改进卷积神经网络的船舶纹理图形分割方法

Improved convolutional neural network method for ship texture image segmentation

船舶从生产到投入使用,在作业中难免会因为焊接节点的设计问题、货物装卸的操作不规范以及诸如线性尺度较大、吃水较多而受风、流等外部环境的影响而对船体结构造成腐蚀,最终导致船体外部出现裂纹,给船舶的航行带来安全隐患。因此加强关于船舶裂纹的排查,是保证船舶安全行驶的关键。本文从改进卷积神经网络入手,以工程船为例,通过对其船舶纹理进行图形分割研究,提升智能检测的精准度,从而为运维检修带来一定的帮助,并为海上作业的安全性检测提供理论支持。

From production to putting into use,it is inevitable that ships will suffer from corrosion to the hull structure due to the design issues of welded joints,non-standard cargo handling operations,and external environments such as large linear scales and high drafts,resulting in cracks on the exterior of the hull,which pose potential safety hazards for shipping trips.Therefore,strengthening the inspection of ship cracks is the key to ensuring the safe operation of ships.This article will attempt to improve the convolutional neural network,take engineering ships as an example,and conduct graphic segmentation research on their ship texture,aiming to improve the accuracy of intelligent detection,thereby bringing some help to relevant operation and maintenance personnel in the maintenance work,and providing theoretical support for the safety detection of offshore industry.