Multi-parameter integrated detection method for telecentric lens

Abstract

The invention discloses a telecentric lens multi-parameter integrated detection method, which comprises the following steps of: after an initial position is manually confirmed, triggering automatic detection through a program, after automatic focusing, moving according to a set path and identifying an upper edge line, then moving rightwards along the line until an angular point 1 appears, then extracting contour features by using an OTSU method and identifying the position of the angular point; recording current coordinates after accurately moving to a corner point 1, starting to move downwards and identifying a right edge line, recording the current coordinates after moving to a corner point 2 along the line and moving to the corner point 2 in the same way, calculating theoretical coordinates of residual corner points according to the obtained accurate coordinates of the corner points and the actual size of the target image, and directly moving to the theoretical positions of the residual corner points; and identifying the position of the angular point and accurately moving the angular point, calculating the magnification, distortion and field curvature value of the lens according to the obtained coordinate data of the target point, and finally displaying the result in a computer equipment interface. According to the invention, the detection efficiency, the focusing precision and the detection stability are improved.

Description

Technical Field

[0001] This invention pertains to lens inspection technology, and particularly relates to a multi-parameter integrated inspection method for telecentric lenses. Background Technology

[0002] Magnification, distortion, and field curvature are three crucial parameters of a lens, significantly impacting its image quality. Compared to ordinary lenses, telecentric lenses exhibit lower distortion, greater depth of field, and constant magnification. Within the depth of field of a telecentric lens, the image remains unchanged regardless of the distance between the object and the lens; therefore, the magnification of telecentric lenses can be measured using target calibration. Currently, target calibration is widely used in lens parameter testing. This method establishes a mapping relationship between image point coordinates and their corresponding spatial coordinates by imaging a precise artificial target with a known geometric structure, thereby accurately estimating the internal and external parameters of the camera or imaging system. The magnification, distortion, and field curvature of a telecentric lens can be calibrated using a two-dimensional planar target. Currently, the testing process for magnification, distortion, and field curvature of telecentric lenses in industrial production relies on manual measurement, which is cumbersome, time-consuming, inefficient, and consumes significant human resources. Furthermore, the testing accuracy is easily affected by subjective human judgment, making it impossible to measure these three parameters in a single test. While methods for detecting magnification, distortion, and field curvature using image processing techniques exist in scientific research, integrated automatic detection of these three parameters has not yet been achieved. Therefore, realizing integrated automatic detection of multiple parameters in telecentric lenses remains a challenging problem.

[0003] In modern optical imaging systems, lens performance directly impacts image quality. Lens magnification, distortion, and field curvature are key parameters affecting image quality. Distortion leads to image distortion, affecting image accuracy; field curvature causes differences in image sharpness and focus across different imaging areas, thus affecting the overall usability of the image; magnification affects aberration balance, depth of field, and other parameters, thus impacting precision visual measurement results. Existing methods utilize frequency domain analysis of fringes or gratings. These methods involve capturing images of black-and-white fringe gratings, performing frequency domain transformation and noise reduction on the obtained data, and then calculating lens magnification and distortion. However, this method is complex, susceptible to environmental interference, and requires sophisticated system equipment, making it unsuitable for industrial inspection environments. Furthermore, capturing projected images of targets for image processing to obtain point information is limited by the target size and camera field of view. Telecentric lenses, with their lower distortion, require more precise identification of the focal plane and target point positions.

[0004] Most existing methods for detecting lens magnification, distortion, and field curvature utilize target calibration, measuring these parameters through a series of image processing steps. However, they cannot output the detection results for all three parameters in a single test. Reference 1 proposes a method for detecting lens distortion and field curvature. This method first adjusts the lens magnification to its theoretical value, projects an image using a DMD chip, controls the movement of the detection camera via a three-axis motion device, marks the coordinates of each point, and directly calculates the field curvature and distortion values ​​of the lens at that point. Reference 2 proposes a method for detecting lens magnification and distortion. This method first integrates the lens under test with a line scan camera and aligns it with a black and white striped grating plate for image acquisition. Then, it performs frequency domain transformation and noise reduction on the image data to extract the actual frequency, calculates the actual grating width, and determines the lens magnification by comparing it with a preset width. Finally, it fits the optimal focal plane based on multiple sampling points and corrects for geometric errors caused by tilt, thereby obtaining the lens distortion. Reference 3 proposes a lens distortion detection method and device. This method first controls a camera to capture an image of a target, obtaining a calibration image. Then, a set of corner points to be fitted are selected from the calibration image for fitting processing to obtain a distortion evaluation value. Finally, the distortion of the lens is detected based on the distortion evaluation value to distinguish different lenses. Reference 4 proposes a lens distortion testing method, device, equipment, and medium based on an industrial camera. Upon receiving a distortion test request, this method activates a projection module to form a projected image through the lens. The projected image is then captured by an industrial camera and binarized. Based on the intrinsic parameter matrix and distortion matrix associated with the industrial camera, the binarized image is subjected to anti-distortion processing to obtain corner coordinates, thereby determining the lateral and longitudinal distortion of the lens.

