A sample height acquisition method and system based on line-spectral confocal profilometry

Abstract

The application belongs to the technical field of optical precision measurement, and discloses a sample height acquisition method and system based on line spectrum confocal profilometry. The method comprises the following steps: S1, collecting multiple gray images of the surface profile of a sample to be measured, calculating the similarity between the gray images of adjacent sampling times and dividing the gray images into multiple groups; S2, performing Gaussian blur mean value processing and Gaussian cut-off on each group of images to obtain Gaussian-processed images, and then performing preprocessing on the Gaussian-processed images to obtain the position coordinates of the pixel points corresponding to the maximum gray value; S3, acquiring the sample height corresponding to the maximum gray value, and obtaining the sample height profile corresponding to the group of images from the sample heights corresponding to all columns; and S4, repeating steps S2 and S3 to obtain the sample height profiles corresponding to all groups of images, thereby obtaining the sample height profiles of the surface profile of the sample to be measured at different positions. Through the application, the problem of low sample height detection precision is solved.

Description

Technical Field

[0001] This invention belongs to the field of optical detection technology, and more specifically, relates to a method and system for obtaining sample height based on line spectrum confocal profile measurement. Background Technology

[0002] In today's society, high-precision measurement technology has permeated all walks of life, becoming an important driving force for promoting scientific and technological progress and industrial development. In fields such as semiconductor manufacturing, medical care, and scientific research, the demand for non-contact, wide-field-of-view, high-resolution, and high-speed measurement is constantly growing.

[0003] Linear spectral confocal profile measurement sensors fully utilize the advantages of optical confocal imaging and spectral analysis, extending traditional point-spectral confocal imaging to linear spectral confocal imaging. By establishing the correspondence between distance and wavelength through the principle of optical dispersion, and using a spectrometer to decode spectral information, they obtain the surface morphology of the sample under test. This enables high-precision and rapid measurement of sample surface morphology in fields such as semiconductors, precision manufacturing, 3C electronics, new energy, and medicine. In recent years, due to its advantages such as self-focusing, large measurement range, and wide measurement range, it has received widespread attention and application. However, due to its high-speed real-time measurement characteristics, peak extraction is often performed directly using the centroid method, resulting in coarse peak localization and affecting the sensor's measurement accuracy. Furthermore, in existing technologies, the accuracy of sample measurement is highly dependent on peak localization accuracy; higher peak localization accuracy leads to higher sample measurement accuracy. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a method and system for obtaining sample height based on line spectrum confocal profile measurement, which solves the problem of low sample height detection accuracy.

[0005] To achieve the above objectives, according to one aspect of the present invention, a method for obtaining sample height based on line spectral confocal profile measurement is provided, the method comprising the following steps:

[0006] S1 calculates the similarity between grayscale images of the surface contour of the sample to be tested that are continuously acquired at a set sampling time, and divides all grayscale images into multiple groups of images according to the similarity value.

[0007] S2 performs Gaussian blur mean processing and Gaussian apodization on each group of images to obtain the Gaussian-processed image. Then, it preprocesses each column of the gray-level matrix corresponding to the Gaussian-processed image to obtain the correspondence between the gray value and the position coordinates of each column in the gray-level matrix of the image, and determines the position coordinates of the pixel corresponding to the maximum gray value in each column.

[0008] S3 uses the known relationship between position coordinates and sample height to obtain the sample height corresponding to the maximum gray value in each column, thereby obtaining the sample height corresponding to each column of the gray matrix, and obtaining the sample height contour corresponding to the group of images from the sample heights of all columns.

[0009] S4 Repeat steps S2 and S3 to obtain the sample height profile corresponding to all groups of images, thereby obtaining the sample height profile of the surface profile of the sample under test at different locations.

[0010] More preferably, in step S2, the Gaussian blur mean processing is performed according to the following steps:

[0011] S21 divides the multiple images in each group into three units: the middle unit, the first unit, and the third unit. The second unit contains images before the middle unit, and the third unit contains images after the middle unit. The middle unit contains only one image.

[0012] S22 performs Gaussian filtering on each image in the first and third units;

[0013] S23 performs mean processing on the Gaussian filtered images in the first and third units to obtain the Gaussian blurred mean images of the first and third units, respectively.

