摘要
建立了一种基于YOLOv3的新型卷积神经网络模型,用于分割QR码在图像中的位置区域。模型在取得较好定位效果的同时,能够在运算能力有限的计算机上保持实时性。并对现有基于比例关系的QR码定位算法作出改进,降低定位比例条件,加入评分系统,提高了算法对污损QR码的定位能力。
A small convolution neural network model based on YOLOv3 is designed to get the position of the QR code in an input image. The model is able to locate the QR code area in real-time even if computational resources are limited. The existing localizing algorithm has been improved by simplifying the position proportional conditions and applying a scoring system. Practically,the developed algorithm is able to well locate the broken QR code.
作者
余先涛
秦岩
YU Xiantao;QIN Yan(School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China)
关键词
卷积神经网络
比例关系
QR码定位
污损QR码
目标检测
convolution neural network
proportion relationship
QR code localization
broken QR code
object detection