[0005] 1. Currently, there are methods for detecting lens magnification, distortion, and field curvature, but these methods do not integrate the detection functions of these three parameters. They require repeated testing using different equipment or methods to achieve this.

[0006] 2. Currently, there are methods that use DMD to automatically detect lens distortion and field curvature. However, due to the gaps between adjacent micromirror units, the projected image will generate some stray light, affecting the determination of the optimal focal plane of the image. In addition, the projected image is composed of square pixel grids, and the edges of the pixel grids are stepped, which will affect the accurate identification of corner positions.

[0007] 3. Currently, there are methods that use frequency domain analysis of fringes or gratings to detect magnification, distortion, and field curvature. These methods require frequency domain transformation of the acquired image data, resulting in high computational load and low efficiency. In addition, noise has a significant impact on the spectrum. Furthermore, eliminating the tilt angle by adjusting the relative position of the lens and the grating plate requires high precision and takes a long time to adjust.

[0008] 4. Currently, there are methods that use target calibration to detect lens distortion. However, most of these methods only perform distortion detection based on a single frame image, which requires high image quality. Vibration can reduce the quality of a single frame image and affect the distortion detection results. In addition, image data is obtained by image binarization processing with a fixed threshold. Since the brightness of the projected image is not constant, it may cause the point recognition in subsequent images to fail.

[0009] References:

[0010] [1] Suzhou Aixian Optoelectronic Technology Co., Ltd. A method and device for detecting lens distortion and field curvature: 202411536270.4 [P]. 2025-02-11.

[0011] [2] Shenzhen Canrui Technology Co., Ltd. A method for detecting lens magnification and distortion: 202311029037.2[P]. 2023-09-12.

[0012] [3] Zhuhai Yiwei Semiconductor Co., Ltd. Lens distortion detection method and distortion detection device: 202311578093.1 [P]. 2025-05-27.

[0013] [4] Goertek Inc. Lens distortion testing method, device, equipment and medium based on industrial camera: 202011059960.7 [P]. 2020-11-10. Summary of the Invention

[0014] To address the shortcomings of the above methods in the integrated automatic detection of telecentric lens magnification, distortion, and field curvature, this invention proposes a multi-parameter integrated detection method for telecentric lenses.

[0015] This invention discloses a multi-parameter integrated detection method for telecentric lenses. After manual confirmation of the initial position, automatic detection is triggered by a program. Following autofocus, the lens moves along a set path and identifies the upper edge line. It then moves right along the line until corner point 1 appears. Subsequently, the OTSU method is used to extract contour features and identify the corner point position. After accurately moving to corner point 1, the current coordinates are recorded. The lens then moves downwards and identifies the right edge line, moving along the line until corner point 2 appears. Similarly, after moving to corner point 2, the current coordinates are recorded. Based on the obtained accurate corner point coordinates and the actual size of the target image, the theoretical coordinates of the remaining corner points are calculated. The lens then moves directly to the theoretical positions of the remaining corner points, identifying and accurately moving the corner points. Based on the obtained target point coordinate data, the lens magnification, distortion, and field curvature are calculated. Finally, the results are displayed on a computer device interface. Specifically, the method includes the following:

[0016] Step 1: The detection system acquires images.

[0017] The inspection system includes a light source, an ACSPL three-axis displacement stage, a USAF1951 resolution plate, a right-angle mirror, an inspection camera, a flat-field lens, and computer equipment.

[0018] The lens under test is connected and fixed to the light source and resolution board. The detection camera and right-angle reflector are mounted on the worktable of the ACSPL three-axis displacement stage. The detection motion control is achieved by the computer equipment sending trajectory planning instructions to the motion controller of the ACSPL three-axis displacement stage. Image acquisition is triggered and controlled by the computer equipment according to the detection process logic, after the initial position is confirmed by the operator, by sending acquisition instructions to the detection camera.

[0019] Since the camera's field of view can only fully contain the stripe patterns of groups 6, 7, 8, and 9 in the resolution test image, stripe pairs are selected from these groups for image processing and recognition.

[0020] Step 2: Target and movement path.

[0021] Target image:

[0022] Before lens inspection, a rectangular frame in the target image is selected as the image size reference for this inspection. At the position of the resolution test map, focus is performed based on the MTF contrast and value of a pair of horizontal and vertical stripes, and the best focal plane at that position is taken as the focal plane closest to the corner point.