[0014] More preferably, in step S2, the Gaussian apodization process is performed in the following manner:

[0015] I'=|I' 2 -(I' 1 +I' 3 ) / 2|

[0016] Where I' is the image after Gaussian apodization, I' 1 and I' 3 These are the images after Gaussian blurring and averaging of the first and third units, respectively. ‘2 It is the image of the intermediate unit.

[0017] More preferably, in step S22, the Gaussian filtering is performed according to the following formula:

[0018] GA(x,y)=(1 / (2πσ 2 ))*exp(-((x 2 +y 2 ) / (2σ 2 )))

[0019] Where GA(x,y) is the value of the Gaussian filter at point (x,y), σ is the parameter that controls the shape of the Gaussian filter, usually called the standard deviation, and x and y are the coordinates of the point in the image.

[0020] More preferably, in step S23, the mean processing is to average the Gaussian filter values ​​corresponding to all images in the first unit or the third unit.

[0021] More preferably, in step S2, the preprocessing is to perform noise reduction on the grayscale image. During the noise reduction process, spectral data in the grayscale image is extracted column by column, and grayscale values ​​greater than a preset threshold are taken as valid data.

[0022] More preferably, in the preprocessing process, when the sample to be tested is a grayscale image obtained for a transparent multilayer medium, the grayscale image needs to be divided into multiple spectral data intervals, and then each spectral data interval is processed as a grayscale matrix to obtain the correspondence between the maximum grayscale value of each column and the position coordinates of each grayscale matrix.

[0023] More preferably, in step S2, the maximum gray value is calculated using the centroid square algorithm.

[0024] More preferably, in step S1, the similarity is calculated according to the following relationship:

[0025]

[0026] Where SSIM(a,b) is the structural similarity index between images a and b, μ a and μ b σ is the average brightness of images a and b. a and σ b σ is the standard deviation of the brightness of images a and b. ab c1 and c2 are two constants, representing the covariance between the brightness of images a and b.

[0027] According to another aspect of the present invention, a measurement system for the sample height acquisition method based on line spectrum confocal profile measurement described above is provided. This system includes an image acquisition module, a Gaussian blur mean processing module, a Gaussian apodization module, a preprocessing module, and a sample height calculation module.

[0028] The image acquisition module is used to acquire grayscale images of the outer contour of the sample to be tested, and to divide the acquired grayscale images into multiple groups of images;

[0029] The Gaussian blur mean processing module is used to perform Gaussian blur mean processing on multiple images;

[0030] The Gaussian apodization module is used to perform Gaussian apodization on the image after Gaussian blur mean processing.

[0031] The preprocessing module is used to perform noise reduction on the image, obtain the correspondence between gray values ​​and position coordinates, and identify the maximum gray value.

[0032] The sample height calculation module is used to obtain the sample height corresponding to the maximum grayscale value by using the pre-calibrated relationship between the position coordinates and the sample height.

[0033] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0034] 1. In this invention, the image is processed by Gaussian blur mean processing and Gaussian apodization, which reduces the FWHM (half-peak width) of the axial response curve, improves the peak positioning accuracy, and enhances the axial resolution of the system. At the same time, the fuzzy filtering in Gaussian blur mean processing solves the problem of low system signal-to-noise ratio caused by actual processing errors, assembly errors and environmental factors.

[0035] 2. This invention addresses the needs of batch measurement of test samples by preprocessing the samples and adjusting the size of the CMOS camera. This reduces the amount of data that needs to be processed when extracting peak values ​​using the centroid square algorithm (the smaller CMOS camera size reduces the number of pixels required for the algorithm), thereby improving the speed of three-dimensional topography reconstruction. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the algorithm flow constructed according to a preferred embodiment of the present invention;

[0037] Figure 2 This is a flowchart illustrating the working process of a spectral confocal profile measurement sensor constructed according to a preferred embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram of the structure of a spectral confocal profile measurement sensor constructed according to a preferred embodiment of the present invention.

