Research On Recognition And Measurement Method Of Ship Hull Components Based On Image

In recent years,with the rise of the concept of "intelligent manufacturing" in the world,all industries have set off a wave of intelligent upgrading of production technology.The premise of intelligent production is to enable the computer to obtain the ability of cognition and classification,and then realize a series of subsequent functions such as product production and detection.The shipbuilding industry is an important industry related to national defense and shipping.It is the general trend to upgrade intelligent production of shipbuilding enterprises.In the manufacturing process of ships,there are a large number of plane members such as plates,manhole covers,holes,etc.,and most of the ship companies still use traditional measuring tools such as tape measure for the size measurement of these plane members.If the target detection technology can be used for the classification and recognition of hull components,and the computer vision technology can be used for the measurement of hull planar components,the efficiency of component sorting and inspection in actual production will be greatly improved,and the labor intensity of enterprise employees will be reduced.Therefore,this paper starts from the identification and measurement of hull planar members,uses deep learning algorithms to complete the identification and positioning of hull members,and proposes an image-based hull planar member measurement method.The main content of the article is as follows:First,due to the lack of ship component image data sets needed to train deep writing learning models,this article uses reptile crawling and on-site collection to complete the collection of ship component image data.A ship component data set SCTD containing five types of image data of profiles,holes,plates,manhole covers and pipes was produced.Using flip,mirror,and translation methods to complete the expansion and enhancement of the number of image data sets.Next,the paper elaborated on the principles of image recognition and target detection technology and the position calibration of the ship component data set is completed.In view of the disadvantage that image recognition technology cannot detect multiple components in the same image,use the improved version of YOLO target detection model,the target detection of ship components is completed.In the follow-up test,this model can have a better target detection effect on ship voids and profiles.Finally,the measurement experiment of plane components of ships using computer vision technology is completed.In the experiment,the camera calibration technique was used to complete the analysis of the pixel error in the shooting.Through image preprocessing technology and perspective transformation technology,the influence of image perspective distortion and noise on measurement accuracy is reduced.In this paper,a square measuring target suitable for ship's planar members is designed to assist the measurement.The feasibility of using computer to measure the plane members of the ship is verified.