[0023] The resolution test image in the upper right corner of the target image is selected as the initial position. The upper right corner is the first corner point to be identified, and the lower right corner is the second corner point to be identified. The right rectangular frame is used as the reference movement path between the two points.

[0024] Edge line position recognition and movement:

[0025] By binarizing with a fixed threshold and drawing edge lines on the contour, all edge line features of the image are obtained. Hough transform is then used for line detection to identify the position of the outer edge line.

[0026] Given the positional relationship between the resolution test image and the nearest corner point, as well as the positional relationship between adjacent corner points, determine the movement direction according to the corresponding positional relationship, and perform multiple movements according to an appropriate step size. After each movement, identify the edge line position and move the displacement stage to keep the target edge line located at the center of the image.

[0027] Step 3: Corner feature extraction, recognition, and movement.

[0028] The OTSU thresholding method is used to extract corner features. The minimum bounding rectangle method is applied directly to the extracted corner features. The vertex corresponding to the bounding rectangle is selected as the corner position based on the positional relationship between the target corner and the rectangle. The extraction of corner contour features at other positions is carried out in the same way.

[0029] Accurate movement of corner points:

[0030] Based on the position information returned by corner point recognition, the three-axis translation stage moves accordingly. Due to errors in the translation stage's movement accuracy and the ratio error between the image pixel distance and the translation stage's movement distance, the translation stage cannot reach the accurate position of the image corner point in a single movement. By adopting a visual servoing method, based on the distance difference between the corner point and the center of the image returned in real time, the positioning error is gradually reduced through multiple movement iterations, eventually approaching the target position, achieving accurate movement of the corner point, and improving the repeatability and reliability of the translation stage's movement.

[0031] Step 4: Calculation of lens magnification, distortion, and field curvature.

[0032] Based on the above detection process, the xyz coordinates of the four corner points and the z coordinate of the center are obtained, where the measured coordinates of each corner point are... The coordinates of the focal plane position measured at the center are: .

[0033] (1) Calculation of multiplier:

[0034] The lens magnification M is calculated using the ratio of the measured length of the 23 sides to the actual length. The formula is as follows:

[0035]

[0036] In the formula, The ideal size is 23 sides.

[0037] (2) Distortion calculation:

[0038] Lens distortion is calculated using the differences between the side lengths of sides 34, 41, and 12 and their ideal side lengths, as well as the distance differences between the measured coordinates of the four angles and one angle and their ideal coordinates. The calculation formula is as follows:

[0039]

[0040] In the formula, The distortion value is for 34 edges; Ideal dimensions for 34 sides; The distortion value is for 41 edges; Ideal dimensions for 41 sides; The distortion value is for 12 sides; The ideal size is for 12 sides.

[0041]

[0042] In the formula, 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 1 jiao Distortion value in direction; The ideal size is 1 cent; 1 jiao Distortion value in direction; The ideal is 1 cent coordinate.

[0043]

[0044] (3) Field calculation:

[0045] Utilize the four corners Coordinates and Center Coordinate calculation of shot field curvature The calculation formula is as follows:

[0046]

[0047] In the formula, The field curvature value is 1. The field curvature value is 2 degrees. The field curvature value is for a 3-angle angle; The field curvature value is for the four corners.

[0048]

[0049] After the lens magnification, distortion, and field curvature are calculated, the results will be displayed on the computer interface for easy viewing.

[0050] Furthermore, in actual working conditions, various factors can cause the entire detection device to vibrate non-periodicly, resulting in jitter in the projected image. With the three axes of the displacement stage as the coordinate system, vibration exists in the XYZ directions of the coordinate system.

[0051] Vibration in the Z direction can affect the sharpness of the acquired image, causing a large deviation in the search for the optimal focal plane position. In order to reduce the impact of vibration on image sharpness, a time-series image sequence of the current position is acquired, and the MTF contrast of a pair of horizontal and vertical stripes is used as an evaluation value of image sharpness to select the image with the best sharpness.

[0052] Vibration in the XY direction can affect the identification of corner positions and increase the error of the obtained XY coordinates of corners. By using the time-series image sequence acquired by the camera, obvious outliers are filtered out based on the corner features of the images. Then, the arithmetic mean of the corner feature coordinates of the remaining images is taken as the final result.

[0053] Furthermore, the minimum bounding rectangle method fits a rectangle with the smallest area to the target contour on the two-dimensional plane, and the side lengths may not be parallel to the image coordinate axes; its mathematical description is: for a given set of discrete points... Representing the outline pixel coordinates, find a rotation angle. and the center of the rectangle This makes the rectangle rotate It can then contain all points, and its area Reach the global minimum.

