Defect detection method for high-reflection battery aluminum shell

Abstract

The invention discloses a defect detection method for a high-reflection battery aluminum shell, which comprises the following steps of: generating sine stripes required by a light source by using computer coding, respectively projecting the sine stripes in x and y directions to the surface of the aluminum shell to be detected through an LCD (Liquid Crystal Display) screen, and acquiring 16 deformed stripe patterns by using a CCD (Charge Coupled Device) camera; drawing a gray value curve of the stripe pattern to obtain a minimum enclosing rectangle of an area where the surface of the to-be-detected aluminum shell is located, and extracting all the surface of the to-be-detected aluminum shell; performing phase extraction and unwrapping on the extracted deformed stripes on the surface of the to-be-detected aluminum shell by using a multi-frequency heterodyne method to obtain a matrix containing phase information in x and y directions; according to the positions of pixel points which are suddenly changed in the phase information, a reflectivity model is used for estimating defect information of the surface of the aluminum shell to be detected, a reflectivity defect graph is generated, gradient information is calculated according to the phase information, and a curvature graph is obtained according to the gradient information and the curvature relation to obtain a clear defect picture.

Description

Technical Field

[0001] The invention relates to a defect detection method for a highly reflective battery aluminum shell. Background Art

[0002] In the surface defect detection of highly reflective objects, due to the strong specular reflection characteristics of their surfaces, traditional image detection methods often have difficulty in effectively identifying defects. The high gloss surface of such objects will cause strong reflection of light, especially under uneven lighting conditions, which easily produces complex light spots and specular reflection effects, causing the brightness information in the imaging to change dramatically, covering or confusing tiny surface defects, resulting in limited detection precision and accuracy. Conventional image processing algorithms are difficult to accurately extract defect information of such reflective surfaces, and the detection effect is easily affected by noise and unstable.

[0003] Battery aluminum shells face stringent quality requirements during manufacturing and subsequent applications. Due to the ductility, machinability and corrosion resistance of aluminum, it is easier to form during the manufacturing process, but it is also more prone to surface defects such as scratches, indentations, microcracks, or bubbles and poor welding during welding. These defects may affect the sealing and structural integrity of the aluminum shell, thereby endangering the safety and service life of the battery.

[0004] To address these challenges, people have gradually invented advanced methods such as structured light projection, line scanning laser, polarized light imaging, and phase measurement deflectometry. Among them, phase measurement deflectometry can accurately restore the three-dimensional surface morphology information of an object by projecting phase-shifted fringes, combined with phase extraction and gradient calculation. Using multi-channel image detection technology, light of different wavelengths, phases or polarization directions will form different reflection images on the surface of an object, so that some defects hidden in a single-channel image can be presented in the comparison of multi-channel images, thereby overcoming the reflection interference problem caused by single-channel imaging. In view of this, the present invention provides a defect detection method for highly reflective battery aluminum shells. Summary of the invention

[0005] The present invention aims to solve the problems existing in the prior art and provide a defect detection method for a highly reflective battery aluminum shell.

[0006] The technical solutions adopted in the present invention are:

[0007] A method for detecting defects in a highly reflective battery aluminum shell, comprising:

[0008] S1: Use computer coding to generate the sinusoidal stripes required by the light source, project the sinusoidal stripes in the x and y directions onto the surface of the aluminum shell to be tested through the LCD display, and use the CCD camera to collect a total of 16 stripe patterns that produce deformation;

[0009] S2: Draw a gray value curve of the stripe pattern to obtain the minimum circumscribed rectangle of the area where the aluminum shell surface to be tested is located, and extract the entire surface of the aluminum shell to be tested;

[0010] S3: Use the multi-frequency heterodyne method to extract and unfold the phase of the stripes extracted from the deformed surface of the aluminum shell to be tested, and obtain a matrix containing the phase information in the x and y directions;

[0011] S4: According to the position of the pixel point where the mutation occurs in the phase information, the reflectivity model is used to estimate the defect information of the aluminum shell surface to be tested and generate a reflectivity defect map. At the same time, the gradient information is calculated from the phase information, and the curvature map is obtained according to the relationship between the gradient information and the curvature to obtain a clear defect image.

