Large optical element detection method based on mechanical arm and ultrahigh-resolution optical coherence tomography imaging system

Abstract

The invention relates to a large optical element detection method based on a mechanical arm and an ultrahigh-resolution optical coherence tomography (OCT) imaging system. According to the method, the ultrahigh-resolution OCT technology and the multi-joint mechanical arm are combined, and high-precision and large-range detection of large optical elements is achieved. By means of the high-resolution imaging capacity of the OCT system, micron-sized defects inside and outside the optical element can be effectively recognized in a non-contact mode, and by means of the flexible movement capacity of the six-axis mechanical arm and a three-dimensional image splicing algorithm, the three-dimensional shape of the complex curved surface optical element can be restored and analyzed, so that the detection efficiency and accuracy are improved. The method has important application value in the fields of optical element quality control and industrial detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical device detection, and particularly relates to a large optical element detection method combining robotic arm motion control and ultra-high resolution optical coherence tomography (OCT) imaging, which is applicable to the detection of micron-level defects on the surface and inside of components such as LED lenses and precision optical lenses. Background Art

[0002] In the modern optical industry, the quality control of large optical elements is of crucial importance. Traditional detection methods such as manual visual inspection and contact measurement have problems such as low efficiency and insufficient accuracy. Optical coherence tomography (OCT) technology, as a non-contact and high-resolution imaging technology, has been widely applied in the biomedical field. However, when applying OCT technology to the detection of large optical elements, challenges such as limited imaging range and insufficient scanning flexibility are faced. The OCT scanning system assisted by a six-axis robotic arm can effectively expand the imaging range and improve scanning flexibility, but in existing research, the OCT axial resolution of the robotic arm is usually above 8 μm, making it difficult to meet the detection requirements for tiny defects. Summary of the Invention

[0003] Aiming at the defects of the existing technology, the present invention proposes a joint detection system and method based on a robotic arm and ultra-high resolution OCT. The system adopts a spectral domain OCT (SD-OCT) structure, with a light source bandwidth of 165 nanometers and an axial resolution of 1.9 micrometers; a six-degree-of-freedom robotic arm provides high-precision motion control, and through three-dimensional space calibration, the accurate mapping of the OCT probe pose and imaging data is achieved.

[0004] To achieve the above object, the technical solution of the present invention is as follows: A large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system, comprising the following steps:

[0005] (1) Build a spectral domain optical coherence tomography system with a light source center wavelength of 850 nanometers and a bandwidth of 165 nanometers;

[0006] (2) Fix the OCT probe at the end of the six-degree-of-freedom robotic arm through an adapter, with the robotic arm repeat positioning accuracy of ±0.03 millimeters and a working radius of 914 millimeters;

[0007] (3) Calibrate the coordinate system at the end of the robotic arm and the OCT imaging space using a three-dimensional calibration target, and calculate the rotation matrix and translation offset through singular value decomposition;

[0008] (4) Verify the system resolution through microsphere imaging experiments, test the system's planar imaging ability using mirror scanning, and evaluate the detection accuracy in combination with defect sample experiments.

[0009] (5) Plan the scanning path of the robotic arm and move step by step along the surface of the optical element with a preset overlap rate;

[0010] (6) Acquire local 3D OCT images and automatically stitch multiple groups of images based on the robotic arm pose data to generate a large field of view detection result of the optical element.

[0011] Among them, the OCT system in step (1) adopts spectral domain OCT technology, and the axial resolution in air is about 1.9 μm.

[0012] Among them, the 3D calibration target in step (3) is a three-peak structure printed by light curing, and the coordinate system transformation matrix is calculated by extracting the peak coordinates under multiple poses of the robotic arm.

[0013] Among them, the motion path of the robotic arm is planned based on the geometric shape and detection requirements of the optical element, and large-range 3D topography reconstruction is achieved by adjusting the overlap rate parameter.