[0054] Furthermore, the specific movement of the corner point is as follows:

[0055] First, determine the size of the target location range for the corner point, and then calculate the center point based on the returned corner point location information. Current position of corner point radial distance At this point, the current position of the corner point is defined. The condition that the target location is already in place is:

[0056]

[0057] In the formula, r is the radius of the target area.

[0058] The beneficial technical effects of this invention are as follows:

[0059] This invention projects a target image onto a USAF1951 resolution plate, utilizes a combination of a displacement stage and a camera to automatically acquire a sequence of time-series images of the target, and uses MTF contrast to optimize and reduce the impact of vibration. Based on the selected horizontal and vertical stripes, MTF contrast is used for focusing, improving focusing accuracy and stability. Adaptive thresholding is used to extract corner features and accurately identify corner positions. Then, closed-loop iterative control is used to precisely move to the corner, automatically acquiring the three-dimensional coordinates of the target point. Finally, the magnification, distortion, and field curvature of the telecentric lens are calculated based on the coordinates. The entire detection process integrates the detection functions of three parameters, improving detection efficiency and accuracy while reducing labor costs. Attached Figure Description

[0060] Figure 1 This is a flowchart of the multi-parameter integrated detection method for telecentric lenses of the present invention.

[0061] Figure 2 This is a schematic diagram of the detection system structure.

[0062] In the figure: 1. ACSPL three-axis displacement stage; 2. Right-angle mirror; 3. Lens under test; 4. Resolution board and light source connection device; 5. Z-axis drive; 6. X-axis drive; 7. Computer equipment; 8. Flat field lens; 9. Inspection camera; 10. Y-axis drive.

[0063] Figure 3 This is a projection image from the USAF resolution test.

[0064] Figure 4 The process of filtering image data based on clarity.

[0065] Figure 5 The results are filtered and averaged for the feature coordinates of corner points.

[0066] Figure 6 Image of a target from a USAF 1951 resolution plate.

[0067] Figure 7 Visualization results of the horizontal edge line contour features.

[0068] Figure 8 Visualization results of the vertical edge line contour features.

[0069] Figure 9 This is the result of corner feature extraction.

[0070] Figure 10 The results show the corner location identification.

[0071] Figure 11 This is the criterion for determining the accurate movement of the corner point.

[0072] Figure 12 This represents the theoretical positional relationship between the corner point and the center.

[0073] Figure 13 This is a screenshot of the interface displaying the test results. Detailed Implementation

[0074] The present invention will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0075] The flowchart of the multi-parameter integrated detection method for telecentric lenses of the present invention is as follows: Figure 1As shown. Due to the limitation of the field of view, the camera image cannot display the complete projected image. Since the projected image contains a lot of information, automatically traversing and identifying the first position is difficult and time-consuming. Therefore, a manual method for determining the initial position is used. In this invention, after the initial position is manually confirmed, automatic detection is triggered by the program. After automatic focusing, it moves along a set path and identifies the upper edge line. Then, it moves to the right along the line until corner point 1 appears. Subsequently, the OTSU method is used to extract contour features and identify the corner point position. After accurately moving to corner point 1, the current coordinates are recorded. It then begins to move downwards and identify the right edge line, moving along the line until corner point 2 appears. Similarly, after moving to corner point 2, the current coordinates are recorded. Based on the obtained accurate corner point coordinates and the actual size of the target image, the theoretical coordinates of the remaining corner points are calculated. It directly moves to the theoretical position of the remaining corner points, identifies the corner point position, and moves accurately. Based on the obtained target point coordinate data, the lens magnification, distortion, and field curvature are calculated. Finally, the results are displayed on the computer device interface. Specifically, it includes the following:

[0076] Step 1: The detection system acquires images.

[0077] Detection system such as Figure 2 As shown, it includes a light source, an ACSPL three-axis displacement stage 1, a USAF1951 resolution board, a right-angle mirror 2, a detection camera 9, a flat-field lens 8, and a computer device 7.

[0078] The lens under test 3 is connected and fixed to the light source and resolution board. The detection camera 9 and the right-angle reflector 2 are installed on the worktable of the ACSPL three-axis displacement stage 1. The detection motion control is achieved by the computer device 7 sending trajectory planning instructions to the motion controller of the ACSPL three-axis displacement stage 1. The image acquisition is triggered and controlled by the computer device 7 according to the detection process logic, after the initial position is confirmed by the human, by sending acquisition instructions to the detection camera 9.

[0079] USAF resolution test projection diagram as shown below Figure 3 As shown, the camera's field of view can only completely include the stripe patterns in groups 6, 7, 8, and 9 of the resolution test image. Therefore, stripe pairs are selected from the above groups for image processing and recognition.

