linker Vision delivered just that with the VisionAI platform and seamless edge deployment by building on Azure. Observ automated what was a highly manual process and delivered precise, real-time reporting. "The process makes their work easier. They don't need to handle a lot of false alarms and report and analyze those. They just focus on real, true alarms," said Cheng.


With Linker Vision's help, the semiconductor manufacturer was able to be more proactive when it came to safety concerns and saw a significant reduction in incidents by predicting and preventing potential hazards. In addition, the Observ platform's virtual fence reduced patrol frequency while data visualization improved productivity with one-click report output.


Linker Vision estimates the customer's:


ROI was up 200%.

Working time on alarms was down 90%.

Time spent on restricted area detection was down nearly 100%.


The future of vision-based AI and Azure IoT in the semiconductor industry is bright. Linker Vision looks ahead to opportunities to build out a smart IoT Park for customers using VisionAI for everything from vehicle detection and secure borders to X-ray machines. Future phases could also include the integration of Azure Machine Learning so that customers can train the model in their own environments.


"All along, we've been using this data-centric approach to AI, focusing on the environmental context and diversity," said Lee. "By doing this, we can drive model performance and accuracy to a better stage so we can ensure that our AI solutions have more reliable and actionable insights."

11 views0 comments

Comments


bottom of page