[0012] Furthermore, in S1, four sinusoidal fringe patterns are drawn by using the four-step phase shift method and the time phase unwrapping method, including two sinusoidal fringe patterns with different frequencies in the x direction and two sinusoidal fringe patterns with the same frequency as the x direction in the y direction;

[0013] Add 0 to each sinusoidal fringe pattern. π, Four phase shifts are used to obtain a total of 16 fringe patterns. The 16 fringe patterns are photographed by a CCD camera and decoded by a computer to obtain the phase information of each fringe pattern. The mathematical expression for the fringe pattern generation is:

[0014]

[0015] Among them, (x, y) is the three-dimensional coordinate of a pixel point on the imaging surface of the CCD camera, I is the fringe light intensity, A is the background light intensity, B is the amplitude modulation intensity, is the phase distribution containing the surface gradient information of the object to be solved, δ(t) is the additional phase shift that changes with time, also known as the phase shift, and its value is

[0016] Furthermore, before the LCD display screen projects the sinusoidal stripes, the gamma value of the LCD display screen is pre-calibrated to obtain the brightness transfer function from the CCD camera to the LCD display screen, and then the input value is pre-corrected before the stripe image is generated, or the distorted stripes captured by the CCD camera are corrected.

[0017] Furthermore, four sinusoidal fringe patterns with frequencies of 1 and 32 are projected in both the x and y directions, and the phase shift steps are 0, π,

[0018] Furthermore, the phase extraction and unfolding of the deformed fringes include the following steps:

[0019] (1) All fringe patterns collected by the CCD camera are converted to double type; the initial fringe patterns are averaged and converted into uniform high grayscale value areas, and the average grayscale values ​​in the x and y directions of the projection are calculated respectively to obtain the corresponding grayscale curve. The mutation part of the grayscale curve represents the edge position, and then the first-order differential of the row and column grayscale value curves is performed to obtain the row and column first-order differential curves. From the two curves, the minimum row and column range of the aluminum shell surface to be tested can be obtained, and the area of ​​interest can be selected;

[0020] (2) Calculate the phase of the fringe pattern in the x and y directions, and only select the value of the aluminum shell surface to be tested. The phase calculation formula is as follows:

[0021]

[0022] in, is the phase, I is the light intensity measurement point, and four different light intensity values ​​are obtained by the four-step phase shift method;

[0023] (3) After calculating the phase information in the x and y directions respectively, the dual-frequency heterodyne method is used to extract the phase information. Perform phase unwrapping to obtain absolute phase information, which includes information about phase mutations containing defects.

[0024] Furthermore, the phase unwrapping information is used to combine the reflectivity and gradient information to obtain a picture containing defects, including the following steps:

[0025] (1) For the phase information in the expanded x and y directions, firstly weight and merge them into the phase information matrix of the entire area, thereby obtaining the phase map to initially observe the defect information;

[0026] (2) Determine the grid spacing, that is, the physical distance between each pixel, and calculate the first-order gradient of the phase information in the x and y directions to obtain the surface slope information in each direction. The gradients along the x and y directions are expressed as:

[0027]

[0028] The curvature reflecting the curvature of the surface is obtained by calculating the second-order derivative of the phase gradient. The specific calculation is as follows:

[0029]

[0030] The overall curvature information image is obtained by taking the square root of the sum of the squares of the curvatures in each direction, which is used to analyze the distribution of the local curvature degree of the object surface;

[0031] (3) According to the reflectivity formula, the reflected light intensity and the incident light intensity are used to construct a reflectivity model related to the gradient information:

[0032]

[0033] Among them, I r (x, y) and I 0 (x, y) are the incident light and reflected light intensities, respectively. Represents gradient information, which is used to correct the reflectivity model.

[0034] Furthermore, the defects of the battery aluminum shell are detected including:

[0035] The outer surface has concave spots, surface scratches, missing materials, random patterns, and blackening;

[0036] Black spots and pits on the bottom;

[0037] Cracks and scratches on edges;

[0038] Internal steps are aluminum wire, embossed, blackened.

[0039] The present invention has the following beneficial effects:

[0040] For the highly reflective aluminum shell surface of solar cells, subtle defects that are difficult to detect with a single channel can be found by comparing images from multiple different channels. The accuracy of the system meets the accuracy requirements of the defect standard. At the same time, the system can continue to participate in the acquisition of three-dimensional surface information after calibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the steps of phase measurement deflectometry.