[0080] In actual working conditions, various factors can cause the entire detection device to vibrate non-periodicly, resulting in jitter in the projected image. With the three axes of the displacement stage as the coordinate system, vibration exists in the XYZ directions of the coordinate system.

[0081] Z-axis vibration can affect the sharpness of acquired images, causing significant deviations in the search for the optimal focal plane position. To reduce the impact of vibration on image sharpness, a time-series image sequence of the current position is acquired. The MTF contrast of a pair of horizontal and vertical stripes is used as an evaluation value for image sharpness, and the image with the best sharpness is selected. The selection process is as follows: Figure 4 As shown.

[0082] Vibration in the XY directions can affect corner location identification, increasing the error in the acquired XY coordinates of the corners. Using a time-series image sequence acquired by the camera, obvious outliers are filtered out based on the corner feature characteristics of the images. Then, the arithmetic mean of the corner feature coordinates of the remaining images is taken as the final result. Figure 5 As shown.

[0083] Step 2: Target and movement path.

[0084] (1) Target image:

[0085] Pre-existing target images from a USAF 1951 resolution plate, such as... Figure 6 As shown, the image only shows the portion near the center, and the completely black rectangular blocks in the image represent USAF resolution test images.

[0086] Compared to the gray-level gradient method, the MTF contrast method is more robust to noise and texture, and its curve exhibits a more pronounced unimodal characteristic. Before lens inspection, a rectangular frame in the target image is selected as the image size benchmark for this inspection. At the resolution test map location, focusing is performed based on the MTF contrast and values ​​of a pair of horizontal and vertical stripes, and the optimal focal plane at that location is taken as the focal plane closest to the corner point.

[0087] Because manual horizontal correction of the target projection image in the camera frame has a certain deviation, and the lens under test also has distortion, the image in the camera frame has a certain angle with the horizontal line. The actual difference between the coordinates of adjacent corner points deviates from the theoretical difference. If the displacement stage moves directly to the theoretical coordinates of the corner point, the acquired image may not contain any corner points. By identifying the edge line position, the displacement stage moves along the edge line to the next corner point. Then, based on the coordinates of the two corner points, the coordinates of the remaining corner points and the center are calculated, thus achieving accurate movement of all points. The upper and lower edges of the rectangular frame intersect with the circular line. To avoid the influence of the circular line on edge line identification, the left or right edge line is chosen as the reference for subsequent movement along the edge line. This invention selects the resolution test image in the upper right corner of the target image as the initial position, the upper right corner as the first corner point to be identified, the lower right corner as the second corner point to be identified, and the right rectangular frame line as the reference movement path between the two points.

[0088] (2) Edge line position recognition and movement:

[0089] Edge lines primarily serve as a reference path for movement; only their approximate location needs to be identified. By binarizing with a fixed threshold and drawing edge lines on the contours, all edge line features of the image can be obtained. The visualization results of the edge line contour features are shown below. Figure 7 and Figure 8As shown, since the acquired image contains both the target edge line and a portion of the resolution test image, the edge features of the resolution test image can interfere with the identification of the target edge line. Based on the fixed relative relationship between the two, the target edge line features can be filtered out. Hough transform is used for line detection to identify the position of the outer edge line.

[0090] Given the positional relationship between the resolution test image and the nearest corner point, as well as the positional relationship between adjacent corner points, determine the movement direction according to the corresponding positional relationship, and perform multiple movements according to an appropriate step size. After each movement, identify the edge line position and move the displacement stage to keep the target edge line located at the center of the image.

[0091] Step 3: Corner feature extraction, recognition, and movement.

[0092] (1) Feature extraction using the OTSU thresholding method:

[0093] The Otsu thresholding method, also known as the Otsu method or the maximum inter-class variance method, is an unsupervised, adaptive global thresholding algorithm for image segmentation based on the statistical characteristics of gray-level histograms. It is mainly used for segmenting the foreground and background of images.

[0094] Due to factors such as ambient light and optical path loss, the overall brightness of the projected images from different lenses fluctuates to some extent. Using a fixed threshold to extract contour features cannot guarantee the completeness of feature extraction, thus affecting the accurate identification of subsequent corner points. The OTSU thresholding method, however, can adaptively adjust the image threshold to achieve complete extraction of corner contour features, providing a solid image data foundation for subsequent corner point identification. The corner feature extraction results are shown below. Figure 9 As shown. The extraction of contour features for other corner points is similar.

