Yard Computer Vision

When vehicles, containers and trailers move through our facility gate and yard, the operations team uses a Yard Management System (YMS) that means they have to be on hand to monitor the flow of traffic and ensure safety of drivers and pedestrians.

This is a challenging and often time-consuming task, exacerbated by operational layouts that were not designed to handle such volume. Often a vehicle is sent on an elongated route or has to move multiple assets to end up with the container they are meant to move. As a result, there are frequent delays, which impact overall productivity.

There had been previous attempts to improve the process with radio-frequency identification (RFID) tags. However, the tags were found to be unreliable, which only increased our facilities’ operational costs and made managing the business more complex.

We wanted to find a smarter solution to monitor movements of both people and machinery at our facilities. Alongside improving the flow and safety of traffic in the yard, the aim was to free up our people’s time to focus on making more strategic operational improvements.

The solution we came up with harnessed computer vision, an AI-powered technology that processes and analyzes digital images, alongside a management platform to ensure every incident is followed up on. The aim was to embed the technology in new and existing cameras (16 in total) that would target and monitor any movement occurring in these areas.

Any unsafe activity such as drivers speeding, people running or even anyone not wearing the correct PPE would be communicated to our team through automated, real-time alerts using speakers mounted in the yard.

The computer vision system seamlessly tracks moving objects such as vehicles and people from the moment they enter the gate, until they reach their final destination. This continuous monitoring enables the capture of any unsafe behaviors, helping us to minimize incidents and track where assets are throughout their time in the yard.

The platform features a centralized dashboard for visualizing activities and unsafe behaviors, sending email or text message notifications to users when a preset condition is met. This means our team is made aware of any issues as they happen.

On the ground, cameras are paired with speakers that are preprogrammed with key alerts, enabling real-time audio communication with drivers and pedestrians. These messages focus on driving speed, walking paths and people-machine interactions.

The system is dynamic, which means that whenever an opportunity to prevent unsafe activity or identify new information on an asset emerges, we can easily update the model. When a new model is released, it only requires approximately 20 minutes of training video to update the system.