基于BP神经网络的镁熔液第一气泡图像检测

郭宏晨; 许四祥; 孙杰; 翟泽

doi:10.15980/j.tzzz.2013.10.002

轻有色合金 | 浏览量 : 0 下载量: 78 CSCD: 4

导出
分享
收藏
专辑

基于BP神经网络的镁熔液第一气泡图像检测
Image Detection of the Initial Bubble in Magnesium Melt Based on BP Neural Network
2013年33卷第10期页码：973-975
纸质出版日期： 2013 ，
DOI： 10.15980/j.tzzz.2013.10.002

扫描看全文

[1]郭宏晨,许四祥,孙杰,翟泽.基于BP神经网络的镁熔液第一气泡图像检测[J].特种铸造及有色合金,2013,33(10):973-975.

Guo Hongchen, Xu Sixiang, Sun Jie, et al. Image Detection of the Initial Bubble in Magnesium Melt Based on BP Neural Network[J]. Special Casting & Nonferrous Alloys, 2013,33(10):973-975.
[1]郭宏晨,许四祥,孙杰,翟泽.基于BP神经网络的镁熔液第一气泡图像检测[J].特种铸造及有色合金,2013,33(10):973-975. DOI： 10.15980/j.tzzz.2013.10.002.

Guo Hongchen, Xu Sixiang, Sun Jie, et al. Image Detection of the Initial Bubble in Magnesium Melt Based on BP Neural Network[J]. Special Casting & Nonferrous Alloys, 2013,33(10):973-975. DOI： 10.15980/j.tzzz.2013.10.002.

摘要

针对复杂背景下镁熔液第一气泡的图像检测

提出一种基于BP神经网络的弱小目标检测算法。首先利用BP神经网络的函数逼近特性对原始图像进行背景估计

再将原始图像和背景估计图像进行差分重构

然后对重构后的图像进行帧差

最后进行自适应阈值处理并结合形态学开运算检测到弱小气泡。结果表明

利用BP神经网络的算法能精确地识别第一气泡

为图像处理技术应用于镁熔液含氢量检测提供了一种新方法。

Abstract

A detection algorithm was presented for weak and dim target on magnesium melt based on the BP neural network

which aimed at image identification of the initial bubble under complex background.At first

the function approximation characteristic of BP neural network was used to estimate the background of original image.Then

the reconstructed images were obtained by subtracting the background from original image.And then

the frame difference was used for the reconstructed images.At last

the weak and dim target was detected by using self-adaptive threshold and morphological opening operation.The results show that BP neural network algorithm is effective for the initial bubble identification

and it can provide a new method for detection of hydrogen content in magnesium melt based on image processing technology.