[0095] (2) Corner point location identification:

[0096] Because the extracted feature image contains some stripes, identifying contour edges using line detection requires filtering out interfering features. Furthermore, the presence of non-zero curvature at feature edges can lead to multiple results for line detection, making it difficult to select the most suitable line. Since the outer edges of the target feature are relatively smooth, the minimum bounding rectangle method can be used to directly obtain the corner positions. The minimum bounding rectangle method fits a rectangle with the smallest area to the target contour on a two-dimensional plane, and the side lengths do not necessarily have to be parallel to the image coordinate axes.

[0097] Its mathematical description is: for a given set of discrete points Representing the outline pixel coordinates, find a rotation angle. and the center of the rectangle This makes the rectangle rotate It can then contain all points, and its area Reach the global minimum.

[0098] This method does not require filtering out interfering stripe features; it can be directly applied to the extracted corner features. The obtained minimum bounding rectangle is used to select the vertices corresponding to the corner points based on their positional relationship with the target corners. The corner point location identification results are as follows: Figure 10 As shown. The extraction of contour features for other corner points is similar.

[0099] (3) Accurate movement of corner points:

[0100] Based on the position information returned by corner point recognition, the three-axis translation stage moves accordingly. Ideally, the corner point should coincide with the center of the image, indicating accurate movement to the corner position. However, due to errors in the translation stage's movement accuracy and the ratio of image pixel distance to the translation stage's movement distance, the translation stage cannot reach the accurate corner position in a single movement. A visual servoing method is employed, using the real-time distance difference between the corner point and the image center, to iteratively reduce the positioning error through multiple movements, ultimately approaching the target position and achieving accurate corner point movement. This improves the repeatability and reliability of the translation stage's movement.

[0101] First, determine the size of the target location range for the corner point, and then calculate the center point based on the returned corner point location information. Current position of corner point radial distance At this point, the current position of the corner point is defined. The condition that the target location is already in place is:

[0102]

[0103] In the formula, r is the radius of the target area.

[0104] The criteria for determining accurate corner movement are as follows: Figure 11 As shown, Figure 11 The dashed circle in the diagram represents the target location range. This represents the deflection angle produced before and after a single movement. When point... When the target position is within the dashed circle, it indicates that the movement has been successfully completed and the target position at the corner has been reached.

[0105] Step 4: Calculation of lens magnification, distortion, and field curvature.

[0106] Based on the above detection process, the xyz coordinates of the four corner points and the z coordinate of the center are obtained. The theoretical positional relationship between the corner points and the center is as follows: Figure 12 As shown. The measured coordinates of each corner point are... The coordinates of the focal plane position measured at the center are: .

[0107] (1) Calculation of multiplier:

[0108] Compared to ordinary lenses, telecentric lenses offer advantages such as no perspective distortion, a large depth of field, and constant magnification within the depth of field. This invention focuses at both the four corners and the center. With ordinary lenses, the perspective model changes after each focusing movement, making the geometric measurement reference unreliable. However, due to its characteristics, the telecentric lens, by focusing on the target position individually, improves the edge blurring problem caused by lens curvature and tilting during installation, thus enhancing the accuracy and stability of corner position identification. This invention calculates the lens magnification M using the ratio of the measured length of the 23 sides to the actual length, as shown in the following formula:

[0109]

[0110] In the formula, The ideal size is 23 sides.

[0111] (2) Distortion calculation:

[0112] Lens distortion is calculated using the differences between the side lengths of sides 34, 41, and 12 and their ideal side lengths, as well as the distance differences between the measured coordinates of the four angles and one angle and their ideal coordinates. The calculation formula is as follows:

[0113]

[0114] In the formula, The distortion value is for 34 edges; Ideal dimensions for 34 sides; The distortion value is for 41 edges; Ideal dimensions for 41 sides; The distortion value is for 12 sides; The ideal size is for 12 sides.

[0115]

[0116] In the formula, 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 1 jiao Distortion value in direction; The ideal size is 1 cent; 1 jiao Distortion value in direction; The ideal is 1 cent coordinate.

[0117]

[0118] (3) Field calculation:

[0119] Utilize the four corners Coordinates and Center Coordinate calculation of shot field curvature The calculation formula is as follows:

[0120]

[0121] In the formula, The field curvature value is 1. The field curvature value is 2 degrees. The field curvature value is for a 3-angle angle; The field curvature value is for the four corners.

[0122]

[0123] After the lens magnification, distortion, and field curvature calculations are completed, the results will be displayed on the computer interface for easy viewing. The results displayed on the interface are as follows: Figure 13 As shown.

[0124] The integrated detection system used in this invention projects images through a USAF1951 resolution plate and uses computer equipment to control the movement of a three-axis displacement stage and the acquisition of camera images. In contrast, existing technologies use DMD to project images and do not mention the application of computer equipment in the detection process.

[0125] This invention can obtain the detection results of telecentric lens magnification, distortion and field curvature through a single complete test, while the existing technology does not integrate the detection function of telecentric lens magnification, distortion and field curvature.

