A rapid detection method for metal strip surface defects based on filter fusion

The Fourier transform and a Qualcommended bandpass filter process the surface image of the metal strip, which solves the problem of time-consuming existing detection systems and realizes the rapid and efficient detection of surface defects of the metal strip.

CN114862843BActive Publication Date: 2024-11-01NORTHEASTERN UNIV CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210634180.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-06
Publication Date
2024-11-01
Estimated Expiration
2042-06-06

AI Technical Summary

Technical Problem

The existing metal strip surface defect detection system is difficult to meet the needs of high-precision and rapid detection. The traditional system processing method takes a long time and is difficult to meet the requirements of downstream product manufacturing companies to improve detection accuracy and speed.

Method used

The surface image of the metal strip is processed through the Fourier transform and the high-pass bandpass fusion filter to suppress low-frequency energy and release medium- and high-frequency energy, so as to achieve rapid detection of surface defects of the metal strip.

Benefits of technology

Through a filtering process, different types of metal strip surface defects can be detected, which shortens the detection steps and time and improves the detection efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114862843B_ABST
    Figure CN114862843B_ABST
Patent Text Reader

Abstract

The present application relates to a method for rapid detection of metal strip surface defects based on filter fusion, the method comprising: obtaining an original image of the metal strip surface; converting the original image of the metal strip surface into a Fourier spectrum diagram through Fourier transform, and centering the Fourier spectrum diagram; constructing a high-pass-bandpass fusion filter based on a high-pass filter and a bandpass filter; enhancing the centralized Fourier spectrum image through a high-pass-bandpass fusion filter; obtaining a corresponding spatial domain image through inverse Fourier transform based on the enhanced Fourier spectrum image, and obtaining metal strip surface defect information based on the spatial domain image. For defects of different characteristics on the surface of metal strips, the present application, based on the above method, can obtain all metal strip surface defect information through only one filtering process, thereby shortening the processing steps, saving processing time, and improving processing efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the technical field of metal surface defect detection, and specifically relates to a method for rapid detection of metal strip surface defects based on filter fusion. Background Art

[0002] The processing and production of metal strips such as copper strips, aluminum strips, and steel strips often have the characteristics of large format sizes and fast production processes. In the process of processing metal strips, due to many factors such as the execution process, production equipment, and surrounding environment, the strips will have a variety of surface defects, such as scratches, burns, pits, bulges, etc. Different defect types have different imaging characteristics.

[0003] Surface defects are important indicators for evaluating product quality, and necessary inspections need to be carried out in the production process to control product quality. At present, the non-contact inspection method of machine vision is mainly used for the inspection of strip surface defect quality. Its working principle is: an industrial camera is used with an auxiliary light source to collect the surface image of the strip in real time, and then the image data is passed to the inspection system for inspection and processing. The recognition results are fed back to the user through the human-computer interface, statistical reports, etc. Surface defect detection is mainly based on graphics detection methods: the abnormal parts of the strip surface are extracted by means of noise reduction, filtering, edge extraction, contour detection, etc., and then the defects are classified and processed through the classification model in machine learning, so as to realize the detection of defects. Due to the different characteristics of strip surface defects, different processing steps and related setting parameters need to be formulated for each defect when extracting abnormal parts.

[0004] At present, as downstream terminal product manufacturers have increasingly higher requirements for strip surface quality, surface inspection systems are required to have higher inspection accuracy and faster inspection speed. Traditional inspection systems can improve accuracy by replacing industrial cameras, but high-precision images have high resolution and large amounts of information, so image processing takes longer, and traditional system processing methods are difficult to meet new inspection requirements.

[0005] In view of this, the present application provides a method for rapid detection of metal strip surface defects based on filter fusion. For different types of defects in metal strips, only one filtering process needs to be performed to detect defects with different characteristics on the strip surface, thereby improving processing efficiency. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides a method for rapid detection of surface defects of metal strips based on filter fusion.

