Laser confocal detection method and system for micro-defects on surface of optical lens

Abstract

The invention belongs to the technical field of confocal detection, and particularly relates to a laser confocal detection method and system for micro-defects on the surface of an optical lens, so as to solve the technical problems of noise pollution caused by image equal weight processing, difficulty in comprehensively representing defects by a single parameter and poor adaptability to complex areas in the conventional polarization detection method. The detection method comprises the following steps: S1, acquiring multiple frames of polarization intensity images corresponding to each polarization angle; s2, calculating an arithmetic mean value of all calculated global structure entropies to obtain an average structure entropy; s3, generating a polarization information composite feature map; and S4, according to the spatial distribution of all the pixel points which are judged to be micro-defect, determining the position and morphology of the micro-defect on the surface of the optical lens. According to the method, through the synergistic effect of information weighting, feature fusion and joint criteria, higher-precision detection of the micro-defects on the surface of the optical lens is realized.

Description

Technical Field

[0001] This invention belongs to the technical field of confocal inspection, and particularly relates to a laser confocal inspection method and system for micro-defects on the surface of optical lenses. Background Technology

[0002] Optical lenses are core components of precision optical systems, and their surface quality directly determines the imaging quality and performance stability of the entire system. During the manufacturing, processing, and use of optical lenses, micro-defects at the micrometer or even sub-micrometer scale inevitably occur on their surfaces. These micro-defects cause light scattering, absorption, and wavefront distortion, which can severely lead to decreased resolution, reduced signal-to-noise ratio, and even damage to high-power laser systems. Therefore, high-precision and high-efficiency detection of micro-defects on the surface of optical lenses is a crucial technological aspect of optical manufacturing and quality control.

[0003] Traditional inspection methods, including visual inspection and optical microscopy, suffer from high subjectivity, low resolution, and difficulty in detecting submicron-level and low-contrast defects. Laser confocal microscopy, with its high resolution, high signal-to-noise ratio, and optical slicing capabilities, has been applied in surface micro-morphology inspection. However, it primarily relies on light intensity contrast imaging, and its sensitivity remains insufficient for detecting micro-defects that only cause a change in light intensity that blurs the image due to a change in the polarization state of the light wave, easily leading to missed detections.

[0004] Polarization is a crucial physical property of light waves, highly sensitive to the fine structure, stress distribution, and material properties of object surfaces. When light interacts with the surface of an optical lens and its micro-defects, its polarization state undergoes a characteristic change. By measuring and analyzing the polarization information of reflected or scattered light, the contrast between micro-defects and intact surfaces can be enhanced, thereby identifying defects that are difficult to detect using traditional light intensity detection methods.

[0005] However, existing polarization-based detection methods still face challenges: First, when calculating Stokes parameters, multiple frames of polarization images are typically processed with equal weights, ignoring the potential differences in noise levels at different polarization angles. This can lead to a noisy frame contaminating the polarization calculation results. Second, a single polarization parameter cannot comprehensively represent all types of micro-defects; fusing multiple polarization features to improve the universality of detection remains a challenge. Third, existing detection methods cannot adapt to the background complexity and defect morphology of different regions, easily leading to misjudgments in complex texture areas or missed detections in sparse defect areas, thus affecting the overall detection accuracy. Summary of the Invention

[0006] This invention provides a laser confocal detection method and system for micro-defects on the surface of optical lenses, in order to solve the technical problems of existing polarization detection methods, such as noise pollution from image weighting, difficulty in fully representing defects with a single parameter, and poor adaptability to complex regions.