[0014] Among them, the OCT 3D image is used to identify micro defects, including scratches, stains, and polishing protrusions, as well as precise measurement of the defect size and position.

[0015] First, use a 3D calibration target to calibrate the coordinate system of the robotic arm end and the OCT imaging space.

[0016] Furthermore, the specific method of this step is as follows: We place the sample on the table that remains stationary in the base coordinate system of the robotic arm, and then program the robotic arm to move multiple small steps (>3) on the x, y, and z axes in the end effector space respectively. Each time it moves, the OCT maintains the same rotation angle to obtain 3D images at each position.

[0017] Then, through a custom MATLAB script, the coordinates of the three peaks in the OCT space can be obtained and averaged. If we assume that the initial value of the average peak coordinates in the OCT space is P0, and the average coordinates of the i-th step calculated are P i ,.

[0018] Furthermore, the coordinates in the initial OCT space can be obtained through the following formula:

[0019] P′ i = P i - P0

[0020] Since the robotic arm moves in the end effector space to obtain imaging data, the coordinates Pr of the end effector in the initial end effector space can be given according to the moving distance. i , if R e2o represents the rotation matrix from the robotic arm end effector to the OCT space, then the following relationship can be obtained:

[0021] [P′1, P′2, P′3, …, P′n] = R e2o *[-Pr1, -Pr2, -Pr2, …, -Pr n

[0022] Secondly, the rotation matrix R can be obtained by performing singular value decomposition (SVD) based on the pseudo-inverse matrix method e2o .

[0023] Then, the translational offset between the end-effector space and the OCT space is calculated. Further, the robotic arm is programmed to aim at the above target in different directions according to the estimated initial offset t0, and then the initial transformation from the end-effector space of the robotic arm to the OCT space can be expressed as:

[0024]

[0025] At each pose of the robotic arm, the pose of the robotic arm and the OCT image are recorded, and the average peak coordinates in the OCT space can be calculated. Since the model is stationary in the base coordinate system of the robotic arm, the peak coordinates calculated at different poses match each other.

[0026] Further, iteration can be performed on the x, y, and z offsets of t0 to calculate the new coordinate Pb in the base coordinate system of the robotic arm i , and when the calculated coordinate difference is minimized, the measured translational offset can be obtained, expressed as:

[0027]

[0028] where is the average value of all coordinates at different robotic arm poses. In this way, the final transformation from the end-effector of the robotic arm to the OCT space can be obtained to navigate the OCT probe for high-resolution wide-field scanning.

[0029] After calibration, during detection, the robotic arm scans the surface of the optical element step by step along a preset path, collects local three-dimensional OCT images and automatically stitches them, and finally generates a complete large-field-of-view detection result. This method optimizes the imaging depth by dynamically adjusting the probe height, combines the overlapping area registration algorithm to eliminate stitching errors, and can detect surface scratches, polishing protrusions, and micron-sized contamination particles simultaneously.

[0030] Compared with the prior art, the advantages of the present invention are as follows:

[0031] ​(1) Due to the broadband light source, unique optical path system, and spectrometer design, the axial resolution of this spectral domain OCT system in air can reach 1.9 μm, exceeding the resolution of general similar systems. It can observe impurity contamination of 2 - 3 μm inside optical components, which is difficult for other similar systems to achieve.

[0032] (2) Cooperating with a six - axis collaborative robotic arm significantly improves the flexibility and degrees of freedom of the system. Combining with the point cloud registration algorithm, it overcomes the limitation of the small field of view of traditional OCT and realizes the detection of large - scale optical components. Brief Description of the Drawings

[0033] Figure 1 is the overall structural schematic diagram of the system of the present invention;

[0034] Figure 2 is the schematic diagram of the three - dimensional overlapping area during the robotic arm scanning

[0035] Figure 3 is the three - dimensional calibration target, its OCT point cloud image, and the feature point extraction process;

