An Adaptive Focusing Evaluation Method for Microscopic Optical Measurement of Film-Coating Aperture in Turbine Blades

Abstract

An adaptive focusing evaluation method for microscopic optical measurement of film pores in turbine blades is disclosed. Based on a microscopic vision measurement system, the method first optimizes the linear parameters of the guided filter using a penalty function method on the original microscopic image I0, and then performs guided filtering to obtain a filtered image. Next, high-contrast image enhancement is performed on the filtered image to obtain an enhanced image that is beneficial for focusing evaluation. The mean absolute deviation (MAD) value is calculated for each pixel in the enhanced image, and a suitable focusing evaluation scale is selected for focusing evaluation. Then, the mean absolute deviation (MAD) value of each pixel in the focusing evaluation image is calculated again, and a suitable filtering scale is selected for mean filtering to obtain a filtered focusing evaluation image. This invention integrates an adaptive scale selection method into the focusing evaluation method and the filtering method, which can effectively reduce the influence of microstructure scale issues on the focusing evaluation of microscopic images.

Description

Technical Field

[0001] This invention belongs to the field of microscopic optical measurement technology, specifically relating to an adaptive focusing evaluation method for microscopic optical measurement of film pores in turbine blades. Background Technology

[0002] With the continuous development of microscopic optical measurement technology, zoom-based 3D measurement technology has become a research focus in the field of microstructure 3D measurement. Zoom measurement is a surface topography measurement method that uses the sharpness of surface images in an optical microscope to determine the surface height at each location along the surface. This technique has been applied to the microscopic 3D measurement of film pores in turbine blades. Determining the height of a point requires comparing the focus measure of pixels at the same pixel coordinate in each image of a focused image sequence. Calculating the focus measure is a crucial basis for accurately determining the height. Focus evaluation methods, as a means of calculating the focus measure, are an important research area in depth-of-field measurement.

[0003] When using the aforementioned focus evaluation operator to calculate the focus evaluation on an image, the texture and color of the surface of the object being measured are unpredictable. The presence of these textures and colors will affect the calculation of the focus measurement. This situation is very common on the surface of the film vents of turbine engine blades, which are made of metal and processed by laser. It is necessary to process them.

[0004] Uniform color areas in an image can lead to errors in focus evaluation in that area. This is because a surface with completely uniform color will have very low gradients between pixels in its microscopic digital image, regardless of whether it is in sharp focus, causing common focus measurement operators to fail. In actual measurement, this manifests as follows: Focusing 3D reconstruction technology involves digital image focus evaluation methods (Liu Mengchen, Pang Changtao, Hao Xue. Research on Sobel operator in the process of sharpness evaluation of air film aperture images [J]. Mechanical Science and Technology, 2020, 39(7):1071-1076), which relies on the inherent irregular micron and nano-scale microstructures of the object surface to ensure the formation of pixel contrast in the focus area. However, the specific scale of these micron and nano-scale microstructures cannot be controlled. The reflected light is projected onto the camera sensor after passing through objective lenses with different optical magnifications, and the number of pixels occupied varies: such as Figure 1 As shown in (a), the image of some small microstructures can be encompassed by a single pixel. In this case, the focus evaluation value of that pixel can accurately represent the actual focus status of that point; while... Figure 1 As shown in (b), the images of some larger structures occupy multiple pixels. On the pixels through which their edges pass, the focus evaluation value can still correctly represent the actual focus status of that point. However, on the pixels inside the structure, due to the small grayscale gradient between them and the surrounding pixels, the focus evaluation function has difficulty in correctly detecting the actual focus status of that point.

