摘要
针对复合材料孔隙含量的光学显微图像法统计过程中出现的统计试样多、统计周期长、人工统计结果存在差异等问题,将计算机图像处理技术与神经网络算法相结合,基于复合材料孔隙人工统计方法形成计算机识别与统计算法,利用大量光学显微图像数据样本的标注与学习结果,开发了基于图像处理技术和深度学习网络技术的复合材料孔隙含量识别与统计系统。同时,为使该系统的适用性更强,开发了有监督的数据更新模块。在复合材料孔隙统计中的试验结果表明,与传统人工统计方法相比,深度学习网络算法能够更好地识别孔隙特征,孔隙统计结果的相对误差小于±10%;并在较大规模数据试验中取得了更好的效果,极大地减少了传统人工统计过程中的诸多弊端,优势更强。
For the issue of many statistical samples,long statistical period and differences in manual statistics in the statisticd process of optical micrographs of composite pore content,computer image processing technology and neural network algorithm are combined to form a computer recognition and statistical algorithm based on the manual statistical method of composite pore.The composite pore content identification and statistical system based on image processing and deep learning network technology is developed based on the results of labeling and learning of a large number of optical micrographs data.At the same time,in order to make the system more applicable,a supervised data update module is developed.The experimental results in the composite pore sta-tistics show that the deep learning network algorithm can identify pore characteristics better than the traditional manual statistical method,and the relative error of results is less than±10%.In the larger scale data experi-ment,it achieves better results,and gteatly reduces many traditional drawbacks in the manual statistical process,its advantages are stronger.
作者
陈健
肖鹏
CHEN Jian;XIAO Peng(COMAC Shanghai Aircraft Manufacturing Co.,Ltd.,Shanghai 201324,China)
出处
《测控技术》
2024年第3期22-27,69,共7页
Measurement & Control Technology
关键词
复合材料
显微图像
孔隙率
统计
神经网络
composite materials
micrographs
porosity
statistic
neural network