[0036] Figure 4 is the single - frame point cloud and surface protrusion diagram of the LED plano - convex lens;

[0037] Figure 5 is the top - view point cloud after stitching 9 - frame point clouds of the LED plano - convex lens;

[0038] Figure 6 is the image after stitching 16 - frame point clouds of the plano - convex lens;

[0039] Figure 7 is the side - view point cloud and internal impurities after stitching 16 - frame point clouds of the plano - convex lens. Detailed Description of the Preferred Embodiments

[0040] Example 1: Refer to Figures 1-7 , a method for detecting large - scale optical components based on a robotic arm and an ultra - high - resolution optical coherence tomography imaging system, including the following steps:

[0041] (1) Build a spectral domain optical coherence tomography system with a light source central wavelength of 850 nm and a bandwidth of 165 nm;

[0042] (2) Fix the OCT probe at the end of the six - degree - of - freedom robotic arm through an adapter. The robotic arm has a repeat positioning accuracy of ±0.03 mm and a working radius of 914 mm;

[0043] (3) Calibrate the coordinate system at the end of the robotic arm and the OCT imaging space using a three - dimensional calibration target, and calculate the rotation matrix and translation offset through singular value decomposition;

[0044] (4) Verify the system resolution through microsphere imaging experiments, test the planar imaging ability of the system using mirror scanning, and evaluate the detection accuracy by combining defect sample experiments.

[0045] (5) Plan the scanning path of the robotic arm and move it step by step along the surface of the optical element with a preset overlap rate;

[0046] (6) Acquire local three-dimensional OCT images, automatically stitch multiple groups of images based on the robotic arm pose data, and generate a large field-of-view detection result of the optical element.

[0047] Among them, the OCT system in step (1) adopts spectral-domain OCT technology, and the axial resolution in air is about 1.9 μm.

[0048] Among them, the three-dimensional calibration target in step (3) is a three-peak structure printed by light curing, and the coordinate system transformation matrix is calculated by extracting the peak coordinates under multiple poses of the robotic arm.

[0049] Among them, the motion path of the robotic arm is planned based on the geometric shape and detection requirements of the optical element, and large-range three-dimensional topography reconstruction is achieved by adjusting the overlap rate parameter.

[0050] Among them, the OCT three-dimensional images are used to identify micro defects, including scratches, stains, and polishing protrusions, as well as to accurately measure the size and position of the defects

[0051] First, use a three-dimensional calibration target to calibrate the coordinate system of the robotic arm end and the OCT imaging space.

[0052] Furthermore, the specific method for this step is as follows: We place the sample on a table that remains stationary in the base coordinate system of the robotic arm, and then program the robotic arm to move in small steps (>3) along the x, y, and z axes in the end effector space respectively. Each time it moves, the OCT maintains the same rotation angle to obtain 3D images at each position.

[0053] Then, through a custom MATLAB script, the coordinates of the three peaks in the OCT space can be obtained and averaged. If we assume that the initial value of the average peak coordinates in the OCT space is P0, and the average coordinates of the i-th step calculated are P i ,。

[0054] Furthermore, the coordinates in the initial OCT space can be obtained through the following formula:

[0055] P′ i =P i -P0

[0056] Since the robotic arm moves in the end - effector space to obtain imaging data, the coordinates Pr of the end - effector in the initial end - effector space can be given according to the moving distance. i , if we use R e2o to represent the rotation matrix from the end - effector of the robotic arm to the OCT space, then the following relationship can be obtained:

[0057] [P′1,P′2,P′3,…,P′ n = R e2o *[-Pr1,-pr2,-pr2,…,-Pr n

[0058] Secondly, the rotation matrix R can be obtained by performing singular value decomposition (SVD) based on the pseudo - inverse matrix method. e2o .

