基于U^(2)-Net的痕迹物证图像分割方法的研究  

Research on Image Segmentation Method of Trace Evidence Based on U^(2)-Net

作　　者：陈雅琪 刘丹[1] CHEN Yaqi;LIU Dan(School of Public Security Information Technology and Intelligence,China Criminal Police College,Shenyang,Liaoning 110854,China)

机构地区：[1]中国刑事警察学院公安信息技术与情报学院,辽宁沈阳110854

出　　处：《自动化应用》2024年第11期35-37,41,共4页Automation Application

摘　　要：针对痕迹物证检测方法精度不高、模型体积较大、部署困难等问题,设计了一种简单且强大的深层网络架构U^(2)-Net,用于研究痕迹物证图像分割方法。该架构特别设计了用于处理复杂背景下的图像分割任务。U^(2)-Net采用了一种新颖的嵌套U型结构,通过多尺度的特征提取和深层次的信息融合,能精确地识别和分割痕迹物证图像中的关键对象。该方法在痕迹物证的识别和提取方面展现了优越的性能,特别是在处理高度复杂或模糊不清的图像时,能有效提高分割的准确性和细节恢复能力。A simple and powerful deep network architecture U^(2)-Net was designed to address the issues of low accuracy,large model size,and difficult deployment in trace evidence detection methods.This architecture was used to study image segmentation methods for trace evidence.This architecture is specifically designed to handle image segmentation tasks in complex backgrounds.U^(2)-Net adopts a novel nested U-shaped structure,which can accurately identify and segment key objects in trace evidence images through multi-scale feature extraction and deep level information fusion.This method demonstrates superior performance in the recognition and extraction of trace evidence,especially in processing highly complex or blurry images,which can effectively improve the accuracy of segmentation and the ability to restore details.

关 键 词：图像分割 U^(2)-Net模型 复杂背景 痕迹物证 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]