We developed an AI object detection system for a public transportation system in Hong Kong.
The object detection system triggers an alarm when it detects tiny metal objects such as nails, screws, keys, or coins that appear at the escalator comb plate. The operation team could then remove the dangerous object before it jams the running escalator, thus reducing maintenance cost, downtime, and potential accidents.
The solution consists of high-resolution cameras, a high-performance edge computing device, and a highly accurate deep learning model. It has been deployed and verified in multiple sites with an accuracy of up to 98%.