[0042] Figure 2 It is a hardware device system diagram.

[0043] Figure 3 This is the schematic diagram of the phase measurement deflectometry.

[0044] Figure 4(a) is the absolute phase diagram of the aluminum shell surface.

[0045] Figure 4(b) is a graph of the curvature of the aluminum shell surface.

[0046] Figure 4(c) is a diagram showing the surface defects of the aluminum shell. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings.

[0048] Figure 2 The hardware system of the present invention includes a CCD camera, an LCD display, a host computer and a test platform; its imaging principle is as follows Figure 3As shown in the figure, when the surface of the object to be tested coincides with the reference plane, a point A on the LCD screen is reflected by point O on the reference plane and enters the CCD camera. If the smooth surface to be tested at point O is deflected at an angle of , then the light entering the CCD camera at the same position will come from point B on the LCD screen, and the angle between the light OA and OB is 2, which means that at each point within the detection range, for the angle between the smooth surface to be tested and the reference plane, there is a similar angle on the LCD display. Figure 3 The change from point A to point B corresponds to this. Since a standard sinusoidal grating is projected on the LCD display, the change in this angle can be converted into a change in the spatial phase of the grating.

[0049] like Figure 1 , the detection method of the present invention comprises:

[0050] S1: Draw the fringe pattern and optimize and improve it, and take pictures;

[0051] The four-step phase shift method and the time phase unwrapping method were used to draw four sinusoidal fringe patterns, including two sinusoidal fringe patterns with different frequencies in the x direction and two sinusoidal fringe patterns with the same frequency in the y direction;

[0052] Add 0 to each image. π, There are four phase shifts, and finally there are 16 fringe patterns. The 16 fringe patterns are photographed by a CCD camera, and the phase information of each fringe pattern is decoded by a computer. The mathematical expression for the fringe pattern generation is:

[0053]

[0054] Among them, (x, y) is the three-dimensional coordinate of a pixel point on the imaging surface of the CCD camera, I is the fringe light intensity, A is the background light intensity, B is the amplitude modulation intensity, is the phase distribution containing the surface gradient information of the object to be solved, δ(t) is the additional phase shift that changes with time, also known as the phase shift, and its value is

[0055] The accuracy of the stripes directly affects the accuracy and effect of the final imaging. There are many kinds of errors in the stripes projected by the light source, such as defocus, overexposure and gamma distortion. The present invention achieves an ideal relative value by adjusting the camera aperture and focal length.

[0056] In order to eliminate the gamma distortion of the system, it pre-calibrates the gamma value of the system, obtains the brightness transfer function from the CCD camera to the LCD display, and then pre-corrects the input value before generating the stripe image or corrects the distorted stripes captured by the CCD camera.

[0057] The present invention uses a projection method combining a four-step phase shift method and a time phase unwrapping method, so the fringe expression that needs to be drawn is as follows:

[0058] I 1 (m,n)=I'(m,n)+I″(m,n)cos[θ(m,n)]

[0059]

[0060] I 3 (m,n)=I'(m,n)+I″(m,n)cos[θ(m,n)+π]

[0061]

[0062] The same coordinate point (m, n), I i (m, n) is set as the gray value of the i-th image at that point, I′(m, n) is the background value of the fringe light intensity, I″(m, n) is the modulation intensity, and θ(m, n) is the phase to be determined.

[0063] According to the four-step phase shift method and time phase unwrapping method, a display screen is used to sequentially project 4 horizontal and 4 vertical sinusoidal stripes with a frequency of 1, and the phase shift steps are 0, π, Similarly, 4 horizontal and vertical sinusoidal stripes with a frequency of 32 are projected again, for a total of 16 images.

[0064] S2: Selecting a region of interest from the captured image.

[0065] The present invention selects the target area for detection (ie the surface of the aluminum shell to be detected) by a grayscale projection method.

[0066] First, the image is processed with global grayscale mean value to identify the high grayscale value area, so as to preliminarily determine the part of interest in the image. Next, the average grayscale value of the fringe pattern in the x and y directions of the image is calculated respectively to obtain the grayscale distribution curves along the row and column directions. On these curves, the area where the grayscale value changes suddenly usually corresponds to the edge position of the object.

