A computer vision-based automatic laser damage judgment method

Abstract

The present application relates to laser damage detection, in particular to a kind of laser damage automatic judging method based on computer vision, the laser radiation before image of the element to be measured is collected, and the image after laser radiation, and the image before laser radiation, the image after laser radiation are preprocessed;Determine the region of interest in the image before laser radiation, the image after laser radiation based on the possible damage location of the element to be measured, and adopt sliding window and move along the region of interest in the image before laser radiation, the image after laser radiation simultaneously;Similarity index between corresponding local window image in each sliding window is calculated based on similarity measurement model;Similarity index is input into classifier, and whether the element to be measured appears laser damage is judged according to the output result of classifier;The technical scheme provided by the present application can effectively overcome the defects that the prior art cannot accurately, efficiently judge whether optical element appears laser damage, and intuitively reflect laser damage detection result.

Description

TECHNICAL FIELD

[0001] The present application relates to laser damage detection, in particular to a computer vision-based automatic laser damage judgment method. BACKGROUND

[0002] The laser damage threshold is one of the important parameters for measuring the laser damage resistance of an optical element, and whether the optical element is damaged by laser is the key to the laser damage threshold test of the optical element. Therefore, in order to determine the laser damage threshold of the optical element, a method for detecting whether the optical element is damaged by laser after being irradiated by laser is needed.

[0003] For laser damage detection, the first method used is the phase contrast microscope method and the plasma flash method. The phase contrast microscope method is a standard detection method proposed by the international standard ISO 11254, and the detection result is relatively accurate, but the workload is large, the time is long, and the efficiency is low. The plasma flash method is greatly affected by human factors during detection, and the flash produced by optical elements of different materials when they are damaged by laser is also different.

[0004] Later, the industry proposed a variety of specific detection methods based on whether the optical performance of the optical element changes, such as scattered light intensity method, photothermal deflection method, and photoacoustic method, etc., but these methods have great limitations in intuitiveness. SUMMARY

[0005] (I) Technical problems to be solved

[0006] In view of the above shortcomings of the prior art, the present application provides a computer vision-based automatic laser damage judgment method, which can effectively overcome the defects of the prior art that cannot accurately and efficiently judge whether the optical element is damaged by laser, and intuitively reflect the laser damage detection result.

[0007] (II) Technical solutions

[0008] In order to achieve the above purpose, the present application is realized by the following technical solutions:

[0009] A computer vision-based automatic laser damage judgment method, comprising the following steps:

[0010] S1, collecting the pre-laser irradiation image and the post-laser irradiation image of the to-be-tested element, and pre-processing the pre-laser irradiation image and the post-laser irradiation image;

[0011] S2, determining the region of interest in the pre-laser irradiation image and the post-laser irradiation image based on the possible damage position of the to-be-tested element, and moving the sliding window along the region of interest in the pre-laser irradiation image and the post-laser irradiation image synchronously;

[0012] S3, calculating a similarity index between corresponding local window images in each sliding window based on a similarity measurement model;

[0013] S4, inputting the similarity index into a classifier, and judging whether the laser damage occurs in the element under test according to an output result of the classifier.

[0014] Preferably, the laser radiation pre-image and the laser radiation post-image of the element under test are collected in S1, and the laser radiation pre-image and the laser radiation post-image are preprocessed, including:

[0015] A 3*3 template is adopted, and the template is moved so that the center point of the template coincides with each pixel point in the image;

[0016] The pixel gray value in the region covered by the template is obtained, and the pixel gray value at the middle position is taken out after being sorted from small to large, and the pixel gray value at the middle position is assigned to the pixel point at the center point of the template.

[0017] Preferably, the region of interest in the laser radiation pre-image and the laser radiation post-image is determined based on the possible damage position of the element under test in S2, including:

[0018] A rectangular region with the center of the image as the center point is selected as the region of interest ROI;

[0019] The length of the region of interest ROI is half of the length of the original image, and the width of the region of interest ROI is half of the width of the original image.

