Sorting System for Ship Outfitting Pipes Based on Enhanced Object Detection

Sorting pipe fittings are a necessary process before ship outfitting. During the sorting process, workers check the quantity and the ID number of the pipe fittings and sort the pipe fittings into different pallets, which are then transported to different assembly areas. Manual searching for paper documents takes up most of the sorting time. To overcome this problem, this paper proposes a digital system incorporating an object detection model to assist in the recognition of pipe fittings. YOLOV4 is adopted as the object detection model. A large number of synthetic 3D model images of pipes are generated to enhance the training dataset for object detection and have improved MAP by 7.64%. The digital system further filters the object detection result by the inputted manual empirical information and has finally achieved a detection accuracy of 82.93%. The proposed system helps to efficiently identify pipe fittings and has reduced time spent on the ship’s pipe sorting process.