[0005] In summary, there is an urgent need to propose a focused evaluation method to address the problem of inconsistent micro- and nano-scale microstructures on the surface of the tested object. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, the present invention aims to provide an adaptive focusing evaluation method for microscopic optical measurement of air film pores in turbine blades, which solves the problem of inconsistent microscale and nanoscale microstructures on the surface of the measured object and improves the accuracy of the focusing evaluation method.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] An adaptive focusing evaluation method for microscopic optical measurement of film pores in turbine blades, based on a microscopic vision measurement system, includes the following steps:

[0009] Step 1: The original microscopic image I0 is preprocessed using a microscopic image preprocessing algorithm that combines guided filtering and high contrast preservation to obtain the enhanced image I2.

[0010] Step 2: Apply MAD-based Brenner-2D adaptive focus evaluation to the enhanced image I2 to obtain the focus evaluation image I. m ;

[0011] Step 3, focus on the evaluation image I m The filtered focus evaluation image I is obtained by using MAD-based adaptive window mean filtering of the focus evaluation value. m ′.

[0012] Step 1 specifically involves: optimizing the linear parameters of the guided filter using the penalty function method on the original microscopic image I0, performing guided filtering to obtain the filtered image I1; then performing high-contrast image enhancement on the filtered image I1: first performing Gaussian filtering to obtain the background image I... b The filtered image I1 and the background image I b Subtraction yields the detailed image I. f , to detail image I f Multiply by the enhancement coefficient n, and then add to the filtered image I1 to obtain the enhanced image I2.

[0013] Step 2 specifically involves: evaluating the enhanced image I2 using Brenner-2D adaptive focus based on MAD: For any pixel (x, y) in the enhanced image I2, select an initial window U1(x, y) of size w1×w1, and calculate the mean absolute deviation (MAD) of the pixel grayscale values ​​within the initial window U1(x, y). Record this MAD value as m1. If m1 is less than a given MAD threshold T1, set the window size w1 to w1+1, and recalculate the m1 value until the m1 value is greater than or equal to the MAD threshold T1, or the w1 value is greater than or equal to the maximum window value w. max Substitute the current window size w1 into the following Brenner-2D focus evaluation function:

[0014] I m (x,y)=|I2(x,y)-I2(x+w1-1,y)|+|I2(x,y)-I2(x,y+w1-1)|

[0015] In the formula: x—pixel horizontal coordinate; y—pixel vertical coordinate;

[0016] The focus evaluation value M(x,y) of pixel (x,y) on the enhanced image I2 is obtained; the above method is applied to each pixel on the enhanced image I2 to obtain the focus evaluation image I. m .

[0017] Step 3 specifically involves: evaluating the focus of image I. m Adaptive window mean filtering based on MAD-based focus evaluation values: for focus evaluation image I m For any pixel (x, y) on the screen, select an initial window U2(x, y) of size w2×w2, and calculate the mean absolute deviation (MAD) of the pixel focus evaluation values ​​within the initial window U2(x, y), and record this MAD value as m2. If m2 is greater than a given MAD threshold T2, then set the window size w2 to w2+1, and recalculate the m2 value until the m2 value is greater than or equal to the MAD threshold T2, or the w2 value is greater than or equal to the maximum window size w. max Using the window size w2 at this time, mean filtering is performed to obtain the focused evaluation image I. m The filtered result M′(x,y) of the focus evaluation value of pixel (x,y) on the image; the focus evaluation image I m Each pixel on the image is processed using the method described above to obtain the filtered focus evaluation image I. m ′.

[0018] The beneficial effects of this invention are:

[0019] This invention achieves image filtering that does not affect focus evaluation by using guided filtering and high-contrast preservation image preprocessing methods. Furthermore, it effectively reduces the impact of microstructure scale issues on focus evaluation by employing MAD-based Brenner-2D adaptive focus evaluation and MAD-based adaptive window mean filtering of focus evaluation values. Specifically:

[0020] (1) Based on the framework of three-dimensional measurement of microscopic zoom, this invention does not require the addition of other equipment in the microscopic vision measurement system, has strong adaptability, and can well fit the microscopic vision measurement system under different usage conditions.