[0039] In all the accompanying drawings, the same reference numerals are used to denote the same elements or structures, wherein:

[0040] 1-Point light source LED, 2-Cylindrical mirror, 3-First collimating lens group, 4-Reflecting mirror, 5-Transmission diffraction grating, 6-Linear lens group, 11-Slit, 12-Second collimating lens group, 13-Holographic grating, 14-Imaging lens group, 15-Camera, 16-Sample to be tested. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0042] like Figure 1 As shown, a method for obtaining sample height based on line spectral confocal profile measurement includes the following steps:

[0043] S1 calculates the similarity between grayscale images of the surface contour of the sample to be tested that are continuously acquired at a set sampling time, and divides all grayscale images into multiple groups of images according to the similarity value.

[0044] In one embodiment of the present invention, the following is employed: Figure 2 The measurement system shown acquires multiple grayscale images. The similarity of grayscale information of the acquired images (SSIM) is calculated according to the following formula:

[0045]

[0046] Where SSIM(a,b) is the structural similarity index between images a and b, μ a and μ b σ is the average brightness of images a and b. a and σ b σ is the standard deviation of the brightness of images a and b. ab c1 and c2 are two constants, representing the covariance between the brightness of images a and b.

[0047] S2 performs Gaussian blur mean processing and Gaussian apodization on each group of images to obtain the Gaussian-processed image. Then, the Gaussian-processed image is preprocessed to extract the effective gray values ​​of each column in the gray matrix corresponding to the image. This yields the correspondence between the gray values ​​and position coordinates of each column in the gray matrix and determines the position coordinates of the pixel corresponding to the maximum gray value.

[0048] S21 In one embodiment of the present invention, due to the modulation of the CMOS sensor size, the camera frame rate is increased by 5 times. It is determined that the high similarity (SSIM>0.95) acquired image includes 5 images, namely I(N-2) to I(N+2). A suitable pair of σ and convolution kernel size is selected, with the first unit being I(N-2) to I(N-1) and the third unit being I(N+1) to I(N+2), and the middle unit being I(N).

[0049] S22 performs piecewise Gaussian filtering according to the following formula;

[0050] GA(x,y)=(1 / (2πσ 2 ))*exp(-((x 2 +y 2 ) / (2σ 2 (2)

[0051] In the formula, GA(x,y) is the value of the Gaussian filter at point (x,y), and σ (called Sigma) is a parameter that controls the shape of the Gaussian filter, usually referred to as the standard deviation.

[0052] S23 Mean Processing

[0053]

[0054] In the formula, I' 1 and I' 3 These are the images after Gaussian blurring and averaging of the first and third units, respectively.

[0055] Different σ values ​​will directly affect the full width at half maximum (FWHM) of the final peak curve.

[0056] The S24 Gaussian apodization generates a line spectrum confocal axial response peak curve with a smaller full width at half maximum (FWHM). The Gaussian apodization formula is as follows:

[0057] I′=|I′ 2 -(I′ 1 +I′ 3 ) / 2| (4)

[0058] In the formula, I' 2 for I(N), I' 2 For any processing of the image sequence N, I' is the result of Gaussian apodization.

[0059] Further, the Gaussian apodized grayscale image is extracted column by column to obtain the axial response curve with a small half-peak width. The centroid square algorithm is used to calculate the peak value of the axial response curve, and the x-coordinate corresponding to the intersection point is extracted. The height of the sample to be tested is obtained from the x-coordinate.

[0060] The centroid leveling method is a common peak extraction algorithm, and its calculation formula is as follows:

[0061]

[0062] In the formula, x c At the peak position, I i To acquire image grayscale values.

[0063] In one embodiment of the present invention, the preprocessing includes:

[0064] (1) Given the fast and wide-range measurement characteristics of the linear confocal dispersion sensor, the application scenarios of this type of sensor are mostly batch measurements of the same type of sample. Therefore, the first sample to be measured is predicted first, and the height range of the first sample is obtained through the three-dimensional reconstruction results. By the correspondence between the sensor height range and the CMOS camera pixel range, the CMOS camera size is adjusted to the corresponding pixel range, and then the subsequent samples of the same type are measured.

[0065] (2) The grayscale image is denoised. A threshold Q is set, and spectral data in the grayscale image is extracted by column. Data greater than the threshold Q is selected as valid data.

