CNN-based fringe analysis has solved various challenging tasks in optical metrology. However, these CNN-based fringe analysis methods perform poorly in high-frequency features because low-frequency features dominate and are easy to learn. In this work, to fully extract the high-frequency features, we present a comparative analysis of off-the-shelf image edges, observing that these image edge priors fail to characterize the high-frequency component of the modulation image. Based on this, we propose fusing modulation soft-edge prior into the fringe analysis network for accurate 3D measurements of complex geometry. Instead of off-the-shelf image edges acquired by binarization preprocessing, the soft edge maximally preserves the modulation features while avoiding false modulation edges. The Soft-Edge Assisted Fringe Analysis (SEAFA) network can maintain high-quality overall measurements while improving the ability to extract the high-frequency component of complex geometry. Extensive experiments demonstrate that the SEAFA network can achieve more accurate 3D measurements than the previous CNN-based methods, especially for complex geometry.
Optical three-dimensional (3D) measurement methods have been widely used in many important fields due to the advantages of non-contact and high efficiency. Among these methods, fringe projection profilometry (FPP) is the most promising method because of its high accuracy. However, the 3D measurement area of FPP is reduced by the misalignment between the field of views (FOVs) of the camera and the projector. The misalignment occurs because the spatial positions of the projector and the camera are different, which is inevitable in the fringe projection system (FPS) only including a single camera. The misalignment leads to two cases: the first case is that the regions without fringes are captured by the camera, and the 3D shape of the regions cannot be restored. The second case is that the regions with fringes are not captured by the camera. In the second case, the 3D shape of the regions can be restored when the regions are captured by the camera. This paper presents a multi-view FPS (MVFPS), which consists of a camera, a projector, and a scanning mirror. The FOV of the camera can be changed by rotating the scanning mirror, which can include the regions in the second case in the camera's FOV. First, multiple rotation angles of scanning mirrors are preset in advance, and the internal and external parameters of the projector and camera corresponding to each angle are calibrated. Then, a set of fringe patterns are projected for reconstructing the 3D shape at each angle. Finally, all 3D shapes obtained at different angles are combined as a whole 3D measurement result by the point cloud registration algorithm. Experiments show the proposed MVFPS, and the 3D measurement area of a plaster statue is expanded after combining the multiple 3D shapes at different angles of the scanning mirror.
Stereo-assisted fringe projection profilometry (FPP) has been broadly applied to 3-D measurements because of its accuracy and stability advantages. However, absolute phase unwrapping requires an ever-increasing number of cameras to continuously enlarge the measurement depth volume without spatial uncertainty, which inevitably reduces calibration reliability and system flexibility. This paper proposes a multithread monocular FPP to implement long-depth-volume 3-D measurements. For achieving the multithread FPP technique, a galvanomirrorbased camera-projector system with variable extrinsic parameters is developed. The typical combination of a camera and a projector is virtually converted into multiple monocular active vision systems through fast viewpoint switching. Based on the spatio-temporal correlation of multiple viewpoints, each wrapped phase can obtain a possible fringe order map. Subsequently, with the geometric constraints of the system, the wrapped phase maps extracted from image sequences are converted into relative phase distributions. Combining the possible fringe orders and the relative phase map, the absolute phase distribution is determined by choosing the correct phase map from all potential candidates. Experimental results demonstrate that the proposed method could accomplish long-depth-volume 3-D measurements with a measuring accuracy of 0.02 mm.
The phase coding approach has been extensively applied to three-dimensional (3-D) shape measurement, in which the fringe order determined by the phase coding patterns is utilized to unfold the wrapped phase calculated by the phase-shifted patterns. However, due to inevitable random noises and projector/camera defocus, the fringe order can not be accurately aligned with the wrapped phase, thereby introducing the abnormal jump phase errors. To this end, this paper presents a simple and practical approach with no accuracy loss and no extra patterns to remove the abnormal jump phase errors. Firstly, the designed phase coding patterns are pre-shifted by half fringe period relative to the phase-shifted sinusoidal patterns for ensuring that the decoding order changes and the 2π phase jumps of the wrapped phase are staggered. Then, by analyzing the intensity variations of the phase-shifted sinusoidal patterns, a virtual binary code is generated, which is essentially shifted by a quarter of the fringe period relative to the phase-shifted sinusoidal patterns. Subsequently, the decoding order is divided into three subregions by referring to the distribution of the virtual binary code and the wrapped phase. After completing the segmentation, a regional correction operation (RCO) is implemented. For each subregion, the decoding order remains unchanged or is adjusted to get the correct fringe order. As a result, the interval of the wrapped phase can be accurately identified at the pixel level for phase unwrapping successfully. Benefitting from the simple pre-shifted strategy used for the phase coding patterns and the virtual binary code extracted by the phase-shifted patterns, the proposed approach can completely avoid the abnormal jump phase errors without projecting additional patterns. Moreover, compared with the conventional median filtering approach, the proposed approach has the capability to preserve sharp edges, and thus it is noninvasive. The measurement results demonstrate the reliability and accuracy of the proposed approach in measuring different objects containing intricate surface profiles.
Fringe projection profilometry (FPP) based on the binary defocusing technique (BD) shows great potential in high-speed 3-D imaging. Due to the constant defocusing degree, existing binary defocusing operations require adopting similar-wavelength fringe patterns, thereby forming a long imaging sequence in multi-frequency temporal phase unwrapping (TPU). In this paper, we propose a few-pattern defocusing FPP for efficient and accurate 3-D imaging. The imaging sequence consists of only 6 hybrid images, namely 2 unit-frequency ramp images, 2 low-frequency, and 2 high-frequency sinusoidal fringe images. Combining unit-frequency ramp and low-frequency fringe images, the unknown average intensity and fringe orders of fringe images can be determined. Consequently, the final absolute phase map can be extracted from the high-frequency fringe images. Moreover, a kernel-optimized dithering technique is presented to generate the projected patterns of hybrid images. In this dithering technique, a dynamic kernel and a dual-objective function ensure the optimal binarization of defocused images with different grayscale variations. Experiment results verify the proposed few-pattern defocusing FPP achieves efficient 3-D imaging with a measurement accuracy of 0.02 mm.
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.