[0007] In a first aspect, the present invention provides a laser confocal detection method for micro-defects on the surface of an optical lens, comprising the following steps: S1. At at least three different polarization angles of the analyzer, a laser confocal microscope is used to scan the same micro-region on the surface of the optical lens point by point to obtain multiple frames of polarization intensity images corresponding to each polarization angle. S2 calculates the global structure entropy of each frame of polarization intensity image, and uses the global structure entropy as a weighting factor to weight multiple frames of polarization intensity images in order to calculate the Stokes parameter. , , Calculate the arithmetic mean of all calculated global structural entropies to obtain the average structural entropy; S3, based on the Stokes parameters , , Calculate the linear polarization degree distribution map and the polarization angle distribution map; perform spatial gradient operation on the polarization angle distribution map to obtain the polarization angle gradient map; multiply the square root of each pixel value in the linear polarization degree distribution map with the gradient value of the corresponding pixel in the polarization angle gradient map point by point to generate a polarization information composite feature map. S4. Determine the neighborhood size based on the relationship between the average structural entropy and the preset nonlinear function; filter the pixels in the polarization information composite feature map. When the composite feature value of a pixel is greater than a first threshold, and the number of pixels with a composite feature value greater than the first threshold in the neighborhood determined by the neighborhood size exceeds a second threshold, then the pixel is determined to be a micro-defect pixel; determine the location and morphology of micro-defects on the surface of the optical lens based on the spatial distribution of all pixels determined to be micro-defects.

[0008] Furthermore, in S1, at at least three different analyzer polarization angles, a laser confocal microscope is used to scan the same microscopic region on the surface of the optical lens point by point, including: The three different analyzers have polarization angles of 0°, 45° and 90°.

[0009] Further, in S2, the global structural entropy of the polarization intensity image for each frame is calculated, including: Gradient calculations are performed on a single-frame polarization intensity image to obtain an intensity gradient map; The gradient values ​​of the intensity gradient map are represented by 256 gray levels; Statistically calculate the probability of each gray level appearing in the intensity gradient map after representing 256 gray levels. ; According to Shannon's information entropy formula The global structural entropy of a single frame polarization intensity image is calculated. .

[0010] Furthermore, in S2, the Stokes parameters are calculated. , , ,include: The global structural entropy H(0), H(45), and H(90) of the three polarization intensity images I(0), I(45), and I(90) acquired when the polarization angle of the analyzer is 0°, 45°, and 90°, respectively, are calculated and weighted according to the following formula: , , .

[0011] Furthermore, in S3, a polarization information composite feature map is generated, including: According to the formula Obtain the linear polarization degree distribution map According to the formula Obtain the polarization angle distribution map ;in, , , These represent the Stokes parameters calculated at pixel coordinates (i, j). , , The value; The polarization angle distribution map was obtained by using a 3×3 Sobel operator. Perform convolution operation to obtain the polarization angle gradient map. ; Linear polarization degree distribution map Square root of each pixel value and polarization angle gradient map The gradient values ​​of corresponding pixels are multiplied point by point to obtain the polarization information composite feature map. The calculation formula is as follows: .

[0012] Further, in S4, determining the neighborhood size includes: According to the formula Calculate neighborhood size ,in, Let be the average structural entropy, exp() be the natural exponential function, and floor() be the floor function; The determined neighborhood is The square pixel area, and N in the formula is an odd number.

[0013] Further, in S4, determining the pixel as a micro-defect pixel includes: The first threshold It is 1.5 times the arithmetic mean of all non-zero pixel values ​​in the polarization information composite feature map; Second threshold for , where N is the neighborhood size, and floor() is the floor function; When the composite feature value of a pixel is greater than the first threshold, and The number of pixels in the neighborhood whose composite feature value is greater than the first threshold exceeds the second threshold. When this happens, the pixel is determined to be a micro-defect pixel.

[0014] Secondly, the present invention provides a laser confocal detection system for micro-defects on the surface of optical lenses, comprising the following modules: The acquisition module is used to scan the same microscopic area on the surface of the optical lens point by point using a laser confocal microscope at at least three different polarization angles of the analyzer, and acquire multiple frames of polarization intensity images corresponding to each polarization angle. The calculation module is used to calculate the global structure entropy of each frame of polarization intensity image, and then uses the global structure entropy as a weighting factor to weight multiple frames of polarization intensity images to calculate the Stokes parameters. , , Calculate the arithmetic mean of all calculated global structural entropies to obtain the average structural entropy; The generation module is used to generate based on the Stokes parameters. , , Calculate the linear polarization degree distribution map and the polarization angle distribution map; perform spatial gradient operation on the polarization angle distribution map to obtain the polarization angle gradient map; multiply the square root of each pixel value in the linear polarization degree distribution map with the gradient value of the corresponding pixel in the polarization angle gradient map point by point to generate a polarization information composite feature map. The determination module is used to determine the neighborhood size based on the relationship between the average structural entropy and a preset nonlinear function; to filter pixels in the polarization information composite feature map; when the composite feature value of a pixel is greater than a first threshold, and the number of pixels with composite feature values ​​greater than the first threshold in the neighborhood determined by the neighborhood size exceeds a second threshold, the pixel is determined to be a micro-defect pixel; and the location and morphology of micro-defects on the surface of the optical lens are determined based on the spatial distribution of all pixels determined to be micro-defects.