[0066] In one embodiment, preprocessing includes: based on the predicted value of the first sample, if the CMOS sensor size of the line spectral confocal profile measurement sensor is larger than R*C pixel, the CMOS sensor size is adjusted to (RX)*C pixel, thus increasing the camera frame rate. The image is filtered according to adaptive wavelet transform, a threshold Q of 20 is selected, and data greater than 20 in the grayscale image are extracted as valid data column by column, where a column refers to the direction of dispersion.

[0067] When the sensor measures a transparent multilayer medium, each layer of material reflects incident light of different wavelengths, thus forming data corresponding to different wavelengths in the dispersive image. Several spectral data intervals are obtained from the grayscale image using the symmetric zero-area method, and the peak positions are obtained by processing the several spectral data intervals separately.

[0068] S3 uses the known relationship between position coordinates and sample height to obtain the sample height corresponding to the maximum gray value, thereby obtaining the sample height corresponding to each column of the gray matrix. The sample height contour corresponding to the image group is obtained from the sample heights corresponding to all columns. The sample height contour is the contour formed by the sample height. The relationship between the known position coordinates and sample height can be obtained through pre-calibration.

[0069] S4 Repeat steps S2 and S3 to obtain the sample height profile corresponding to all groups of images, thereby obtaining the sample height profile of the surface profile of the sample under test at different locations.

[0070] In this embodiment, the following is adopted: Figure 3 The grayscale image acquired by the line spectral confocal profile measurement sensor includes a light source, a dual-axis confocal unit, and an imaging unit, wherein...

[0071] The light source provides a linear light source; the dual-axis confocal unit illuminates the surface of the sample under test with the light emitted from the light source and focuses the reflected light from the sample to form a confocal structure. It includes a first linear dispersion module, a second linear dispersion module, and a slit 11. The first and second linear dispersion modules are respectively distributed on both sides of the sample under test and are symmetrical about the sample. They have the same structure and form two symmetrical optical axes on both sides of the sample under test. The first linear dispersion module focuses the linear light from the light source onto the surface of the sample under test. The light of each wavelength is focused on a plane perpendicular to the surface of the sample under test in order of wavelength from high to low or from low to high. The second linear dispersion module receives the light reflected from the surface of the sample under test and focuses the light onto the slit. The slit is located behind the second linear dispersion module to filter out defocused light. The imaging unit images the light from the dual-axis confocal unit onto the camera 15 to achieve the measurement of the sample 16.

[0072] The light source includes a point light source 1 and a cylindrical mirror 2. The point light source 1 is a high-power white LED, and the cylindrical mirror 2 is used to convert the light emitted by the point light source into line light.

[0073] The back focal planes of slit 11 and cylindrical mirror 2 are conjugate. Cylindrical mirror 2 and the biaxial confocal unit form a confocal structure. Both the first linear dispersion module and the second dispersion module include a first collimating lens group 3, a reflecting mirror 4, a transmission diffraction grating 5, and a linear lens group 6. The first collimating lens group 3 is an achromatic collimating lens used to achromatic collimate linear rays; the reflecting mirror 4 is used to change the direction of light propagation; the transmission diffraction grating 5 is used to disperse white light; and the linear lens group is used to focus the dispersed light. The first collimating lens group 3 and the second collimating lens group have the same function, both used for achromatic collimation of white light; the linear lens group 6 focuses the dispersed light into a focused beam linearly aligned with the sample height. The imaging unit includes a second collimating lens group 12, a holographic grating 13, an imaging lens group 14, and a camera 15. The second collimating lens group 12 is used to collimate the light, the holographic grating 13 is used to disperse the light, and the imaging lens group 14 is used to image the dispersed light onto the camera surface.