[0067] On this basis, the grayscale distribution curves in the row and column directions are subjected to first-order differential operations to obtain the row and column gradient curves that reflect the grayscale change rate. The peak part of the first-order differential corresponds to the position where the grayscale value changes significantly, that is, the position of the edge area. By analyzing the peak points of these two gradient curves, the significant position range of the edge can be identified, thereby determining the minimum boundary range of the target area in the row and column directions.

[0068] Finally, the minimum rectangle of the target area, i.e., the region of interest, is extracted using the identified row and column minimum boundary range. This method can effectively locate high grayscale areas and clarify their boundaries. It is suitable for automatically extracting target areas from complex backgrounds, especially in terms of accurate boundary positioning.

[0069] S3: extract and unfold the phase of the captured fringe pattern;

[0070] First, convert the image to double type, calculate the phase of the fringe pattern in the x and y directions, and select only the values ​​in the target area. The phase calculation formula is as follows:

[0071]

[0072] in is the phase, I is the light intensity measurement point, and these are four different light intensity values ​​obtained by the four-step phase shift method.

[0073] The present invention uses a multi-frequency heterodyne method to extract the absolute phase from the long and short wavelength phase expansion. For each pixel, the difference between the long-period phase multiplied by the multiple k and the short-period phase is calculated and divided by 2π to obtain an integer multiple of the period number. Then, this integer multiple of the period number is used to expand the short-period phase to obtain the absolute phase (Figure 4 (a) is an absolute phase diagram of the aluminum shell surface). The formula is as follows:

[0074]

[0075] in To unwrap the phase, is the short wavelength phase, is the long wavelength phase, and k is the multiple relationship between long and short wavelengths.

[0076] S4: Use the phase unwrapping information, combined with the reflectivity and gradient information to obtain a picture containing defects.

[0077] The total phase unwrapping amount is obtained by expanding and square-adding the phase information of x and y.

[0078] Furthermore, the total phase unwrapping amount is gray-scale modulated to obtain a preliminary defect map.

[0079] Furthermore, a grid spacing is set, that is, the physical distance between each pixel. The first-order gradient of the phase unwrapping information in the x and y directions is calculated to obtain the surface slope information in each direction. The gradients along the x and y directions are expressed as:

[0080]

[0081] Calculate the second derivative (curvature) of the phase gradient. The curvature is calculated as follows:

[0082]

[0083] Finally, the overall curvature information image is obtained by taking the square root of the sum of the squares of the curvatures in each direction, which is used to analyze the distribution of the local curvature degree of the object surface (Figure 4(b) is the curvature diagram of the aluminum shell surface). The formula is as follows:

[0084]

[0085] At the same time, based on the phase information and gradient information, a reflectivity model is estimated based on the reflectivity to analyze the surface defects of the object. A reflectivity model related to the gradient information is constructed using the reflected light intensity and the incident light intensity:

[0086]

[0087] Among them I r (x, y) and I 0 (x, y) are the incident light and reflected light intensities respectively. Represents gradient information, which is used to correct the model.

[0088] Furthermore, a median filter is used on the reflectance image to replace the value of each pixel with the median value of the pixels in its neighborhood, thereby reducing the local extreme value.

[0089] Furthermore, a multi-channel comparison is performed on the generated phase image, curvature image, and reflectivity, and the defects therein are marked for subsequent defect detection (Figure 4(c) is a display of defects on the aluminum shell surface).

[0090] The present invention can clearly image highly reflective aluminum shells and other mirror-like reflective objects based on phase measurement deflectometry, and screen out defects with ideal accuracy through multi-channel image comparison.

[0091] The above description is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be regarded as within the protection scope of the present invention.

Claims

1. A method for detecting defects in highly reflective battery aluminum shells, characterized in that: include S1: Use computer coding to generate the sinusoidal stripes required by the light source, project the sinusoidal stripes in the x and y directions onto the surface of the aluminum shell to be tested through the LCD display, and use the CCD camera to collect a total of 16 stripe patterns that produce deformation; S2: Draw a gray value curve of the stripe pattern to obtain the minimum circumscribed rectangle of the area where the aluminum shell surface to be tested is located, and extract the entire surface of the aluminum shell to be tested; S3: Use the multi-frequency heterodyne method to extract and unfold the phase of the stripes extracted from the deformed surface of the aluminum shell to be tested, and obtain a matrix containing the phase information in the x and y directions; S4: According to the position of the pixel point where the mutation occurs in the phase information, the reflectivity model is used to estimate the defect information of the aluminum shell surface to be tested and generate a reflectivity defect map. At the same time, the gradient information is calculated from the phase information, and the curvature map is obtained according to the relationship between the gradient information and the curvature to obtain a clear defect image.

