基于深度学习的ESPI条纹图骨架线智能提取

Intelligent skeleton extraction of ESPI fringe pattern based on deep learning

为实现电子散斑干涉条纹图骨架线的自动提取,引入了深度学习技术。网络利用带噪声的电子散斑干涉条纹图和对应的骨架线图进行训练,然后将所有待测电子散斑干涉条纹图同时输入训练后的网络中,即可直接得到相应的骨架线图像。提取200幅实验电子散斑干涉条纹图的骨架线只需11.7 s,且对于存在断裂的电子散斑干涉条纹图,传统骨架线法会失效,利用训练完成的网络可获得完整骨架线。通过对该网络在模拟电子散斑干涉条纹图和实验电子散斑干涉条纹图上性能的评估,证实了基于深度学习的电子散斑干涉条纹图骨架线智能提取的可行性。

In order to realize automatic extraction of the skeleton of electronic speckle interference fringe pattern,deep learning technology is introduced. The network is trained by the noisy electronic speckle interference fringe patterns and the corresponding skeleton images. After the training,all electronic speckle interference fringe patterns to be measured are input into the network at the same time,and the corresponding skeleton images can be directly obtained.It only takes 11. 7 s to extract the skeletons of 200 experimental electronic speckle interference fringe patterns,and for the broken electronic speckle interference fringe pattern,the traditional skeleton method will fail,and the complete skeleton can be obtained by using trained network. Through evaluating the performance of this network on simulated and experimental electronic speckle interference fringe patterns,the feasibility of intelligent skeleton extraction of electronic speckle interference fringe pattern based on deep learning is verified.