[0021] (2) Since the present invention uses guided filtering and high contrast retention as image preprocessing methods, compared with ordinary filtering, it can effectively retain the features of the focused area of ​​the microscopic image, filter out noise in non-aggregated areas, and achieve image preprocessing that has no impact on focus evaluation.

[0022] (3) By dynamically selecting the focusing evaluation scale, the present invention effectively reduces the impact of microstructure scale problems on focusing evaluation, reduces erroneous focusing evaluation values, and is beneficial to high-precision three-dimensional measurement. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the microstructure scale problem in digital images; Figure (a) includes only one pixel; Figure (b) includes multiple pixels.

[0024] Figure 2 This is a flowchart of the present invention.

[0025] Figure 3 This is a schematic diagram of the high contrast retention method used in this invention.

[0026] Figure 4 Figure 1 shows the microscopic focusing image processing process involved in the embodiments of the present invention; wherein (a) is the original microscopic image I0 of the air film aperture acquired by the CCD camera; (b) is the filtered image I1 after being filtered by the guided filter; (c) is the enhanced image I2 obtained after high contrast preservation processing; (d) is the focusing evaluation image obtained directly using Brenner-2D; and (e) is the focusing evaluation image I1 obtained by Brenner-2D adaptive focusing evaluation based on MAD. m Figure (f) shows the filtered focus evaluation image I after MAD-based adaptive window filtering of the focus evaluation value. m ′. Detailed Implementation

[0027] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0028] Reference Figure 2An adaptive focusing evaluation method for microscopic optical measurement of film pores in turbine blades, based on a microscopic vision measurement system, includes the following steps:

[0029] Step 1: Perform guided filtering on the original microscopic image I0 to obtain the filtered image I1; then perform high-contrast preservation on the filtered image I1 to obtain the enhanced image I2;

[0030] Specifically, the method for applying guided filtering to the original microscopic image I0 is as follows:

[0031] Guided filters solve the edge-preserving filtering problem through linear optimization. By assuming that the pixel grayscale values ​​before and after filtering have a linear relationship within the filtering window, they maintain the relativity of the gradient, thereby achieving edge preservation.

[0032]

[0033] Where: ω—calculation window; a—first-order linear coefficient; b—constant term; x—pixel x-coordinate; y—pixel y-coordinate; i—pixel x-coordinate within the window; j—pixel y-coordinate within the window;

[0034] For different filtering windows, the filtering strength can be controlled by changing the values ​​of a and b. When the window is on the edge, a larger a is used to weaken smoothing; when the window is not on the edge, a smaller a is used to enhance smoothing. Since the filtered image should minimize the variance compared to the unfiltered image, a penalty function method is used to optimize a and b. The cost function is set as follows:

[0035]

[0036] In the formula; cost — cost; ∈ — relative to a (x，y) The penalty coefficient, a (x，y) The larger the size, the greater the punishment.

[0037] Taking the partial derivatives of the above equation with respect to a and b respectively, we get:

[0038]

[0039]

[0040] In the formula; cost — cost; ∈ — penalty coefficient;

[0041] Setting the above two equations to 0, we find the extreme values, solve the system of equations, and obtain:

[0042]

[0043]

[0044] In the formula: |ω| — number of pixels in the window; μ — mean of the guide image within the window; σ — standard deviation of pixel grayscale within the window;

[0045] For the same pixel, it will appear multiple times in the filtering window during the filtering process, so it is necessary to take the average value, that is:

[0046]

[0047] Specifically, such as Figure 3 As shown, the method for high-contrast-preserving image enhancement of the filtered image is as follows: First, Gaussian filtering is applied to the filtered image I1 to obtain the background image I. b Then compare the filtered image I1 with the background image I b Subtraction yields the detailed image I. f Then, the detailed image I f Multiply by the enhancement coefficient n, and finally add to the filtered image I1 to obtain the enhanced image I2;