[0015] Furthermore, in the acquisition module, at at least three different analyzer polarization angles, a laser confocal microscope is used to scan the same microscopic area on the surface of the optical lens point by point, including: The three different analyzers have polarization angles of 0°, 45° and 90°.

[0016] Furthermore, the calculation module calculates the global structural entropy of each frame of polarization intensity image, including: Gradient calculations are performed on a single-frame polarization intensity image to obtain an intensity gradient map; The gradient values ​​of the intensity gradient map are represented by 256 gray levels; Statistically calculate the probability of each gray level appearing in the intensity gradient map after representing 256 gray levels. ; According to Shannon's information entropy formula The global structural entropy of a single frame polarization intensity image is calculated. .

[0017] The beneficial effects are as follows: This invention utilizes global structural entropy to weight multiple frames of acquired polarization intensity images, which can suppress the adverse effects of noise or low-quality images when calculating Stokes parameters, thus improving the accuracy and reliability of polarization fundamental information calculation. This invention deeply fuses linear polarization degree and polarization angle gradient information to generate a composite feature map of polarization information that can more comprehensively represent defect characteristics, enhancing the contrast between micro-defects and the intact background, and improving the detection capability for defects with diverse shapes and weak contrast. In the defect identification stage, a dual threshold criterion based on the pixel's own feature value and the number of feature points in its neighborhood is used for judgment, combining the intensity information of the point with the spatial clustering characteristics of the region, reducing the false alarm rate. This invention achieves higher precision detection of micro-defects on the surface of optical lenses through the synergistic effect of information weighting, feature fusion, and joint criteria. Attached Figure Description

[0018] Figure 1 This is a flowchart of a laser confocal detection method for micro-defects on the surface of optical lenses; Figure 2 This is a schematic diagram illustrating the acquisition of multi-frame polarization intensity images. Figure 3 A schematic diagram illustrating the weighted calculation of Stokes parameters and average structural entropy; Figure 4 This is a schematic diagram for determining micro-defect pixels. Detailed Implementation

[0019] An embodiment of the laser confocal detection method for micro-defects on the surface of optical lenses provided by this invention: like Figure 1As shown, a laser confocal method for detecting micro-defects on the surface of an optical lens includes the following steps: S1. At at least three different polarization angles of the analyzer, a laser confocal microscope is used to scan the same microscopic area on the surface of the optical lens point by point to obtain multiple frames of polarization intensity images corresponding to each polarization angle.

[0020] Specifically, a helium-neon laser with a wavelength of 633 nm was selected as the light source. After linearly polarized light was generated by a polarizer, it was focused onto the sample surface of the optical lens through the objective lens. The scanning galvanometer of the microscope drove the light spot to perform a two-dimensional grid scan on the sample surface. The reflected light passed through the objective lens again, and after passing through the analyzer, it was received by the photomultiplier tube (PMT). The polarization angle of the analyzer was set sequentially to 0°, 45°, and 90°. Each setting of a polarization angle completed a complete scan of the target area, thus obtaining three registered polarization intensity images, denoted as I(0), I(45), and I(90), respectively. Figure 2 As shown.

[0021] In an optional embodiment, in S1, at at least three different analyzer polarization angles, a laser confocal microscope is used to scan the same microscopic area on the surface of the optical lens point by point, including: The three different analyzers have polarization angles of 0°, 45° and 90°.

