深度学习的实例分割技术应用研究综述  

A Review of Application Research on Instance Segmentation Technology in Deep Learning

作　　者：孙鹏 孙传聪 蔡青 徐要要 甄珍 吴翠杨 邹田甜 Sun Peng;Sun Chuancong;Cai Qing;Xu Yaoyao;Zhen Zhen;Wu Cuiyang;Zou Tiantian(Department of Medical Devices,Shandong Drug and Food Vocational College,Weihai,Shandong 264200,China)

机构地区：[1]山东药品食品职业学院医疗器械系,山东威海264200

出　　处：《机电工程技术》2025年第3期1-6,63,共7页Mechanical & Electrical Engineering Technology

基　　金：山东省教育科学研究项目(19SR001);基于制图技术的校园数字化地图应用研究(2020094);威海市科技智库调研课题(人工智能提高威海旅游业个性化体验与服务质量的研究)。

摘　　要：为促进深度学习实例分割算法的应用研究,对其应用情况做系统的梳理总结,研究了基于深度学习的图像实例分割方法在各种常见的图像分割任务的应用进展。首先,介绍基于深度学习的图像实例分割技术现状;随之,分析和比较常用的数据库;然后,针对不同的应用场景,总结了各类实例分割方法在医学影像诊断、自动驾驶、视频监控等领域的应用情况。结果表明:随着深度学习的日趋成熟,基于深度学习的图像实例分割方法已经被广泛应用于各行各业中,也已成为该领域新兴研究热点,结合实例分割技术当下存在的不足,提出可行的解决方案,并展望了实例分割技术的发展未来。To promote the application research of deep learning instance segmentation algorithms,a systematic review and summary of their applications are conducted,and the application progress of deep learning based image instance segmentation methods in various common image segmentation tasks is studied.Firstly,the current status of image instance segmentation technologyn is introduced based on deep learning.Subsequently,the used databases are analyzed and compared commonly.Then,based on different application scenarios,the application of various instance segmentation methods in fields such as medical image diagnosis,autonomous driving,and video surveillance is summarized.The results indicate that with the increasing maturity of deep learning,image instance segmentation methods based on deep learning are widely applied in various industries and have become emerging research hotspots in this field.Based on the current shortcomings of instance segmentation technology,feasible solutions are proposed,and the future development of instance segmentation technology is prospected.

关 键 词：数据库 遥感图像 人脸识别 医学影像 自动驾驶 文字识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP751[自动化与计算机技术—计算机科学与技术]