[0074] Those skilled in the art will readily understand that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for obtaining sample height based on line spectral confocal profile measurement, characterized in that, The method includes the following steps: S1 For multiple grayscale images of the surface contour of the sample to be tested that are continuously acquired at a set sampling time, calculate the similarity between grayscale images at adjacent sampling times, and divide all grayscale images into multiple groups of images according to the similarity value. S2 performs Gaussian blur mean processing and Gaussian apodization on each group of images to obtain the Gaussian-processed image. Then, it preprocesses each column of the gray-level matrix corresponding to the Gaussian-processed image to obtain the correspondence between the gray value and the position coordinates of each column in the gray-level matrix of the image, and determines the position coordinates of the pixel corresponding to the maximum gray value in each column. S3 uses the pre-calibrated relationship between position coordinates and sample height to obtain the sample height corresponding to the maximum gray value in each column, thereby obtaining the sample height corresponding to each column of the gray matrix, and obtaining the sample height contour corresponding to the group of images from the sample heights corresponding to all columns. S4 Repeat steps S2 and S3 to obtain the sample height range corresponding to all groups of images, thereby obtaining the sample height profile of the surface profile of the sample under test at different locations. In step S2, the Gaussian blur mean processing is performed according to the following steps: S21 divides the multiple images in each group into three units: the middle unit, the first unit, and the third unit. The first unit contains images before the middle unit, and the third unit contains images after the middle unit. The middle unit contains only one image. S22 performs Gaussian filtering on each image in the first and third units; S23 performs mean processing on the Gaussian filtered images in the first and third units to obtain the Gaussian blurred mean images of the first and third units respectively. In step S2, the Gaussian apodization process is performed as follows: in, This is the image after Gaussian apodilation. and These are the images after Gaussian blurring and averaging of the first and third units, respectively. I ‘2 It is the image of the intermediate unit.

2. The sample height acquisition method based on line spectral confocal profile measurement as described in claim 1, characterized in that, In step S22, the Gaussian filtering is performed according to the following formula: in, GA(x, y) Is it a Gaussian filter at point (x, y) The value at that location, σ It is a parameter that controls the shape of the Gaussian filter, usually called the standard deviation, where x and y are the coordinates of a point in the image.

3. The sample height acquisition method based on line spectral confocal profile measurement as described in claim 1, characterized in that, In step S23, the mean processing is to average the Gaussian filter values ​​corresponding to all images in the first unit or the third unit.

4. A method for obtaining sample height based on line spectral confocal profile measurement as described in claim 1 or 2, characterized in that, In step S2, the preprocessing is to denoise the grayscale image. During the denoising process, spectral data in the grayscale image is extracted column by column, and grayscale values ​​greater than a preset threshold are taken as valid data.

5. The sample height acquisition method based on line spectrum confocal profile measurement as described in claim 4, characterized in that, In the preprocessing process, when the sample to be tested is a grayscale image obtained for a transparent multilayer medium, the grayscale image needs to be divided into multiple spectral data intervals, and then each spectral data interval is processed as a grayscale matrix to obtain the correspondence between the maximum grayscale value of each column and the position coordinates of each grayscale matrix.

6. A method for obtaining sample height based on line spectral confocal profile measurement as described in claim 1 or 2, characterized in that, In step S2, the maximum gray value is calculated using the centroid square algorithm.

7. A method for obtaining sample height based on line spectral confocal profile measurement as described in claim 1 or 2, characterized in that, In step S1, the similarity is calculated according to the following relationship: in, SSIM(a, b) yes a and b Structural similarity index between two images μa and μb It is an image a and b The average brightness, σa and σb It is an image a and b The standard deviation of brightness, σab It is an image a and b Covariance between brightness levels c1 and c2 They are two constants.

8. A measurement system for a sample height acquisition method based on line spectral confocal profile measurement as described in any one of claims 1-7, characterized in that, The system includes an image acquisition module, a Gaussian blur mean processing module, a Gaussian apodization module, a preprocessing module, and a sample height calculation module. The image acquisition module is used to acquire grayscale images of the outer contour of the sample to be tested, and to divide the acquired grayscale images into multiple groups of images; The Gaussian blur mean processing module is used to perform Gaussian blur mean processing on multiple images; The Gaussian apodization module is used to perform Gaussian apodization on the image after Gaussian blur mean processing. The preprocessing module is used to perform noise reduction on the image, obtain the correspondence between gray values ​​and position coordinates, and identify the maximum gray value. The sample height calculation module is used to obtain the sample height corresponding to the maximum grayscale value by using the pre-calibrated relationship between the position coordinates and the sample height.

Images