[0022] Specifically, the polarization transmission axis of the analyzer is adjusted to a 0° position parallel to the reference direction. The image acquisition device is activated to perform a full scan of the target micro-region on the surface of the optical lens and acquire the first frame of polarization intensity image, denoted as I(0). For example, the target micro-region is a 1mm × 1mm square area. The first frame of polarization intensity image records the distribution of light intensity through the analyzer in each pixel within the target micro-region at this polarization angle. The analyzer is rotated so that the polarization transmission axis is at 45° to the reference direction. The position and focal length of the image acquisition device and the optical lens remain unchanged. A second scan is performed on the identical 1mm × 1mm micro-region, and the second frame of polarization intensity image is acquired and saved, denoted as I(45). The polarization transmission axis of the analyzer is rotated to a 90° position perpendicular to the reference direction, and the same region is scanned again. The third frame of polarization intensity image is acquired and saved, denoted as I(90). Through the above process, three accurately registered polarization intensity images of the same physical region were obtained. These images reflect the light intensity information in different polarization directions, providing the necessary data basis for subsequent polarization state calculations.

[0023] S2 calculates the global structure entropy of each frame of polarization intensity image, and uses the global structure entropy as a weighting factor to weight multiple frames of polarization intensity images in order to calculate the Stokes parameter. , , Calculate the arithmetic mean of all calculated global structural entropies to obtain the average structural entropy.

[0024] Specifically, for the three polarization intensity images I(0), I(45), and I(90), gray-level histograms are calculated respectively, and the probability of each gray level appearing in the gray-level histogram is determined accordingly. The global structural entropy H(0), H(45), and H(90) of each polarization intensity image are calculated. Weighting factors for each polarization intensity image are defined. For the corresponding structural entropy The reciprocal of the formula is normalized. The Stokes parameters are calculated using a weighted formula. Average structural entropy. It equals the sum of H(0), H(45), and H(90) divided by 3, such as Figure 3 As shown.

[0025] In an optional embodiment, in S2, the global structural entropy of each frame of polarization intensity image is calculated, including: Gradient calculations are performed on a single-frame polarization intensity image to obtain an intensity gradient map; The gradient values ​​of the intensity gradient map are represented by 256 gray levels; Statistically calculate the probability of each gray level appearing in the intensity gradient map after representing 256 gray levels. ; According to Shannon's information entropy formula The global structural entropy of a single frame polarization intensity image is calculated. .

[0026] Specifically, a frame of polarization intensity image is selected for gradient calculation, usually implemented using a gradient operator (such as the Sobel operator). This operator calculates the rate of change of intensity of each pixel relative to its neighboring pixels in the horizontal and vertical directions, thereby generating a new image, i.e., a gradient map. In this gradient map, pixel values ​​are higher in areas of sharp edge or texture changes in the polarization intensity image, while pixel values ​​are lower in smooth areas.

[0027] The gradient values ​​in the gradient map are represented as follows: Assuming the gradient value range is 0 to 510, this range is linearly mapped to gray levels from 0 to 255. For example, a point with an original gradient value of 255 will be represented as having 127 gray levels. After representation, the number of pixels at each gray level in the 256-level gradient map is counted. For example, assuming the gradient map has a total of 1 million pixels, there are 50,000 pixels with gray level 0, 30,000 pixels with gray level 1, and so on. The probability of each gray level occurring can then be calculated. ,For example =0.05 =0.03. Substituting all 256 probability values ​​into the Shannon information entropy formula, each... The values ​​are calculated, summed, and then negative to obtain a single value H, which is the global structural entropy of the polarization intensity image of that frame.

[0028] In an optional embodiment, in S2, the calculation of the Stokes parameter... , , ,include: The global structural entropy H(0), H(45), and H(90) of the three polarization intensity images I(0), I(45), and I(90) acquired when the polarization angle of the analyzer is 0°, 45°, and 90°, respectively, are calculated and weighted according to the following formula: , , .

