Progress monitoring of pipe installation by applying deep learning to process image and installation BOM

초록

This study was conducted to identify the list of parts being assembled and installed on the ship's piping installation process. During the outfitting installation process, An attempt was made to extract valve items from the image information of the piping installation workshop first. As a method for object recognition from image images, we further apply the function of extracting the number of objects based on the object recognition algorithm YOLO-V4, especially learning about valves among piping parts. Through this, the company detection and classifies Work-In-Process (WIP) of the valve being installed and uses the number of items detected and Bill of Material (BOM) data from the installation plan to determine the progress of the installation process.