Research On Multi-Frequency Phase Shift 3D Imaging For AOI

Printed circuit board(PCB)is the basis of electronic products.Currently,the size of electronic components is becoming smaller,which is increasingly bringing difficulty to PCB manufacturing quality control and defect detection.Automatic optical inspection(AOI)technology is widely adapted in the quality control of PCB manufacturing products due to its non-contact,high-speed,and high-precision characteristics.3D AOI technology,based on structured light three-dimensional imaging,is enable to accurately restore the complete three-dimensional shape of the circuit board.Compared with 2D's,the more comprehensive detection function and higher stability makes 3D AOI becoming the mainstream choice of electronics manufacturers.This paper mainly studies the high-precision,high-robust multi-frequency phaseshift structured light method,as well as the key issues such as phase analysis,phase unwrapping and high dynamic range imaging for AOI scenes.The main contents of the study are summarized as follows:(1)Multi-frequency phase shift optimization.Affected by sensor noise,out-offocus,or low reflectivity of the measured object in actual measurement,multifrequency heterodyne easily leads to accumulation of phase errors and large jump errors.In this paper,three methods of multi-frequency phase unwrapping are discussed and error models are established respectively.Consequently,a phase optimization method based on frequency selection is proposed,which significantly suppresses the occurrence of phase jump errors,so as to obtain high-precision,high-stability absolute phase.(2)Absolute phase retrieval based on multi-view geometry.The optimized multifrequency phase shift method can reduce the number of fringe patterns as much as possible.However,the measurement target with high precision and complex illumination requires the fringe frequency to be high enough,which means more reference fringe projection.In order to further reduce the number of fringe projections,this paper introduces the principle of stereo vision to solve the phase ambiguity problem.Consequently,a fast phase unwrapping method based on multi-viewpoint phase consistency and epipolar constraints is proposed,which improves the reliability of point cloud while taking into account the efficiency of scanning.(3)Multi-camera point cloud fusion.Due to the occlusion and large reflectivity range of printed circuit board components,the quality of depth data obtained by the camera at different angles is various,and it even leads to data missing in a certain viewing angle.Based on the orderliness of point clouds from structured light method,this paper proposes a multi-view point cloud fusion by phase quality guidance,which achieves multi-view information complementation and improve the stability of the point cloud.In order to ensure the speed of 3D imaging calculation,the proposed algorithms are designed in parallel and integrated in a software based on GPU,which is capable of calculating 10 million point clouds per second.It is proved that the accuracy of 3D sensor is 5?m through gauge block measurement.Overall,the 3D sensor built in this paper can restore the real profile of the PCB surface with high precision,stability and efficiency,which provides a reliable data for 3D AOI.