[0048] Step 2, evaluate the enhanced image I2 using MAD-based Brenner-2D adaptive focus: For any pixel (x, y) in the enhanced image I2, select an initial window U1(x, y) of size w1×w1:

[0049] U1(x,y)={I2(i,j)|0≤(ix)<w1∩0≤(jy)<w1}

[0050] In the formula: U1—focus evaluation window, w1—focus rating scale, and w1 is generally taken as 3 in the initial case;

[0051] Calculate the mean absolute deviation (MAD) of pixel grayscale values ​​within the initial window U1(x,y), and denote this MAD value as m1:

[0052]

[0053] Where: m1—MAD value;

[0054] If m1 is less than the given MAD threshold T1, then set the window size w1 to w1+1, and recalculate the value of m1 until the value of m1 is greater than or equal to the MAD threshold T1, or the value of w1 is greater than or equal to the maximum window size w. max Substitute the current window size w1 into the following Brenner-2D focus evaluation function to obtain the focus evaluation value of point (x, y):

[0055] I m(x,y)=|I2(x,y)-I2(x+w1-1,y)|+|I2(x,y)-I2(x,y+w1-1)|

[0056] The focus evaluation value M(x,y) of pixel (x,y) on the enhanced image I2 is obtained; by applying the above method to each pixel on the enhanced image I2, the focus evaluation image I can be obtained. m ;

[0057] Step 3, focus on the evaluation image I m Adaptive window mean filtering based on MAD-based focus evaluation values: for focus evaluation image I m For any pixel (x, y) on the graph, select an initial window U2(x, y) of size (2×w²+1)×(2×w²+1):

[0058] U2(x, y)={I m (i,j)||ix|≤w2∩|jy|≤w2}

[0059] Where: U2 — filter window;

[0060] The mean absolute deviation (MAD) of pixel focus evaluation values ​​within the initial window U2(x,y) is calculated and denoted as m2.

[0061]

[0062] Where: m2 — MAD value;

[0063] If m2 is greater than the given MAD threshold T2, then set the window size w2 to w2+1, and recalculate the value of m2 until the value of m2 is greater than or equal to the MAD threshold T2, or the value of w2 is greater than or equal to the maximum window size w. max Use the current window size w2 for mean filtering:

[0064]

[0065] Obtain the focused evaluation image I m Focus evaluation value filtering result I of pixel (x, y) on m ′(x, y); for focusing evaluation image I m By applying the above method to each pixel, the filtered focus evaluation image I can be obtained. m ′.

[0066] Figure 4Figure 1 shows the microscopic focusing image processing procedure involved in this embodiment; where (a) is the original microscopic image I0 of the air film aperture acquired by the CCD camera; (b) is the filtered image I1 after being filtered by the guided filter; (c) is the enhanced image I2 obtained after high contrast preservation processing; (d) is the focusing evaluation image obtained directly using Brenner-2D; and (e) is the focusing evaluation image I1 obtained by Brenner-2D adaptive focusing evaluation based on MAD. m Figure (f) shows the filtered focus evaluation image I after MAD-based adaptive window filtering of the focus evaluation value. m The comparison shows that the original microscopic image I0 has a lot of noise in the high-focal region; the noise in the high-focal region of the filtered image I1 is removed; the filtered image I1 has large-sized microstructures; the focus evaluation of the microstructures in the focus evaluation image obtained directly using Brenner-2D fails; the focus evaluation image I... m The focus evaluation of the internal microstructure was improved, but the focus evaluation changed drastically; filtered focus evaluation image I m The filtered focused evaluation value is relatively smooth.

[0067] The above embodiments are merely one implementation of the present invention. Any modifications and improvements made without departing from the concept of the present invention should fall within the protection scope of the present invention.