[0059] Then, the translational offset between the end - effector space and the OCT space is calculated. Further, the robotic arm is programmed to aim at the above - mentioned target in different directions according to the estimated initial offset t0, and then the initial transformation from the end - effector space of the robotic arm to the OCT space can be expressed as:

[0060]

[0061] At each pose of the robotic arm, the pose of the robotic arm and the OCT image are recorded, and the average peak coordinates in the OCT space can be calculated. Since the model is stationary in the base coordinate system of the robotic arm, the peak coordinates calculated at different poses match each other.

[0062] Further, iteration can be performed on the x, y, and z offsets of t0 to calculate the new coordinates Pb i in the base coordinate system of the robotic arm. When the calculated coordinate difference is the smallest, the measured translational offset can be obtained, expressed as:

[0063]

[0064] where is the average value of all coordinates at different robotic arm poses. In this way, the final transformation from the end - effector of the robotic arm to the OCT space can be obtained to navigate the OCT probe for high - resolution wide - field scanning.

[0065] After calibration, during detection, the robotic arm scans the surface of the optical element step by step along the preset path, collects local 3D OCT images and automatically stitches them, and finally generates a complete large - field - of - view detection result. This method optimizes the imaging depth by dynamically adjusting the probe height, combines the overlapping - area registration algorithm to eliminate stitching errors, and can detect surface scratches, polishing protrusions, and micron - scale contamination particles simultaneously. ​

[0066] Example 2: Taking the detection of a plano-convex lens with a diameter of 20 mm and a planar LED lens with a diameter of 100 mm as an example: First, build a spectral domain optical coherence tomography system, install a six-degree-of-freedom robotic arm (ELITE ROBOT EC66), and fix a customized OCT probe at the end. Use a three-peak calibration target printed by stereolithography for hand-eye calibration. Through the robotic arm moving step by step along the X / Y / Z axes multiple times, collect the OCT images of the target and calculate the coordinate system transformation matrix, as Figure 3 shown. After calibration, fix the sample on the stage, set the overlapping rate of adjacent areas for scanning the planar LED lens to 20%, and set the overlapping rate of adjacent areas for scanning the plano-convex lens to 60%. The schematic diagram of the overlapping rate is as Figure 2 shown. The robotic arm scans the two samples respectively with a single scan area of 3.2 mm × 3.2 mm, performs a 3×3 step-by-step scan on the planar lens, and a 4×4 step-by-step scan on the plano-convex lens. Dynamically adjust the height of the probe from the sample surface by extracting the lens surface in the B-scan, so that the sample is always at a suitable scanning position. As Figure 4 shown is the point cloud result of a single scan of the planar LED lens. Clear grinding marks and fine scratches can be seen, and a polishing protrusion with a height of about 6 μm can be easily observed. After the scan is completed, use the iterative closest point (ICP) algorithm to stitch 9 groups of point cloud images of the planar LED lens into a 13.5 mm × 13.5 mm three-dimensional point cloud, as Figure 5 shown; stitch 16 point cloud images of the plano-convex lens into an 8.2 mm × 8.2 mm three-dimensional point cloud, as Figure 6 shown, and detect internal impurities as small as about 3 μm, as shown in the enlarged part in Figure 7 .

[0067] It can be understood that the present invention is described through specific embodiments. Those skilled in the art know that without departing from the spirit and scope of the present invention, various changes or equivalent substitutions can be made to these features and embodiments. In addition, under the teaching of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application belong to the scope protected by the present invention.

Claims

1. A large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system, characterized in that, It includes the following steps: (1) Set up a spectral domain optical coherence tomography system with a light source having a central wavelength of 850 nm and a bandwidth of 165 nm; (2) Fix the OCT probe to the end of a six-degree-of-freedom robotic arm through an adapter. The repeat positioning accuracy of the robotic arm is ±0.03 mm, and the working radius is 914 mm; (3) Calibrate the coordinate system at the end of the robotic arm and the OCT imaging space using a three-dimensional calibration target, and calculate the rotation matrix and translation offset through singular value decomposition; (4) Verify the system resolution through microsphere imaging experiments, test the planar imaging ability of the system using mirror scanning, and evaluate the detection accuracy in combination with defect sample experiments; (5) Plan the scanning path of the robotic arm and move step by step along the surface of the optical component at a preset overlap rate; (6) Acquire local three-dimensional OCT images, and automatically stitch multiple groups of images based on the robotic arm pose data to generate a large-field-of-view detection result of the optical component.

