Operator task tracking

A new vision-based tracking system has been developed at Volvo. This system is unique due to its simplicity of using cameras as sensors instead of more expensive systems like ultra wide band trackers. A vision based system does not require a tracker mounted on what is tracked and a combination of deep learning and tracking of markers is used.

The system us using a number of normal cameras that are mounted above the work area. The images are streamed to a server where the objects that are tracked are being identified. The position of the found object is calculated and the position in space is saved. In the first pilot, the tools are tracked.

The cameras needs to be mounted and installed in the station. Curretly, aruco markers are used for tracking the position. But in the future, the position could be inferred using machine learning.

This will be used internally at volvo as a smart sensor system. These cameras are also used for other tasks, like quality inspection and internal logistics.

These are currently internal tools used at Volvo, but most of the developed code is open source.