Printed circuit board (PCB) is a crucial component of electronic products, and its mounting quality directly affects the stability and reliability of electronic products. Therefore, three-dimensional (3D) reconstruction of PCB is essential for accurately detecting solder joint defects and component mounting positions, etc., which is vital for ensuring PCB mounting quality. At present, the structured light method is widely used in PCB 3D reconstruction due to its simple operation, versatility, and high precision. However, the accuracy of PCB 3D reconstruction is compromised by two main factors. First, the varying reflectivity of the PCB surface can lead to abnormal and missing fringe phase information. Second, calibration errors of the structured light system can further affect accuracy. We propose a PCB 3D reconstruction method based on fringe phase characteristics, which effectively addresses the issue of missing 3D point clouds when measuring high dynamic range surfaces and improves the accuracy of 3D reconstruction.

The structured light system consists of a structured light generator and a camera. Initially, a multi-exposure image fusion rule based on adaptive weights is established, which calculates the weight map through sample images captured at different exposure time and synthesizes the fringe image sequences. The weight map is determined by two factors: the relationship between each pixel’s intensity value and the overall brightness of the current image, and the variation of the same pixel between neighboring exposure images. Next, the phase value is solved using an asynchronous multi-frequency phase shift method. To improve the calibration accuracy of the structured light system, we build a height reconstruction model based on the perspectives of the camera and the structured light generator. The optimal function is formulated by examining the offset between the measured value and the actual value of the gage blocks. Subsequently, the Levenberg-Marquardt algorithm is used to fit and optimize the 3D reconstruction parameters in the optimization function. Finally, the 3D reconstruction results of the proposed method are verified using selected dark and high reflectivity areas on the PCB surface, and the accuracy is confirmed with a standard ball.

Our method can effectively optimize the image intensity of the dark region and saturated region while preserving the structure of the fringe in the synthesized image. This significantly addresses the issue of fringe loss on the surface with complex reflectivity (Fig. 7). In addition, the accuracy and stability of surface phase calculation are enhanced by increasing phase shift steps in the main phase shift fringe sequence, which greatly influences phase unwrapping precision. Consequently, the proposed method can accurately reconstruct the 3D shape of the measured objects (Fig. 8) and the reconstruction rate of the samples is 98.3% (Table 3). The performance of the proposed method is significantly superior to the other three methods. In the precision evaluation experiment, the average diameter error of the standard ball measured by the proposed method is 0.0188 mm and the root mean square error (RMSE) is 0.0441 (Table 4). These results demonstrate that the proposed method can reduce the error in 3D reconstruction parameters using gage blocks to further optimize the parameters of the camera and the structured light generator. When compared to the PCB 3D reconstruction results of the triangular stereo model and phase-height model, the proposed method exhibits better accuracy and stability, leading to improved measured outcomes.

To improve the accuracy and reliability of PCB 3D reconstruction, we propose a method based on fringe phase characteristics. The main conclusions are as follows. Firstly, the multi-exposure image fusion method based on adaptive weights is used to synthesize the coded fringe images, and the phase field is then solved using the asynchronous multi-frequency phase shift method, enhancing the measurement success rate for high dynamic range and complex surface. This approach effectively addresses the issue of missing point clouds on high dynamic range surfaces. Secondly, the system calibration error, caused by camera error transmission and simplification of the structured light generator model, is reduced using gage block constraints to optimize 3D reconstruction parameters. Thirdly, experimental results demonstrate that the proposed method can accurately reconstruct the 3D shape of high dynamic range and complex surface and significantly reduce the problem of point cloud loss. In the precision evaluation experiment, the proposed method measured the average diameter of a standard ball with a diameter of 20.0148 mm, achieving an average diameter error of 0.0188 mm. Compared to the triangular stereo model and the phase-height model, the proposed method offers superior precision and stability in PCB measurement results.