Dataset collection for use in automatic quay crane research

Abstract

A lack of public datasets in Automated Quay Crane operations has hindered the growth in smart ports. The objective of this research is to enhance safety and efficiency in port operations by focusing on the detection of humans in operational areas.

The research commenced with the collection of sensory data using sophisticated camera technology. This data was then processed, meticulously labeled, and augmented to ensure its accuracy and diversity. The core of this study involved the development and fine-tuning of a machine learning model using the YOLOv8 framework, which is designed for real-time object detection. The performance of the model was evaluated using metrics like precision, recall, and Mean Average Precision, with the results indicating a high level of accuracy in detecting human presence.

The evaluation showed that the model achieved a high level of accuracy in identifying humans. The model was also tested in live tracking where it was able to detect humans with high accuracy.