[0029] Specifically, the global structure entropy of each of the three polarization intensity images I(0), I(45), and I(90) is calculated to obtain three scalar values ​​H(0), H(45), and H(90). For example, after calculation, H(0) is 2.1, H(45) is 2.5, and H(90) is 1.9. The global structure entropy reflects the overall texture complexity of each frame of polarization intensity image. Polarization intensity images with more defects usually have higher global structure entropy. Calculate the Stokes parameter map pixel by pixel: For any pixel in the polarization intensity image, for example, the pixel with coordinates (10, 20), the intensity values ​​in the three frames of polarization intensity images are I(0)(10, 20) = 150, I(45)(10, 20) = 180, and I(90)(10, 20) = 50, respectively. Use the global structure entropy value obtained in the previous step as the weight and substitute it into the formula for calculation. The Stokes parameter of this pixel. Value 410, Stokes parameter The value is 220, Stokes parameter The value is 490. This weighted calculation process is repeated for each pixel in the polarization intensity image, generating three new images, namely the Stokes parametric maps. , and .

[0030] S3, based on the Stokes parameters , , The linear polarization degree distribution map and the polarization angle distribution map are calculated; spatial gradient operation is performed on the polarization angle distribution map to obtain the polarization angle gradient map; the square root of each pixel value in the linear polarization degree distribution map is multiplied point by point with the gradient value of the corresponding pixel in the polarization angle gradient map to generate a polarization information composite feature map.

[0031] In an optional embodiment, in S3, generating a polarization information composite feature map includes: According to the formula Obtain the linear polarization degree distribution map According to the formula Obtain the polarization angle distribution map ;in, , , These represent the Stokes parameters calculated at pixel coordinates (i, j). , , The value; The polarization angle distribution map was obtained by using a 3×3 Sobel operator. Perform convolution operation to obtain the polarization angle gradient map. ; Linear polarization degree distribution map Square root of each pixel value and polarization angle gradient map The gradient values ​​of corresponding pixels are multiplied point by point to obtain the polarization information composite feature map. The calculation formula is as follows: .

[0032] Specifically, using the calculations obtained above , , Three Stokes parametric plots were used to calculate the linear polarization degree (DoLP) and polarization angle (AoP) pixel by pixel. For example, for pixel (i, j), the Stokes parametric values ​​were respectively... =410 =220 =490. Therefore, the linear polarization degree of pixel (i, j) is approximately 1.31, which is usually normalized to between 0 and 1 due to physical limitations. The polarization angle of pixel (i, j) is approximately 0.57 radians. Repeat this process for all pixels to generate the DoLP and AoP maps.

[0033] Gradient calculations are performed on the generated polarization angle distribution map AoP: the standard 3×3 Sobel operator is used to convolve the AoP map to obtain a new polarization angle gradient map G. The polarization angle gradient map G can highlight regions where the polarization angle changes drastically, which often correspond to physical defects on the surface, such as scratches or pits.

[0034] The polarization information composite feature map F is generated by fusing information: For each pixel (i, j), the value is read from the DoLP map, the square root is calculated, and the gradient value at the corresponding position is read from the G map. The two values ​​are then multiplied. For example, if a pixel has a DoLP value of 0.64 and a polarization angle gradient value of 15, then the value of this pixel in the polarization information composite feature map F is 12. In this way, the composite feature map F simultaneously fuses information on polarization intensity and polarization direction change rate, making micro-defect features more prominent.

[0035] S4. Determine the neighborhood size based on the relationship between the average structural entropy and the preset nonlinear function; filter the pixels in the polarization information composite feature map. When the composite feature value of a pixel is greater than a first threshold, and the number of pixels with a composite feature value greater than the first threshold in the neighborhood determined by the neighborhood size exceeds a second threshold, then the pixel is determined to be a micro-defect pixel; determine the location and morphology of micro-defects on the surface of the optical lens based on the spatial distribution of all pixels determined to be micro-defects.

[0036] Specifically, a binary image is constructed from all pixels identified as micro-defects, where defect pixels have a value of 1 and background pixels have a value of 0. This binary image is processed using an 8-connected component labeling algorithm to divide spatially connected defect pixels into independent sets, each set representing an independent micro-defect. For each labeled micro-defect, its location coordinates are determined by calculating the geometric center containing the pixels, its area is calculated by counting the total number of pixels, and its size and contour are represented by calculating the length and width of the minimum bounding rectangle, thus completing the determination of the micro-defect's location and shape.

