图像识别技术在铸件缺陷检测中的应用

打开文本图片集

Abstract：This paper focuses on the problems of low eficiency and inconsistent standards of manual inspection of traditional casting surfacedefects，anddesigns asetof intelligent inspection system based onmachine vision. Bycomparingandanalyzing the mean filter，median filterand Gaussian filter，the median filterisselected as the preprocessing method.Intheresearchofsegmentationalgorithm，the improved U-Netnetwork iscompared withadaptive thresholdsegmentation，Cannyoperatorandregiongrowthmethod，andthedvatagesoftheimprovedU-Netntworkin objectintegrityandedgeclarityareverifed.Thesystemhasproveditstechnicaladvantagesintestingccuracyfiency androbustnessin practical industrialapplications，which provides a practicalreference for the intellgentupgradingof foundry industry.

Keywords：castingsurface defects; median filtering;improved U-Netnetwork;deep learning;image recognition

0 引言

铸件表面缺陷检测是保障铸件质量的关键环节，传统检测方式主要依赖工人目视检查和手动测量，这种方法存在检测效率低、判断标准不统一、误检漏检率高等问题。（剩余4584字）