[0020] Preferably, the sliding window is moved along the region of interest in the laser radiation pre-image and the laser radiation post-image in S2, including:

[0021] A 300 pixel*300 pixel sliding window is set, and the left upper corner of the region of interest ROI in the laser radiation pre-image and the laser radiation post-image is started with a step of 80, and the local window images in the laser radiation pre-image and the laser radiation post-image are sequentially cut from left to right and from top to bottom.

[0022] Preferably, the similarity index between corresponding local window images in each sliding window is calculated based on a similarity measurement model in S3, including:

[0023] The similarity index SSIM between corresponding local window images in each sliding window is calculated by the following formula:

[0024] SSIM(x,y)=[l(x,y)] α [c(x,y)] β [s(x,y)] γ -λ·meanDiff

[0025] wherein x, y are local window image signals in the image before laser irradiation, the image after laser irradiation respectively; l(x, y) is a luminance function, c(x, y) is a contrast function, s(x, y) is a structure information function, a, b, g are coefficients for adjusting the proportion of luminance, contrast, structure information in the similarity evaluation, a, b, g are all greater than 0;

[0026] meanDiff is a penalty term in the similarity evaluation, meanDiff = |m y - m x |, m x is the average gray value of the local window image in the image before laser irradiation, x i is the gray value of the i-th pixel point of the local window image in the image before laser irradiation, m y is the average gray value of the local window image in the image after laser irradiation, y i is the gray value of the i-th pixel point of the local window image in the image after laser irradiation, N is the number of pixel points of the local window image in the image before laser irradiation, the image after laser irradiation, and l is a penalty term coefficient, l = 0.01.

[0027] Preferably, the luminance function l(x, y) is defined as:

[0028]

[0029] The contrast function c(x, y) is defined as:

[0030]

[0031] The structure information function s(x, y) is defined as:

[0032]

[0033] wherein s x is the gray standard deviation of the local window image in the image before laser irradiation, s y is the gray standard deviation of the local window image in the image after laser irradiation, s x s y is the gray covariance between the local window images in the image before laser irradiation, the image after laser irradiation,

[0034] C1, C2, C3 are constants for avoiding the denominator being 0, C1 = (K1L) 2 , C2 = (K2L) 2 , K1 = 0.01, K2 = 0.03, L = 255.

[0035] Preferably, the similarity index is input into the classifier in S4, and whether the laser damage occurs in the element under test is determined according to the output result of the classifier, including:

[0036] According to the minimum similarity index SSIM between all the local window images in each sliding window min The classifier is constructed, and the minimum similarity index SSIM min is input into the classifier.

[0037] When the minimum similarity index SSIM min <0.8, the classifier outputs the determination result that the laser damage occurs in the element under test, otherwise the classifier outputs the determination result that the laser damage does not occur in the element under test.

[0038] (Three) beneficial effects

[0039] Compared with the prior art, the laser damage automatic determination method based on computer vision provided by the present application collects the pre-laser radiation image and the post-laser radiation image of the element under test, uses a spatial domain median filter to reduce the background noise interference, locates the possible damage position of the element under test through a fast search strategy, combines a similarity measurement model and constructs a classifier to accurately and efficiently determine whether the optical element has laser damage after laser radiation, and can intuitively reflect the laser damage detection result. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0041] Figure 1 The flowchart of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0043] A computer vision-based automatic laser damage judgment method, as shown in Figure 1 The method comprises the following steps: ①collecting a pre-laser radiation image (i.e., a standard image without laser damage) and a post-laser radiation image of a to-be-tested element, and preprocessing the pre-laser radiation image and the post-laser radiation image, specifically including:

[0044] A 3*3 template is adopted, and the template is moved so that the center point of the template coincides with each pixel point in the image.

[0045] The pixel gray value in the region covered by the template is obtained, the pixel gray value at the middle position is taken out after being sorted from small to large, and the pixel gray value at the middle position is assigned to the pixel point at the center point of the template.

[0046] The above technical solution can effectively reduce background noise interference and effectively weaken the blurring effect of the target boundary by using a spatial domain median filter to preprocess the pre-laser radiation image and the post-laser radiation image.

[0047] ②determining a region of interest in the pre-laser radiation image and the post-laser radiation image based on the possible damage position of the to-be-tested element, and moving a sliding window along the region of interest in the pre-laser radiation image and the post-laser radiation image synchronously.