[0008] (II) Technical solution

[0009] In order to achieve the above objectives, this application adopts the following technical solutions:

[0010] In a first aspect, the present application provides a method for rapid detection of metal strip surface defects based on filter fusion, comprising:

[0011] S1, obtaining the original image of the metal strip surface;

[0012] S2. Converting the original image of the metal strip surface into a Fourier spectrum through Fourier transform, and centering the Fourier spectrum;

[0013] The Fourier spectrum diagram is a two-dimensional matrix of the same size as the original image of the metal strip surface;

[0014] S3, constructing a high-pass-bandpass fusion filter based on the high-pass filter and the band-pass filter;

[0015] The high-pass filter and the band-pass filter are: a two-dimensional matrix of the same size as the Fourier spectrum obtained by setting the flux in the interval [0,1]; wherein the flux is a value corresponding to a coordinate position in the filter;

[0016] The high-pass-band-pass fusion filter is: a two-dimensional matrix with the same size as the Fourier spectrum diagram constructed according to the fluxes corresponding to the high-pass filter and the band-pass filter;

[0017] The high-pass-bandpass fusion filter is used to suppress low-frequency energy in the Fourier spectrum diagram based on the flux corresponding to the high-pass-bandpass fusion filter, and release intermediate frequency and high frequency energy in the Fourier spectrum diagram;

[0018] S4, enhancing the centralized Fourier spectrum image through a high-pass-bandpass fusion filter;

[0019] S5. Based on the enhanced Fourier spectrum image, obtain a corresponding spatial domain image through inverse Fourier transform, and obtain surface defect information of the metal strip based on the spatial domain image.

[0020] Optionally, the converting of the original image of the metal strip surface into a Fourier spectrum diagram by Fourier transform in S2 includes:

[0021] The original image of the metal strip surface is subjected to Fourier transform, specifically:

[0022]

[0023] Wherein, f(x,y) represents the M×N matrix corresponding to the original image of the metal strip surface, where x=0,1,2,…,M-1 and y=0,1,2,…,N-1; F(u,v) represents the Fourier transform of f(x,y).

[0024] Optionally, the step S2 converts the original image of the metal strip surface into a Fourier spectrum diagram by Fourier transform, and centralizes the Fourier spectrum diagram, including:

[0025] S2-1, taking the center point of the Fourier spectrum graph as the boundary, dividing the Fourier spectrum graph into four regions: upper left, upper right, lower left and lower right;

[0026] S2-2. Swap the upper left area with the lower right area in the Fourier spectrum graph, and swap the upper right area with the lower left area in the Fourier spectrum graph to obtain a centralized Fourier spectrum graph.

[0027] Optionally, the step of constructing a high-pass-band-pass fusion filter based on the high-pass filter and the band-pass filter in S3 includes:

[0028] S3-1, obtaining a second high-pass filter based on the first high-pass filter;

[0029] The first high-pass filter is an ideal high-pass filter, and the ideal high-pass filter is a high-pass filter whose flux is only 0 or 1;

[0030] The ideal high-pass filter is used to completely suppress low-frequency energy and completely release high-frequency energy according to the flux corresponding to the ideal high-pass filter;

[0031] S3-2, superimposing the second high-pass filter and the Butterworth band-pass filter to construct a high-pass-band-pass fusion filter.

[0032] Optionally, the obtaining of the second high-pass filter based on the first high-pass filter in S3-1 includes:

[0033] The first high-pass filter is expressed as:

[0034]

[0035] Where u corresponds to the abscissa of the Fourier spectrum graph, v corresponds to the ordinate of the Fourier spectrum graph, dis(u,v) represents the distance from the coordinate point (u,v) in the Fourier spectrum graph to the center point of the Fourier spectrum graph, and D0 represents the radius of the high-pass filtered image.

[0036] Optionally, the obtaining of the second high-pass filter based on the first high-pass filter in S3-1 includes:

[0037] The radius of the Fourier spectrum image passing through the first high-pass filter is set to 10 pixels, and the first high-pass filter is subjected to channel attenuation processing to obtain a second high-pass filter, which is expressed as:

[0038]

[0039] Where C is the sum of the long side and the wide side of the original image on the metal strip surface.

[0040] Optionally, the step of superimposing the second high-pass filter and the Butterworth band-pass filter in S3-2 to construct a high-pass-band-pass fusion filter includes:

[0041] The Butterworth bandpass filter is expressed as:

[0042]

[0043] Where W represents the passband width and n represents the edge gradient smoothness.