2. The defect detection method for a highly reflective battery aluminum shell according to claim 1, characterized in that: In S1, four sinusoidal fringe patterns are drawn by using the four-step phase shift method and the time phase unwrapping method, including two sinusoidal fringe patterns with different frequencies in the x direction and two sinusoidal fringe patterns with the same frequency as the x direction in the y direction; Add 0 to each sinusoidal fringe pattern. π, Four phase shifts are used to obtain a total of 16 fringe patterns. The 16 fringe patterns are photographed by a CCD camera and decoded by a computer to obtain the phase information of each fringe pattern. The mathematical expression for the fringe pattern generation is: Among them, (x, y) is the three-dimensional coordinate of a pixel point on the imaging surface of the CCD camera, I is the fringe light intensity, A is the background light intensity, B is the amplitude modulation intensity, is the phase distribution containing the surface gradient information of the object to be solved, δ(t) is the additional phase shift that changes with time, also known as the phase shift, and its value is 3. The defect detection method for a highly reflective battery aluminum shell according to claim 2, characterized in that: Before the LCD display screen projects sinusoidal stripes, the gamma value of the LCD display screen is pre-calibrated to obtain the brightness transfer function from the CCD camera to the LCD display screen. Then, the input value is pre-corrected before the stripe image is generated, or the distorted stripes captured by the CCD camera are corrected.

4. The defect detection method for highly reflective battery aluminum shell according to claim 2, characterized in that: Four sinusoidal fringe patterns with frequencies of 1 and 32 are projected in both the x-direction and the y-direction, and the phase shift steps are 0, π, 5. The defect detection method for highly reflective battery aluminum shell according to claim 1, characterized in that: Phase extraction and unfolding of deformed fringes include the following steps: (1) All fringe patterns collected by the CCD camera are converted to double type; the initial fringe patterns are averaged and converted into uniform high grayscale value areas, and the average grayscale values ​​in the x and y directions of the projection are calculated respectively to obtain the corresponding grayscale curve. The mutation part of the grayscale curve represents the edge position, and then the first-order differential of the row and column grayscale value curves is performed to obtain the row and column first-order differential curves. From the two curves, the minimum row and column range of the aluminum shell surface to be tested can be obtained, and the area of ​​interest can be selected; (2) Calculate the phase of the fringe pattern in the x and y directions, and only select the value of the aluminum shell surface to be tested. The phase calculation formula is as follows: in, is the phase, I is the light intensity measurement point, and four different light intensity values ​​are obtained by the four-step phase shift method; (3) After calculating the phase information in the x and y directions respectively, the dual-frequency heterodyne method is used to extract the phase information. Phase unwrapping is performed to obtain absolute phase information, which includes information about phase mutations containing defects.

6. The defect detection method for highly reflective battery aluminum shell according to claim 5, characterized in that: Using phase unwrapping information, combined with reflectivity and gradient information, to obtain images containing defects includes the following steps: (1) For the phase information in the x and y directions, the weighted information is first combined to form a phase information matrix for the entire region, thereby obtaining a phase map to initially observe defect information; (2) Determine the grid spacing, that is, the physical distance between each pixel, and calculate the first-order gradient of the phase information in the x and y directions to obtain the surface slope information in each direction. The gradients along the x and y directions are expressed as: The curvature reflecting the curvature of the surface is obtained by calculating the second-order derivative of the phase gradient. The specific calculation is as follows: The overall curvature information image is obtained by taking the square root of the sum of the squares of the curvatures in each direction, which is used to analyze the distribution of the local curvature degree of the object surface; (3) According to the reflectivity formula, the reflected light intensity and the incident light intensity are used to construct a reflectivity model related to the gradient information: Among them, I r (x, y) and I0(x, y) are the incident light and reflected light intensities, respectively. Represents gradient information, which is used to correct the reflectivity model.

Images