摘要
机器视觉检测任务通常需通过图像拼接获取高质量、宽视野的被检对象,图像拼接融合是实现机器视觉图像拼接的关键步骤,该文系统评述常见机器视觉检测图像拼接融合技术,包括基于平滑过渡的机器视觉图像拼接融合技术、基于缝合主线的机器视觉图像拼接融合技术以及基于深度学习的机器视觉图像拼接融合技术等,阐述各技术方法的主要数学模型、工作机理以及性能特点,以及总结当前图像拼接融合技术先进方法与图像拼接配准技术值得关注的方向。
Machine vision detection tasks usually require high-quality,wide-field objects through image stitching.Image blending is a key step to achieve image stitching.This paper systematically reviews common image blending method in machine vision detection,including transition smoothing-based blending,optimal seam-based blending and blending based on deep learning,etc.The paper explains the main mathematical models,working theory and performance characteristics of each method,summarizes the current advanced image blending method,and proposes several research directions which deserve attention.
作者
刘桂雄
张瑜
蔡柳依婷
LIU Guixiong;ZHANG Yu;CAI Liuyiting(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640,China)
出处
《中国测试》
CAS
北大核心
2020年第1期1-6,共6页
China Measurement & Test
基金
广州市产业技术重大攻关计划(2018020300006)
广东省特检院科研项目(2020CY14)
关键词
机器视觉
图像拼接
图像融合
深度学习
machine vision
image stitching
image blending
deep learning