Claims

1. An adaptive focus evaluation method for micro-optical measurement of turbine blade film hole, based on a micro-vision measurement system, characterized in that, The method comprises the following steps: Step 1, pre-process the original microscopic image I0 by using a microscopic image pre-processing algorithm combining guided filtering with high-contrast reservation to obtain an enhanced image I2; Step 2. Using the MAD-based Brenner-2D adaptive focus evaluation on the enhanced image I2, the focus evaluation image I m ; in particular: Brenner-2D adaptive focus evaluation based on MAD is used for the enhanced image I2: for any pixel (x, y) on the enhanced image I2, an initial window U1(x, y) with size w1 x w1 is selected, and the mean absolute deviation (MAD) of the pixel gray values in the initial window U1(x, y) is calculated, and the MAD value is recorded as m1, if m1 is less than a given MAD threshold T1, then the window size w1 is equal to w1 + 1, and the m1 value is calculated again, until the m1 value is greater than or equal to the MAD threshold T1, or the w1 value is greater than or equal to the maximum window size w max ; the window size w1 at this time is substituted into the following Brenner-2D focus evaluation function: Brenner-2D adaptive focus evaluation based on MAD is used for the enhanced image I2: for any pixel (x, y) on the enhanced image I2, an initial window U1(x, y) with size w1 x w1 is selected, and the mean absolute deviation (MAD) of the pixel gray values in the initial window U1(x, y) is calculated, and the MAD value is recorded as m1, if m1 is less than a given MAD threshold T1, then the window size w1 is equal to w1 + 1, and the m1 value is calculated again, until the m1 value is greater than or equal to the MAD threshold T1, or the w1 value is greater than or equal to the maximum window size w max ; the window size w1 at this time is substituted into the following Brenner-2D focus evaluation function: Brenner-2D adaptive focus evaluation based on MAD is used for the enhanced image I2 I m (x,y) = |I(x,y) - I(x + w1 - 1,y) | + |I(x,y) - I(x,y + w1 - 1) | where: x - horizontal pixel coordinate; y - vertical pixel coordinate; obtaining a focus evaluation value M(x, y) of the pixel (x, y) on the enhanced image I2; using the above method for each pixel on the enhanced image I2, i.e. obtaining a focus evaluation image I m ; Step 3. Evaluating the image I for focus m Adaptive window mean filtering of the focus evaluation image I using the MAD-based focus merit function results in a filtered focus evaluation image I m ′.

2. The self-adaptive focusing evaluation method for micro-optical measurement of turbine blade film hole according to claim 1, characterized in that, The step 1 is specifically: using a penalty function method to optimize the linear parameters of the guided filter for the original microscopic image I0, performing the guided filter to obtain a filtered image I1; then performing high-contrast reservation image enhancement on the filtered image I1: first performing Gaussian filtering to obtain a background image I b Subtracting the background image I b from the filtered image I1 to obtain a detail image I f Multiplying the detail image I f by an enhancement coefficient n, and then adding the filtered image I1 to obtain an enhanced image I2.

3. The self-adaptive focusing evaluation method for micro-optical measurement of turbine blade film hole according to claim 1, characterized in that, The step 3 is specifically: calculating the focus evaluation image I m Adaptive window mean filtering using the MAD-based focus evaluation value: selecting an initial window U2(x, y) with a size of w2xw2 for any pixel (x, y) on the focus evaluation image I m , and calculating the mean absolute deviation (MAD) of the focus evaluation values of the pixels within the initial window U2(x, y), and recording the MAD value as m2. If m2 is less than a given MAD threshold T2, then the window size w2 is equal to w2+1, and the m2 value is calculated again until the m2 value is greater than or equal to the MAD threshold T2, or the w2 value is greater than or equal to the maximum window size w max . Mean filtering is performed using the window size w2 at this time to obtain the focus evaluation value filtering result I m ′(x, y) of the pixel (x, y) on the focus evaluation image I m . On the focused evaluation image I m Each pixel on the focused evaluation image I m is filtered using the above method, i.e. the filtered focused evaluation image I

Images