The Design Of The Automatic Draft Detection System Based On Image Processing

Automatic draft detection system based on image processing is a new technology that introduces machine vision into the field of draft survey.The traditional draft survey is basically based on the human eye observation method.However,due to the subjective factors of the method,the unrecorded observable process data,and few observation points,it is easy to create a dispute among the ship,the port and the third party.The result of the system that can be traced later is only related to the draft video,includes not only the final reading but also the continuous fluctuation of the draft reading.This technology converts the fluctuating draft video into a fluctuation curve,which improves the objectivity,accuracy and scientificity of the ship’s draft reading.This paper first studies the framework design of the automatic draft detection system,and designs the hardware and software framework of the system around image acquisition,wireless transmission and image processing module.The hardware platform includes an image acquisition system,a wireless transmission module,a stable acquisition bracket,and the software platform includes a development framework,a video decoding library,a deep learning framework,and an image processing library.Then the ship’s waterline detection algorithm is studied from the perspectives of edge detection,color feature and deep learning.The edge detection principle and the local derivative operator are used to detect the waterline.In the color feature perspective,the segmentation algorithm based on RGB histogram,the segmentation algorithm based on RGB space vector and SVM,the waterline edge detection algorithm based on RGB vector space are used to detect the waterline;In the deep learning perspective,the Faster R-CNN algorithm is used to detect the waterline,which is much better than the traditional image processing method.After that,this paper studies the draft character detection algorithm which contains the character detection algorithm based on deep learning and the outer frame of character position correction algorithm and the character detection algorithm based on template matching for difficult-to-recognize characters.Finally,the draft reading algorithm is studied.The least squares fitting algorithm and cubic spline interpolation algorithm are used to approximate the waterline.The oblique character and waterline spacing algorithm and the overlooking image correction algorithm are proposed for the non-standard draft image.Finally,the automatic draft detection system with outlier algorithm is realized.Afterwards,a large number of draft video tests are performed on the software to analyze its accuracy.Based on the overall scheme of ship draft,this paper studies waterline recognition algorithm,character recognition algorithm and draft reading algorithm,and completes the overall software implementation and verification.Experiments show that the image-based automatic draft detection system can effectively get the right reading of the draft video and display the fluctuation curve.The accuracy of the software reading is in line with expectations.