[0037] In an optional embodiment, in S4, determining the neighborhood size includes: According to the formula Calculate neighborhood size ,in, Let be the average structural entropy, exp() be the natural exponential function, and floor() be the floor function; The determined neighborhood is The square pixel area, and N in the formula is an odd number.

[0038] Specifically, the average structure entropy is obtained by adding the global structure entropy H(0), H(45), and H(90) previously calculated for the three original images I(0), I(45), and I(90) and then dividing by 3. For example, if H(0) = 2.1, H(45) = 2.5, and H(90) = 1.9, then the average structure entropy is... The value is 2.17. The average structural entropy reflects the overall surface texture complexity of the region.

[0039] Calculating the exponent: The natural exponent function exp results in approximately 0.272. Multiplying this by 4 and adding 1.5 gives 2.588. Rounding this down using the floor function gives 2. Multiplying the integer by 2 and adding 1 gives 5. For this example, the defined neighborhood size is a 5×5 pixel square. This formula makes the image more complex... The higher the value, the smaller the calculated N value, and vice versa, thus allowing subsequent defect judgments to adjust the analysis scale according to image characteristics.

[0040] In an optional embodiment, in S4, determining the pixel as a micro-defect pixel includes: The first threshold It is 1.5 times the arithmetic mean of all non-zero pixel values ​​in the polarization information composite feature map; Second threshold for , where N is the neighborhood size, and floor() is the floor function; When the composite feature value of a pixel is greater than the first threshold, and The number of pixels in the neighborhood whose composite feature value is greater than the first threshold exceeds the second threshold. When this happens, the pixel is determined to be a micro-defect pixel.

[0041] Specifically, two decision thresholds are calculated: the first threshold Second threshold Iterate through the generated polarization information composite feature map F, sum up all pixel values ​​greater than zero, divide by the total number of non-zero pixels, and obtain the arithmetic mean. 1.5 times this arithmetic mean is the first threshold. Assuming the arithmetic mean is 20.0, then Let's set it to 30.0. Calculate the second threshold using the neighborhood size N determined above. For example, N=5, substituting into the formula yields... =8.

[0042] Defect determination is performed using a dual threshold criterion: each pixel in the polarization information composite feature map F is examined using a sliding window approach. For example... Figure 4 As shown, assume the current center pixel coordinates are (x, y), and the value is F(x, y). Determine if F(x, y) is greater than... (i.e., 30.0). If the condition is not met, the pixel is not a defective pixel, and the next pixel is checked. If F(x, y) is greater than 30.0, the second condition is checked. Statistical analysis is performed on pixels with values ​​greater than 30.0 within a 5×5 neighborhood centered at (x, y). How many pixels are there in total, including the center point itself? Let's assume the count is 10. Because 10 is greater than the second threshold. (i.e., 8), so the second condition is also met. The center pixel (x, y) is then determined to be a micro-defect pixel. This process eliminates isolated noise points and retains only feature regions with a certain scale and intensity as defects.

[0043] Embodiments of the laser confocal detection system for micro-defects on the surface of optical lenses provided by this invention: A laser confocal detection system for micro-defects on the surface of an optical lens includes the following modules: The acquisition module is used to scan the same microscopic area on the surface of the optical lens point by point using a laser confocal microscope at at least three different polarization angles of the analyzer, and acquire multiple frames of polarization intensity images corresponding to each polarization angle. The calculation module is used to calculate the global structure entropy of each frame of polarization intensity image, and then uses the global structure entropy as a weighting factor to weight multiple frames of polarization intensity images to calculate the Stokes parameters. , , Calculate the arithmetic mean of all calculated global structural entropies to obtain the average structural entropy; The generation module is used to generate based on the Stokes parameters. , , Calculate the linear polarization degree distribution map and the polarization angle distribution map; perform spatial gradient operation on the polarization angle distribution map to obtain the polarization angle gradient map; multiply the square root of each pixel value in the linear polarization degree distribution map with the gradient value of the corresponding pixel in the polarization angle gradient map point by point to generate a polarization information composite feature map. The determination module is used to determine the neighborhood size based on the relationship between the average structural entropy and a preset nonlinear function; to filter pixels in the polarization information composite feature map; when the composite feature value of a pixel is greater than a first threshold, and the number of pixels with composite feature values ​​greater than the first threshold in the neighborhood determined by the neighborhood size exceeds a second threshold, the pixel is determined to be a micro-defect pixel; and the location and morphology of micro-defects on the surface of the optical lens are determined based on the spatial distribution of all pixels determined to be micro-defects.

