Rongsheng Lu’s research while affiliated with Hefei University of Technology and other places

Stabilizing orthogonal phase bias is highly required in Mach-Zehnder interferometer (MZI) to achieve sufficiently high sensitivity and long-term stability. To this end, we demonstrated a stabilized MZI using dual-wavelength feedback and a geometric phase shifter (GPS). The two beams at the wavelengths of 632.8 nm and 532 nm, respectively, propagate almost in the common path in the MZI, so that the beams have the same sensitivity to ambient disturbance. The beam at 632.8 nm is used to sense the displacement, and the other one at 532 nm is used for feedback and orthogonal phase bias stabilization. The phase shift error of the two separate wavelengths can be compensated by a GPS consisting of a rotating half waveplate sandwiched between two static quarter waveplates. The results of the proof-of-concept experiment showed that the phase shift error between 632.8 nm and 532 nm was reduced to within (-0.08π, 0.09π) by the GPS, and can be further reduced by two orders of magnitude if more appropriate wavelengths or waveplates were selected. The stabilized MZI showed a displacement resolution of 5 nm, and was proved to be effective in suppressing the low-frequency noise less than 10 Hz.

Fringe projection contouring is a widely used technique in various three-dimensional (3D) reconstruction applications. However, achieving high reconstruction precision typically requires the use of numerous projected patterns, which limits its practicality in dynamic scenes. To address this challenge, we utilized defocused fringes to significantly enhance the projection frame rate of projectors, enabling adaptation to high-speed environments. In this paper, we propose what we believe is a novel method that integrates defocusing technology with composite fringes and an optimization approach for defocusing composite fringes. Notably, our method achieves high-precision 3D reconstruction using only five patterns under slight defocusing. Furthermore, owing to the multifrequency information inherent in composite fringes, each image can be independently utilized for 3D reconstruction, effectively increasing the frame rate of the reconstruction process. The experimental results demonstrate the effectiveness of our method in generating low-bit composite fringes, achieving reconstruction speeds of up to 500 fps. The proven efficacy and efficiency of our approach make it a promising solution for high-speed 3D reconstruction in dynamic environments.

The paper investigates the structural parameters of laser triangulation measurement system (LTMS) based on the utilization of commercial lenses, and offers a basis for choosing the most suitable configuration. Firstly, the object-image relationship model and optical sensitivity model of the LTMS are studied respectively. The Scheimpflug condition and transverse magnification of a given point on the axial field are used to reduce the number of unknown parameters in the model. Furthermore, the value of the magnification and resolution are preliminarily determined by means of focal length, sensor size and measurement range. Moreover, the optimum acceptance angle is concluded based on the maximum sensitivity of the LTMS. Finally, the selected parameters are evaluated by nonlinear error. The experimental results demonstrate that optimizing the structural parameters of the LTMS can elevate the nonlinearity to nearly twice its original value when compared to the structural parameters without optimization, under the identical hardware configuration. And selecting the acceptance angle based on maximum sensitivity has rationality in terms of the nonlinear error results.

Crack detection is an important task to ensure structural safety. Traditional manual detection is extremely time-consuming and labor-intensive. However, existing deep learning-based methods also commonly suffer from low inference speed and continuous crack interruption. To solve the above problems, a novel bilateral crack detection network (BiCrack) is proposed for real-time crack detection tasks. Specifically, the network fuses two feature branches to achieve the best trade-off between accuracy and speed. A detail branch with a shallow convolutional layer is first designed. It preserves crack detail to the maximum and generates high-resolution features. Meanwhile, the semantic branch with fast-downsampling strategy is used to obtain enough high-level semantic information. Then, a simple pyramid pooling module (SPPM) is proposed to aggregate multi-scale context information with low computational cost. In addition, to enhance feature representation, an attention-based feature fusion module (FFM) is introduced, which uses space and channel attention to generate weights, and then fuses input fusion features with weights. To demonstrate the effectiveness of the proposed method, it was evaluated on 5 challenging datasets and compared with state-of-the-art crack detection methods. Extensive experiments show that BiCrack achieves the best performance in the crack detection task compared to other methods.

Precise alignment of the system scan geometry is crucial to the reconstruction quality of computed laminography (CL). Different from computed tomography (CT), a challenging task for CL image reconstruction is to deal with rotational axes that are no longer symmetrical. In this paper, we take advantage of the fact that the difference between the inverse of the projected transverse coordinates of two points rotated about the axis of rotation under the CL scan geometry sums to zero when the projection angle differs by 180 deg. Based on this fact, the projection data under the scanned geometry with errors are aligned to the theoretical projection geometry. To validate this calibration method, a numerical printed circuit board (PCB) phantom was simulated. The results demonstrate that this method is capable of achieving high accuracy compared to the previous methods. The experimental results on printed circuit boards (PCB) demonstrate that the method can effectively improve geometric alignment accuracy of CL and obtain high-resolution reconstructed images.

With the continuous improvement of quality requirements for optical components, the detection of subsurface defects in optical components has become a key technology. However, there is a problem with existing detection techniques, which is that they cannot simultaneously and independently detect subsurface defects at the micrometer and nanometer levels. This article analyzes the scattering field model of subsurface scratches and conducts simulation experiments on the relationship between scattering light intensity and system aperture. Based on the simulation results, a dual channel experimental system with adjustable spot size was designed to achieve automated measurement of subsurface defects. The narrow channel was used to detect micrometer-level subsurface defects and the wide channel was used to detect nanometer-level subsurface defects. The experimental results verified the correctness of the simulation experiment. In order to improve the sensitivity of the system, we designed an aperture based on the scattering field distribution of surface and subsurface defects, which is used to block the interference signal on the sample surface and improve the signal-to-noise ratio of the subsurface defect signal. The experimental results show that this aperture plays an important role, and the detection sensitivity of the system reaches 100 nm. We used four algorithms for data processing and found that the IQR algorithm is most suitable for this system. Finally, the detection results were compared under different spot sizes, and it was found that small spot sizes have better detection effects on nanoscale subsurface defects. In practice, the spot size can be dynamically adjusted according to the detection needs to achieve the optimal configuration of detection speed and sensitivity.