[0126] Compared with existing methods that use frequency domain analysis of stripes or gratings, the target calibration method of this invention can directly measure and calculate magnification, distortion, and field curvature. Furthermore, due to the limitation of the camera's field of view, the detection method of this invention requires manual determination of the initial position before detection, while the target image of the prior art can be fully displayed in the camera's field of view, eliminating the need to determine the initial position before detection.

[0127] This invention acquires time-series image sequences at target locations and uses MTF contrast optimization and arithmetic mean of non-outlier coordinates to reduce the impact of vibration on the Z and XY coordinates of the target points, respectively. In contrast, most existing technologies use single-frame images, which are easily affected by vibration.

[0128] This invention determines the movement path based on edge line recognition and ensures that the displacement stage moves accurately to the corner point through image-based visual servo control. In contrast, existing technologies use preset fixed points, which can be fixed by image recognition or by directly moving the displacement stage to the fixed point.

[0129] Compared with the prior art, the beneficial effects of the present invention are:

[0130] (1) The present invention projects images using the USAF1951 resolution plate target image. Using a three-axis displacement stage and camera, only the initial position needs to be determined manually. The subsequent detection process is automatically implemented, and the detection results are finally displayed on a computer device. This reduces the technical threshold and the amount of participation required for manual operation and avoids the subjective errors introduced by manual detection.

[0131] (2) This invention utilizes the characteristic that the magnification of a telecentric lens is basically constant within the focal depth range, and integrates the detection of three major parameters of telecentric lens magnification, distortion and field curvature, so as to realize the output of the detection results of the three parameters in a single automatic detection, which changes the current technology that can only output the results of some parameters in a single detection and improves the detection efficiency.

[0132] (3) The present invention introduces a method of MTF contrast selection, corner coordinate screening and arithmetic mean of time-series image sequences, which avoids the instability of single frame image quality and reduces the impact of vibration on image data stability.

[0133] (4) The present invention uses the MTF contrast of two orthogonal directional stripes selected by the USAF1951 resolution board as the focus evaluation function, which has stronger noise resistance compared with the traditional gray-scale gradient change method; by extracting corner features through adaptive threshold and introducing an image-based visual servo control method, the corner feature positioning error of the image is reduced to the pixel level, which improves the measurement accuracy and repeatability.