[0048] 1) determining a region of interest in the pre-laser radiation image and the post-laser radiation image based on the possible damage position of the to-be-tested element, comprising:

[0049] A rectangular region with the center of the image as the center point is selected as the region of interest ROI.

[0050] The length of the region of interest ROI is half of the length of the original image, and the width of the region of interest ROI is half of the width of the original image.

[0051] 2) moving a sliding window along the region of interest in the pre-laser radiation image and the post-laser radiation image synchronously, comprising:

[0052] A 300*300 sliding window is set, and the left upper corner of the region of interest ROI in the pre-laser radiation image and the post-laser radiation image is started with a step of 80, and the local window images in the pre-laser radiation image and the post-laser radiation image are sequentially cut from left to right and from top to bottom.

[0053] The technical scheme is characterized in that, after pre-processing of the image before laser irradiation and the image after laser irradiation, the local window images are processed by using a fast sliding window mechanism. In actual application, the position of laser irradiation is usually located at the center of the optical element, i.e. the center of the image field, so a rectangular region with the image center as the center point is selected as the region of interest (ROI), and the region of interest (ROI) in the image before laser irradiation and the image after laser irradiation includes the possible damage position of the element to be tested.

[0054] ③calculating the similarity index between the corresponding local window images in each sliding window based on a similarity measurement model, specifically including:

[0055] The similarity index between the corresponding local window images in each sliding window is calculated by using the following formula:

[0056] SSIM(x,y)=[l(x,y)] α [c(x,y)] β [s(x,y)] γ -λ·meanDiff

[0057] wherein x and y are respectively the local window image signals in the image before laser irradiation and the image after laser irradiation; l(x,y) is a luminance function, c(x,y) is a contrast function, s(x,y) is a structure information function, and α, β and γ are respectively coefficients for adjusting the proportions of luminance, contrast and structure information in the similarity evaluation, and α, β and γ are all greater than 0.

[0058] meanDiff is a penalty term in the similarity evaluation, and meanDiff=|μ y -μ x |, μ x is the average gray value of the local window image in the image before laser irradiation, x i is the gray value of the i-th pixel point of the local window image in the image before laser irradiation, μ y is the average gray value of the local window image in the image after laser irradiation, y i is the gray value of the i-th pixel point of the local window image in the image after laser irradiation, N is the number of pixel points of the local window image in the image before laser irradiation and the image after laser irradiation, and λ is a penalty term coefficient, λ=0.01.

[0059] Specifically, the luminance function l(x,y) is defined as:

[0060]

[0061] The contrast function c(x,y) is defined as:

[0062]

[0063] The structural information function s(x, y) is defined as:

[0064]

[0065] Wherein, σ x is the gray scale standard deviation of the local window image in the image before laser irradiation, σ y is the gray scale standard deviation of the local window image in the image after laser irradiation, σ x σ y is the gray scale covariance between the local window images in the image before laser irradiation and the image after laser irradiation,

[0066] C1, C2, C3 are constants for avoiding denominator being zero, C1 = (K1L) 2 , C2 = (K2L) 2 , K1 = 0.01, K2 = 0.03, L = 255.

[0067] (4) inputting the similarity index into the classifier, judging whether the tested element appears laser damage according to the output result of the classifier, specifically including:

[0068] constructing the classifier according to the minimum similarity index SSIM min between all the corresponding local window images in each sliding window, and inputting the minimum similarity index SSIM min into the classifier;

[0069] when the minimum similarity index SSIM min < 0.8, the classifier outputs the judging result that the tested element appears laser damage, otherwise the classifier outputs the judging result that the tested element does not appear laser damage.

[0070] In theory, the value range of the similarity index SSIM between the corresponding local window images in each sliding window is 0-1, the closer the similarity index SSIM to 0, the greater the probability that the tested element appears laser damage.