[0044] In an optional embodiment, the acquisition module performs point-by-point scanning of the same micro-region on the surface of the optical lens using a laser confocal microscope at at least three different analyzer polarization angles, including: The three different analyzers have polarization angles of 0°, 45° and 90°.

[0045] In an optional embodiment, the calculation module calculates the global structural entropy of each frame of polarization intensity image, including: Gradient calculations are performed on a single-frame polarization intensity image to obtain an intensity gradient map; The gradient values ​​of the intensity gradient map are represented by 256 gray levels; Statistically calculate the probability of each gray level appearing in the intensity gradient map after representing 256 gray levels. ; According to Shannon's information entropy formula The global structural entropy of a single frame polarization intensity image is calculated. .

[0046] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

Claims

1. A method for laser confocal detection of surface microdefects of an optical lens, characterized in that, The method comprises the following steps: S1, under at least three different polarizer polarization angles, a same micro area on the surface of the optical lens is scanned point by point by using a laser confocal microscope to obtain multiple frames of polarization intensity images corresponding to each polarization angle; S2, calculate the global structure entropy of each frame of polarization intensity image, and use the global structure entropy as a weight factor to weight the multi-frame polarization intensity image, so as to calculate the Stokes parameter , , ; calculate the arithmetic mean of all calculated global structure entropies to obtain the average structure entropy; S3, calculating a linear polarization degree distribution map and a polarization angle distribution map based on the Stokes parameters , , , performing a spatial gradient operation on the polarization angle distribution map to obtain a polarization angle gradient map; and multiplying the square root of each pixel value in the linear polarization degree distribution map with the gradient value of the corresponding pixel in the polarization angle gradient map point by point to generate a polarization information composite feature map. S4, according to the relationship between the average structure entropy and a preset nonlinear function, a neighborhood size is determined; pixel points in the polarization information composite feature map are screened, when the composite feature value of a pixel point is greater than a first threshold value, and within a neighborhood of the pixel point determined by the neighborhood size, the number of pixel points with the composite feature value greater than the first threshold value exceeds a second threshold value, the pixel point is determined as a micro defect pixel point; according to the spatial distribution of all the pixel points determined as the micro defect pixel points, the position and morphology of the micro defects on the surface of the optical lens are determined.

2. The method for laser confocal detection of surface micro-defects of optical lenses according to claim 1, characterized in that, In S1, under at least three different polarizer polarization angles, a same micro area on the surface of the optical lens is scanned point by point by using a laser confocal microscope, comprising: The three different polarizer polarization angles are 0°, 45° and 90° respectively.

3. The method of laser confocal detection of surface micro-defects of optical lenses according to claim 1, characterized in that, In S2, the global structure entropy of each frame of polarization intensity image is calculated, comprising: Gradient operation is performed on a single frame of polarization intensity image to obtain an intensity gradient map; The gradient values of the intensity gradient map are represented by 256 levels of gray scale; The probability of occurrence of each gray level in the intensity gradient map after statistical 256-level gray scale representation ; According to the Shannon information entropy formula The global structure entropy of the single-frame polarization intensity image is calculated .

4. The method for laser confocal detection of surface micro-defects of optical lenses according to claim 3, characterized in that, In S2, the Stokes parameter is calculated , , , comprising: Global structure entropies H(0), H(45), H(90) of the three frames of polarization intensity images I(0), I(45), I(90) obtained when the polarization angle of the polarizer is 0°, 45° and 90° respectively are calculated, and weighted calculation is performed according to the following formula: , , .

