摘要
针对固体火箭发动机喷管喉径的精确测量问题,提出了一种基于机器视觉的方案,采用了基于深度学习的图像分割技术和单目测量技术。通过平行面激光得到喷管喉部的投影图像,对获得的图像利用图像语义分割网络模型进行提取,进而得到准确的喷管喉部区域,对该喉部区域通过单目测量技术完成喷管喉径的测量。通过实验测量了不同条件下的发动机喷管喉径,验证了方法的可行性,结果表明,相比于传统测量方法,在精度、效率和稳定性上有了显著提升。该方法对于固体火箭发动机的性能测试具有重要意义,在实际工程中具有良好应用前景。
Aiming at the task of accurately measuring the nozzle throat diameter of solid rocket motor,a measurement system based on machine vision was proposed.The monocular measurement technology and the image segmentation based on deep learning were applied in this system.The projection image of nozzle throat was obtained by a parallel plane laser source,and the image was extracted by a semantic segmentation network based on improved deep learning to obtain the nozzle throat region.Finally,the nozzle throat diameter was calculated by monocular measurement technology.Through the experiments,real nozzle throat diameter in various conditions was measured and feasibility of the method was verified.The results show that compared to the traditional method,the accuracy,efficiency and stability of this method have been significantly improved.The method is of great significance for the performance test of solid rocket motor and has good application prospect in practical engineering.
作者
孙旭阳
沈飞
谢俊彦
SUN Xuyang;SHEN Fei;XIE Junyan(Xi'an Aerospace Propulsion Testing Technology Research Institute,Xi'an 710025,China)
出处
《固体火箭技术》
CAS
CSCD
北大核心
2023年第1期119-127,共9页
Journal of Solid Rocket Technology
关键词
喷管喉径
测量
机器视觉
图像处理
深度学习
nozzle throat diameter
measurement
machine vision
image processing
deep learning
