基于轻量级Unet架构的UWF图像的凹陷深度检测方法  

A Method for Detecting Depression Depth in UWF Images Based onLightweight Unet Architecture

作　　者：魏岸若 WEI Anruo(Chongqing Industry&Trade Polytechnic,Chongqing 408000)

机构地区：[1]重庆工贸职业技术学院,重庆408000

出　　处：《软件》2024年第8期18-21,共4页Software

基　　金：重庆市教委科技项目(KJQN202203605);重庆工贸职业技术学院科研项目(ZR202111)。

摘　　要：凹陷深度检测是UWF图像处理的关键,但是当前UWF图像的凹陷深度检测存在缺陷,在实际中检测误差比较大,检测结果置信水平比较低,无法到达预期的检测效果。因此,本文提出基于轻量级Unet架构的UWF图像的凹陷深度检测方法。采用各向异性滤波法对UWF图像进行平滑处理,利用轻量级Unet架构对UWF图像进行语义分割,提取UWF图像凹陷区域,根据图像灰度与深度的线性关系计算UWF图像的凹陷深度,实现基于轻量级Unet架构的UWF图像的凹陷深度检测。实验证明,应用本文方法,UWF图像的凹陷深度检测误差得到了有效降低,置信水平得到了明显提升,轻量级Unet架构在UWF图像的凹陷深度检测领域具有良好的应用前景。Depression depth detection is a key aspect of UWF image processing,but there are shortcomings in the current detection of depression depth in UWF images.In practice,the detection error is relatively large,and the confidence level of the detection results is low,which cannot achieve the expected detection effect.Therefore,this article proposes a depression depth detection method for UWF images based on a lightweight Unet architecture.The anisotropic filtering method is used to smooth the UWF image,and the lightweight Unet architecture is used for semantic segmentation of the UWF image.The concave regions of the UWF image are extracted,and the concave depth of the UWF image is calculated based on the linear relationship between image grayscale and depth,achieving concave depth detection of UWF images based on the lightweight Unet architecture.Experimental results have shown that the method proposed in this paper effectively reduces the detection error of indentation depth in UWF images and significantly improves the confidence level.The lightweight Unet architecture has good application prospects in the field of indentation depth detection in UWF images.

关 键 词：轻量级Unet架构 UWF图像 凹陷深度 检测 各向异性滤波法 语义分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]