5. The method of laser confocal detection of surface micro-defects of optical lenses according to claim 1, characterized in that, In S3, the polarization information composite feature map is generated, comprising: According to the formula , a linear polarization degree distribution map is obtained , according to the formula , a polarization angle distribution map is obtained ; wherein, , , respectively represent the numerical value of the Stokes parameter , , calculated at the pixel coordinates (i, j). The polar angle distribution map is convolved with a 3x3 Sobel operator to obtain a polar angle gradient map ; a linear polarization degree distribution map a square root of each pixel value in the polarization angle gradient map the gradient values of the corresponding pixels are multiplied point by point to obtain a polarization information composite feature map The calculation formula of the polarization information composite feature map is as follows: .

6. The method of laser confocal detection of surface microdefects of an optical lens according to any one of claims 1 to 5, characterized in that, In S4, the neighborhood size is determined, comprising: The neighborhood size is calculated according to the formula wherein wherein is the average structural entropy, exp() is the natural exponential function, and floor() is the floor function. The determined neighborhood is a square pixel region of N pixels, and N is an odd number in the formula.

7. The method of laser confocal detection of surface microdefects of optical lenses according to claim 6, characterized in that, In S4, the pixel point is determined as a micro defect pixel point, comprising: the first threshold 1.5 times the arithmetic mean of all non-zero pixel values in the polarized information composite feature map; the second threshold for where N is the neighborhood size, floor() is the floor function; When the composite feature value of the pixel point is greater than the first threshold value, and the number of pixels in the neighborhood whose composite feature value is greater than the first threshold value exceeds the second threshold value the pixel point is determined as a micro-defect pixel point.

8. A laser confocal detection system for surface micro-defects of an optical lens, characterized in that, The method comprises the following modules: The acquisition module is configured to, under at least three different polarizer polarization angles, scan a same micro area on the surface of the optical lens point by point by using a laser confocal microscope to obtain multiple frames of polarization intensity images corresponding to each polarization angle; A computing module is configured to calculate the global structure entropy of each frame of polarized intensity image, and use the global structure entropy as a weight factor to weight the multiple frames of polarized intensity image, so as to calculate the Stokes parameter , , ; The calculation module is configured to calculate the arithmetic mean value of all the calculated global structure entropies to obtain an average structure entropy; generating a linear polarization degree distribution map and a polarization angle distribution map based on the Stokes parameters , , , performing a spatial gradient operation on the polarization angle distribution map to obtain a polarization angle gradient map; and multiplying the square root of each pixel value in the linear polarization degree distribution map with the gradient value of the corresponding pixel in the polarization angle gradient map point by point to generate a polarization information composite feature map. The determination module is configured to, according to the relationship between the average structure entropy and a preset nonlinear function, determine a neighborhood size; screen pixel points in the polarization information composite feature map, when the composite feature value of a pixel point is greater than a first threshold value, and within a neighborhood of the pixel point determined by the neighborhood size, the number of pixel points with the composite feature value greater than the first threshold value exceeds a second threshold value, the pixel point is determined as a micro defect pixel point; according to the spatial distribution of all the pixel points determined as the micro defect pixel points, the position and morphology of the micro defects on the surface of the optical lens are determined.

9. The system for laser confocal detection of surface micro-defects of optical lenses according to claim 8, characterized in that, In the acquisition module, under at least three different polarizer polarization angles, a same micro area on the surface of the optical lens is scanned point by point by using a laser confocal microscope, comprising: The three different polarizer polarization angles are 0°, 45° and 90° respectively.

10. The system for laser confocal detection of surface micro-defects of optical lenses according to claim 8, characterized in that, In the calculation module, the global structure entropy of each frame of polarization intensity image is calculated, comprising: Gradient operation is performed on a single frame of polarization intensity image to obtain an intensity gradient map; The gradient values of the intensity gradient map are represented by 256 levels of gray scale; The probability of occurrence of each gray level in the intensity gradient map after statistical 256-level gray scale representation ; According to the Shannon information entropy formula The global structure entropy of the single-frame polarization intensity image is calculated .

Images