High-speed three-dimensional (3D) shape measurement has become a very important technology in industrial manufacturing, motion detection and other scientific research. Although there are some methods to measure 3D surface patterns, it is still difficult to accurately measure the rapidly changing 3D high-speed scenes. Multi-frequency phase unwrapping usually uses a combination of noisy fringe images with different fringe frequencies for phase unwrapping, which has high accuracy and reliability. Benefiting from the success of deep learning in the field of computer vision in recent years, we combine multi-frequency phase-shifting and phase unwrapping with deep learning, and propose the high-speed 3D shape measurement from noisy fringe images using deep learning. Compared with traditional methods, this method can achieve more convenient and robust phase retrieval at high speed. Based on a good training model, the deep learning neural network can directly achieve the corresponding high-quality phase results after extensive learning of the data set collected at high speed. The experimental results demonstrate that this method can achieve 3D shape of the measured object with an accuracy of about 51μm at the camera frame rate of 700 frames per second.
Aiming at the serious noise and low signal-to-noise ratio of the fringe images obtained by industrial cameras under high-speed projection, a high-speed fringe projection measurement method based on convolutional neural network is proposed, which can achieve high-quality phase recovery and high-precision three-dimensional reconstruction in high-speed scenes. Using the designed convolutional neural network, the noise fringe images obtained at high frame rate and the wrapped phase images recovered by the traditional 12-step phase-shifting method at low frame rate are input into the convolutional neural network for training. After learning the mapping relationship between a large number of noise fringe images in the data set and the corresponding high-quality wrapped phase, a trained network model is obtained. And using this model, the high-quality wrapped phase information can be directly recovered from the input noise fringe images. The experiment results demonstrate that the method proposed in this paper can achieve with an accuracy of about 32μm through three noise fringe images at the camera frame rate of 700 frames per second.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.