MEASUREMENT OF PEOPLE AND VEHICLES IN FISHING PORTS BY USING AI IMAGE RECOGNITION

Abstract

 This study aims to quantify people and vehicles in a fishing port by using AI image recognition, and to examine how to use it to improve objectivity in planning and project evaluation which is conventionally conducted based on interviews from users. A demonstration test held at a fishing port with one of the largest catches in Hokkaido. In the fishing port, facilities have been improved for hygiene control and work efficiency. We quantified the use condition before and after commencement of new facilities.

 We developed an object detection model and an object tracking model using CNN （Convolutional Neural Network）, which is a kind of deep learning. The object detection model captured temporal changes in the number of workers, forklifts, and birds by using AI image analysis. These highly objective data such as total work hours, can be used as a basis for project evaluation.