[0044] Optionally, the step of superimposing the second high-pass filter and the Butterworth band-pass filter in S3-2 to construct a high-pass-band-pass fusion filter includes:

[0045] The values ​​corresponding to the coordinate positions in the second high-pass filter and the Butterworth band-pass filter are used as fluxes; based on the fluxes of the second high-pass filter and the Butterworth band-pass filter, a high-pass-band-pass fusion filter is constructed;

[0046] Among them, for the value corresponding to the coordinate position, 0 means that the energy is blocked, 1 means that all the energy passes, and the interval between (0, 1) means that the energy partially passes.

[0047] Optionally, the step of acquiring a corresponding spatial domain image by inverse Fourier transform based on the enhanced Fourier spectrum image in S5, and acquiring the surface defect information of the metal strip based on the spatial domain image, includes:

[0048] S5-1, performing inverse Fourier transform on the enhanced Fourier spectrum image to obtain a spatial domain image, specifically:

[0049]

[0050] Where f'(x,y) represents the inverse Fourier transform of F(u,v);

[0051] S5-2, performing binarization processing on the obtained spatial domain image to obtain a binary-spatial domain image;

[0052] The binary-spatial domain image is a spatial domain image represented by values ​​0 and 255;

[0053] S5-3. Obtaining surface defect information of metal strip based on binary-spatial domain images.

[0054] Optionally, the step of acquiring the surface defect information of the metal strip based on the Fourier spectrum image in S5 includes:

[0055] The surface defect information of the metal strip includes roller marks, scratches, bulges, black stripes and holes.

[0056] (III) Beneficial effects

[0057] The present application constructs a high-pass-bandpass fusion filter by fusing a high-pass filter and a band-pass filter, thereby achieving: obtaining all types of surface defect information of metal strips through only one filtering process; further, shortening the processing steps of defect detection, saving the processing time of defect detection, and improving the processing efficiency of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The present application is described with the aid of the following drawings:

[0059] Figure 1 It is a schematic diagram of the process of the rapid detection method of metal strip surface defects based on filter fusion;

[0060] Figure 2 It is a schematic diagram of the process of low-frequency centering of Fourier spectrum;

[0061] Figure 3-1 is the original two-dimensional Fourier spectrum;

[0062] Figure 3-2 It is a two-dimensional Fourier spectrum after centralization;

[0063] Figure 4 It is a schematic diagram of an ideal high-pass filter;

[0064] Figure 5-1 This is a schematic diagram of the effect obtained based on an ideal high-pass filter;

[0065] Figure 5-2 This is a schematic diagram of the binarization effect obtained based on an ideal high-pass filter;

[0066] Figure 6 is a schematic diagram of a bandpass filter;

[0067] Figure 7-1 A schematic diagram of the effect obtained based on the bandpass filter;

[0068] Figure 7-2 This is a schematic diagram of the binarization effect obtained based on the bandpass filter;

[0069] Figure 8 It is a schematic diagram of a high-pass-bandpass fusion filter;

[0070] Figure 9-1 This is a schematic diagram of the effect obtained based on the high-pass-bandpass fusion filter;

[0071] Figure 9-2 This is a schematic diagram of the binarization effect obtained based on the high-pass-bandpass fusion filter;

[0072] Fig.10 Schematic diagram of defect information included on the surface of metal strip. DETAILED DESCRIPTION

[0073] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below in conjunction with the accompanying drawings through specific implementation methods. It is understood that the specific embodiments described below are only used to explain the relevant inventions, rather than to limit the invention. It should also be noted that the embodiments and features in the embodiments of this application can be combined with each other in the absence of conflict; for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0074] Embodiment 1 provides a method for rapid detection of metal strip surface defects based on filter fusion, such as Figure 1 As shown, the specific steps are as follows:

[0075] S1. Obtaining an original image of the metal strip surface.

[0076] S2. Convert the original image of the metal strip surface into a Fourier spectrum diagram through Fourier transform, and center the Fourier spectrum diagram.

[0077] In this embodiment, the Fourier spectrum diagram is a two-dimensional matrix of the same size as the original image of the metal strip surface.

[0078] S3. Based on the high-pass filter and the band-pass filter, a high-pass-band-pass fusion filter is constructed.

