铝合金压铸件X射线图像多任务分割方法研究  

Multi-task Segmentation Method for X-ray Images of Aluminum Alloy Die Castings

作　　者：杜旺哲 刘元铭 牛小淼 王涛 DU Wangzhe;LIU Yuanming;NIU Xiaomiao;WANG Tao(College of Mechanical and Vehicle Engineering,Taiyuan University of Technology,Taiyuan 030024;Engineering Research Center of Advanced Metal Composites Forming Technology and Equipment,Taiyuan University of Technology,Taiyuan 030024;National Key Laboratory of Metal Forming Technology and Heavy Equipment,Taiyuan University of Technology,Taiyuan,030024)

机构地区：[1]太原理工大学机械与运载工程学院,太原030024 [2]太原理工大学先进金属复合材料成形技术与装备教育部工程研究中心,太原030024 [3]太原理工大学金属成形技术与重型装备全国重点实验室,太原030024

出　　处：《特种铸造及有色合金》2025年第1期15-22,共8页Special Casting & Nonferrous Alloys

基　　金：国家重点研发计划资助项目(2021YFB3401000);山西省基础研究计划资助项目(20210302124321);河北省科技重大专项资助项目(23280101Z)。

摘　　要：为满足对铝合金压铸件内部缺陷和关键位置的精确分割需求,提出一种铝合金压铸件X射线图像多任务分割算法。设计一种关键位置边界感知模块和局部参数共享编码器,提高对低相关性任务的特征提取能力。开发一种交互式注意力模块,探究关键位置及其边界区域特征交互融合机制。结果表明,提出的多任务缺陷分割算法可以对X射线图像中的缺陷和关键位置区域同时进行分割,且分割精度在mIoU指标上分别为42.3和82.5,满足实际需求。In order to meet the requirements of high-accuracy segmentation of internal defects and key locations of alu⁃minum alloy die-casting parts,a multi-task segmentation algorithm for X-ray images of aluminum alloy die-casting parts was proposed.A key position boundary awareness module and a local parameter sharing encoder were designed to improve the feature extraction capability for low-relation tasks.An interactive attention module was proposed to improve the segmentation accuracy of key positions and their boundary areas.Results indicate that the proposed multi-task defect segmentation algorithm can simultaneously segment defect areas and key location areas in X-ray images with an mIoU accuracy up to 42.3 and 82.5,meeting the actual needs.

关 键 词：缺陷分割 铝合金压铸件 多任务分割 X射线图像 深度学习 

分 类 号：TG146.21[一般工业技术—材料科学与工程] TG249.2[金属学及工艺—金属材料] TP311[金属学及工艺—金属学]