[0071] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A computer vision-based automatic judgment method for laser damage, characterized by: The method comprises the following steps: S1, collecting a pre-laser radiation image and a post-laser radiation image of a to-be-tested element, and preprocessing the pre-laser radiation image and the post-laser radiation image; S2, determining a region of interest in the pre-laser radiation image and the post-laser radiation image based on a possible damage position of the to-be-tested element, and moving a sliding window along the region of interest in the pre-laser radiation image and the post-laser radiation image synchronously; S3, calculating a similarity index between corresponding local window images in each sliding window based on a similarity measurement model; S4, inputting the similarity index into a classifier, and judging whether the to-be-tested element is damaged by laser according to an output result of the classifier; The similarity index between corresponding local window images in each sliding window is calculated based on the similarity measurement model in S3, comprising: The similarity index SSIM between corresponding local window images in each sliding window is calculated by using the following formula: SSIM(x,y) = [l(x,y)] α [c(x,y)] β [s(x,y)] γ - λ - meanDiff Wherein, x and y are local window image signals in the pre-laser radiation image and the post-laser radiation image; l(x, y) is a luminance function, c(x, y) is a contrast function, s(x, y) is a structure information function, and α, β and γ are coefficients for adjusting the proportion of luminance, contrast and structure information in similarity evaluation, all of which are greater than 0; meanDiff is a penalty term in the similarity evaluation, meanDiff = |μ y -μ x |, μ x is the gray mean value of the local window image in the image before laser irradiation, x i is the gray value of the i-th pixel point of the local window image in the image before laser irradiation, μ y is the gray mean value of the local window image in the image after laser irradiation, y i is the gray value of the i-th pixel point of the local window image in the image after laser irradiation, N is the number of pixel points of the local window image in the image before laser irradiation, the image after laser irradiation, and λ is a penalty term coefficient, λ = 0.

01.

2. The computer vision-based automatic judging method of laser damage according to claim 1, characterized in that: The pre-laser radiation image and the post-laser radiation image of the to-be-tested element are collected and preprocessed in S1, comprising: A 3*3 template is used, and the template is moved so that the center point of the template coincides with each pixel point in the image; The pixel gray value in the template coverage area is obtained, sorted from small to large, and the pixel gray value at the middle position is taken out, which is assigned to the pixel point at the center point of the template.

3. The computer vision-based automatic judging method of laser damage according to claim 2, characterized in that: The region of interest in the pre-laser radiation image and the post-laser radiation image is determined based on the possible damage position of the to-be-tested element in S2, comprising: A rectangular region with the center of the image as the center point is selected as the region of interest ROI; Wherein, the length of the region of interest ROI is half of the length of the original image, and the width of the region of interest ROI is half of the width of the original image.

4. The computer vision-based automatic judging method of laser damage according to claim 3, characterized in that: The sliding window is moved along the region of interest in the pre-laser radiation image and the post-laser radiation image synchronously in S2, comprising: A 300*300 sliding window is set, and the left upper corner of the region of interest ROI in the pre-laser radiation image and the post-laser radiation image is started with a step of 80, and the local window images in the pre-laser radiation image and the post-laser radiation image are sequentially cut from left to right and from top to bottom.

5. The computer vision-based automatic judging method of laser damage according to claim 1, characterized in that: The luminance function l(x, y) is defined as: The contrast function c(x, y) is defined as: The structure information function s(x, y) is defined as: wherein σ x is the standard deviation of the grey values of the local window image in the image before the laser radiation, σ y is the standard deviation of the grey values of the local window image in the image after the laser radiation, σ x σ y is the covariance of the grey values between the local window image in the image before the laser radiation and the local window image in the image after the laser radiation, C1, C2, C3 are constants to avoid denominator zero, C1 = (K1L) 2 , C2 = (K2L) 2 , K1 = 0.01, K2 = 0.03, L = 255.

6. The computer vision-based automatic judging method of laser damage according to claim 1, characterized in that: The similarity index is input into the classifier in S4, and whether the to-be-tested element is damaged by laser is judged according to the output result of the classifier, comprising: According to the minimum similarity index SSIM between all the local window images in each sliding window min Constructing a classifier and inputting the minimum similarity index SSIM min Inputting the classifier; When the minimum similarity index SSIM min When the minimum similarity index SSIM <0.8, the classifier outputs a judgment result that the tested element has laser damage, otherwise the classifier outputs a judgment result that the tested element has no laser damage.

Images