Claims

1. A multi-parameter integrated detection method for telecentric lenses, characterized in that, After the initial position is manually confirmed, automatic detection is triggered by the program. After autofocus, it moves along a set path and identifies the upper edge line. Then, it moves to the right along the line until corner point 1 appears. Subsequently, the OTSU method is used to extract contour features and identify the corner point position. After accurately moving to corner point 1, the current coordinates are recorded. It then begins to move downwards and identify the right edge line, moving along the line until corner point 2 appears. Similarly, after moving to corner point 2, the current coordinates are recorded. Based on the obtained accurate corner point coordinates and the actual size of the target image, the theoretical coordinates of the remaining corner points are calculated. It moves directly to the theoretical position of the remaining corner points, identifies the corner point position, and moves accurately. Based on the obtained target point coordinate data, the lens magnification, distortion, and field curvature are calculated. Finally, the results are displayed on the computer device interface. Specifically, this includes the following: Step 1: The detection system acquires images; The detection system includes a light source, an ACSPL three-axis displacement stage (1), a USAF1951 resolution plate, a right-angle mirror (2), a detection camera (9), a flat-field lens (8), and computer equipment (7). The lens under test (3) is connected and fixed to the light source and resolution plate. The detection camera (9) and the right-angle reflector (2) are installed on the worktable of the ACSPL three-axis displacement stage (1). The detection motion control is achieved by the computer device (7) sending trajectory planning instructions to the motion controller of the ACSPL three-axis displacement stage (1). The image acquisition is triggered and controlled by the computer device (7) according to the detection process logic, after the initial position is confirmed by the human, by sending acquisition instructions to the detection camera (9). Since the camera's field of view can only completely contain the stripe patterns of groups 6, 7, 8, and 9 in the resolution test image, stripe pairs are selected from the above groups for image processing and recognition. Step 2: Target and movement path; Target image: Before lens inspection, a rectangular frame in the target image is selected as the image size reference for this inspection. At the position of the resolution test map, focus is performed based on the MTF contrast and value of a pair of horizontal and vertical stripes, and the best focal plane at that position is taken as the focal plane closest to the corner point. The resolution test image in the upper right corner of the target image is selected as the initial position. The upper right corner is the first corner point to be identified, and the lower right corner is the second corner point to be identified. The right rectangular frame is used as the reference movement path between the two points. Edge line position recognition and movement: By binarizing with a fixed threshold and drawing edge lines on the contour, all edge line features of the image are obtained. Hough transform is used for line detection to identify the position of the outer edge line. Given the positional relationship between the resolution test image and the nearest corner point, as well as the positional relationship between adjacent corner points, determine the movement direction according to the corresponding positional relationship, perform multiple movements according to an appropriate step size, identify the edge line position after each movement and move the displacement stage to keep the target edge line located at the center of the image; Step 3: Corner feature extraction, recognition, and movement; The OTSU thresholding method is used to extract corner features. The minimum bounding rectangle method is applied directly to the extracted corner features. The vertex corresponding to the bounding rectangle is selected as the corner position based on the positional relationship between the target corner and the rectangle. The extraction of corner contour features at other positions is carried out in the same way. Accurate movement of corner points: Based on the position information returned by corner point recognition, the three-axis translation stage moves accordingly. Due to errors in the translation stage's movement accuracy and the ratio error between the image pixel distance and the translation stage's movement distance, the translation stage cannot reach the accurate position of the image corner point in a single movement. By adopting a visual servoing method, based on the distance difference between the corner point and the center of the image returned in real time, the positioning error is gradually reduced through multiple movement iterations, eventually approaching the target position, achieving accurate movement of the corner point, and improving the repeatability and reliability of the translation stage's movement. Step 4: Calculation of lens magnification, distortion, and field curvature; Based on the above detection process, the xyz coordinates of the four corner points and the z coordinate of the center are obtained, where the measured coordinates of each corner point are... The coordinates of the focal plane position measured at the center are: ; (1) Calculation of multiplier: The lens magnification M is calculated using the ratio of the measured length of the 23 sides to the actual length. The formula is as follows: ； In the formula, Ideal dimensions for 23 sides; (2) Distortion calculation: Lens distortion is calculated using the differences between the side lengths of sides 34, 41, and 12 and their ideal side lengths, as well as the distance differences between the measured coordinates of the four angles and one angle and their ideal coordinates. The calculation formula is as follows: ； In the formula, The distortion value is for 34 edges; Ideal dimensions for 34 sides; The distortion value is for 41 edges; Ideal dimensions for 41 sides; The distortion value is for 12 sides; Ideal dimensions for 12 sides; ； In the formula, 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 4 corners Distortion value in direction; For the ideal of 4 angles coordinate; 1 jiao Distortion value in direction; The ideal size is 1 cent; 1 jiao Distortion value in direction; The ideal is 1 cent coordinate; ； (3) Field calculation: Utilize the four corners Coordinates and Center Coordinate calculation of shot field curvature The calculation formula is as follows: ； In the formula, The field curvature value is 1. The field curvature value is 2 degrees. The field curvature value is for a 3-angle angle; The field curvature values ​​at the four corners; ； After the lens magnification, distortion, and field curvature are calculated, the results will be displayed on the computer interface for easy viewing.

2. The multi-parameter integrated detection method for telecentric lenses according to claim 1, characterized in that, Step 1 also includes image screening; in actual working conditions, various factors may cause the entire detection device to vibrate non-periodicly, which will cause the projected image to shake. With the three axes of the displacement stage as the coordinate system, vibration exists in the XYZ directions of the coordinate system. Vibration in the Z direction can affect the sharpness of the acquired image, causing a large deviation in the search for the optimal focal plane position. In order to reduce the impact of vibration on image sharpness, a time-series image sequence of the current position is acquired, and the MTF contrast of a pair of horizontal and vertical stripes is used as an evaluation value of image sharpness to select the image with the best sharpness. Vibration in the XY direction can affect the identification of corner positions and increase the error of the obtained XY coordinates of corners. By using the time-series image sequence acquired by the camera, obvious outliers are filtered out based on the corner features of the images. Then, the arithmetic mean of the corner feature coordinates of the remaining images is taken as the final result.

3. The multi-parameter integrated detection method for telecentric lenses according to claim 1, characterized in that, The minimum bounding rectangle method fits a rectangle with the smallest area to the target contour on a two-dimensional plane, and the side lengths may not be parallel to the image coordinate axes; its mathematical description is: for a given set of discrete points... Representing the outline pixel coordinates, find a rotation angle. and the center of the rectangle This makes the rectangle rotate It can then contain all points, and its area Reach the global minimum.

4. The multi-parameter integrated detection method for a telecentric lens according to claim 1, characterized in that, The specific movement of the corner point position is as follows: First, determine the size of the target location range for the corner point, and then calculate the center point based on the returned corner point location information. Current position of corner point radial distance At this point, the current position of the corner point is defined. The condition that the target location is already in place is: ； In the formula, r is the radius of the target area.

Images