2. The large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system according to claim 1, characterized in that, The OCT system in step (1) adopts spectral domain OCT technology, and the axial resolution in air is about 1.9 μm.

3. A large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system according to claim 1, characterized in that, The three-dimensional calibration target in step (3) is a three-peak structure printed by light curing, and the coordinate system transformation matrix is calculated by extracting the peak coordinates under multiple poses of the robotic arm.

4. A large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system according to claim 1, characterized in that, The motion path of the robotic arm is planned based on the geometric shape and detection requirements of the optical component, and large-range three-dimensional topography reconstruction is achieved by adjusting the overlap rate parameter.

5. The large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system according to claim 1, wherein Use the OCT three-dimensional image to identify micro defects, including scratches, stains, and polishing protrusions, as well as accurately measure the size and position of the defects.

6. The large optical element detection method based on a robotic arm and an ultra-high resolution optical coherence tomography imaging system according to claim 1, wherein Step (3) uses a three-dimensional calibration target to calibrate the coordinate system at the end of the robotic arm and the OCT imaging space, The specific method is as follows: Place the sample on a table that remains stationary in the base coordinate system of the robotic arm, and then program the robotic arm to move multiple small steps (>3) on the x, y, and z axes in the end effector space respectively. Each time it moves, the OCT maintains the same rotation angle to obtain 3D images at each position, Then, through a custom MATLAB script, the coordinates of the three peaks in the OCT space are obtained and averaged. If the initial value of the average peak coordinate in the OCT space is set as P0, and the average coordinate at the i-th step calculated is P i , Obtain the coordinates in the initial OCT space through the following formula: P i ′ = P i - P0 Since the robotic arm moves in the end - effector space to obtain imaging data, the coordinates Pr of the end - effector in the initial end - effector space are given according to the moving distance. i , if we use R e2o to represent the rotation matrix from the end - effector of the robotic arm to the OCT space, then the following relationship is obtained: [P′1, P′2, P′3, …, P′ n = R e2o *[-Pr1, -Pr2, -Pr2, …, -Pr n ​ Secondly, the rotation matrix R is obtained by performing singular value decomposition (SVD) based on the pseudo-inverse matrix method e2o , Then calculate the translation offset between the end effector space and the OCT space. Program the robotic arm to aim at the above target in different directions according to the estimated initial offset t0, and then the initial transformation from the end effector space of the robotic arm to the OCT space can be expressed as: At each pose of the robotic arm, the pose of the robotic arm and the OCT image are recorded, and the average peak coordinates in the OCT space are calculated. Since the model is stationary in the base coordinate system of the robotic arm, the peak coordinates calculated at different poses match each other. Iterate over the x, y, and z offsets at t0 to calculate the new coordinates Pb in the base coordinate system of the robotic arm i , when the calculated coordinate difference is minimized, the measured translational offset is obtained, expressed as: Among them is the average value of all coordinates under different robotic arm postures. In this way, the final transformation from the end effector of the robotic arm to the OCT space is obtained to navigate the OCT probe for high-resolution wide-field scanning. After calibration, during detection, the robotic arm scans the surface of the optical element step by step along the preset path, collects local 3D OCT images and automatically stitches them together, and finally generates a complete large-field-of-view detection result. This method optimizes the imaging depth by dynamically adjusting the probe height and combines the overlapping area registration algorithm to eliminate stitching errors, and can detect surface scratches, polishing protrusions and micron-sized contamination particles at the same time.

Images