[0079] In this embodiment, the high-pass filter and the band-pass filter are: a two-dimensional matrix of the same size as the Fourier spectrum diagram is obtained by setting the flux in the interval [0, 1]; wherein the flux is a value corresponding to the coordinate position in the filter.

[0080] In this embodiment, the high-pass-bandpass fusion filter is a two-dimensional matrix of the same size as the Fourier spectrum diagram constructed according to the fluxes corresponding to the high-pass filter and the band-pass filter.

[0081] In this embodiment, the high-pass-bandpass fusion filter is used to suppress low-frequency energy in the Fourier spectrum diagram based on the flux corresponding to the high-pass-bandpass fusion filter, and release intermediate-frequency and high-frequency energy in the Fourier spectrum diagram.

[0082] S4. Enhance the centralized Fourier spectrum image through a high-pass-bandpass fusion filter.

[0083] S5. Based on the enhanced Fourier spectrum image, obtain a corresponding spatial domain image through inverse Fourier transform, and obtain surface defect information of the metal strip based on the spatial domain image.

[0084] In this embodiment, the surface defect information of the metal strip includes roller marks, scratches, bulges, black stripes and holes. It should be noted that: it includes but is not limited to the above five types of defect information.

[0085] Based on the metal strip surface defect rapid detection method based on filter fusion in Example 1, by fusing the high-pass filter and the band-pass filter to construct a high-pass-bandpass fusion filter, it can be achieved that: all types of metal strip surface defect information can be obtained through only one filtering process; the processing steps of defect detection are further shortened, the processing time of defect detection is saved, and the processing efficiency of defect detection is improved.

[0086] Embodiment 2 provides a method for rapid detection of metal strip surface defects based on filter fusion, and the specific method steps are as follows:

[0087] S1. Obtaining an original image of the metal strip surface.

[0088] Based on the above step S1, an industrial camera may be selected to obtain an original image of the surface of the metal strip.

[0089] S2. Convert the original image of the metal strip surface into a Fourier spectrum diagram through Fourier transform, and center the Fourier spectrum diagram.

[0090] Based on the above step S2, for the Fourier transform process, it should be noted that:

[0091] The idea of ​​Fourier transform is that any function can be very accurately approximated to the sum of infinite sine functions and cosine functions. The Fourier transform process of the original image of the metal strip surface can be implemented based on the following formula (1), specifically:

[0092]

[0093] In formula (1), f(x, y) represents the M×N matrix corresponding to the original image of the metal strip surface, where x=0, 1, 2, ..., M-1 and y=0, 1, 2, ..., N-1; F(u, v) represents the Fourier transform of f(x, y).

[0094] Based on the above step S2, for the Fourier spectrum diagram, it should be noted that:

[0095] The Fourier spectrum is a two-dimensional matrix with the same size as the original image of the metal strip surface. It describes the image information through parameters such as frequency and amplitude. The frequency is divided into high frequency and low frequency. The high frequency part describes the edge information in the image, and the low frequency part describes the basic outline of the image. The amplitude is the value of any point in the Fourier spectrum. The amplitude describes the energy of the original image at any point. Specifically, the higher the amplitude, the more information at that frequency is in the original image.

[0096] Based on the above step S2, it should be noted that for the centering of the Fourier spectrum graph:

[0097] Taking the center point of the Fourier spectrum graph as the boundary, the Fourier spectrum graph is evenly divided into four areas: upper left, upper right, lower left and lower right; however, for the Fourier spectrum graph, it is usually the case that "the low-frequency energy part is distributed in the upper left, upper right, lower left and lower right positions of the Fourier spectrum graph, while the high-frequency energy part is in the center of the Fourier spectrum graph". As a result, the Fourier spectrum graph cannot be directly filtered using high-pass filters and band-pass filters, and the low-frequency energy part needs to be moved to the center of the Fourier spectrum graph. This process is the centralization process of the Fourier spectrum graph.

[0098] For the centering of the Fourier spectrum graph, an optional processing process is: divide the Fourier spectrum graph into two parts in the horizontal direction and in the vertical direction, and obtain the four regions of upper left, upper right, lower left and lower right mentioned above, and mark the four regions of upper left, upper right, lower left and lower right as A, B, C, D respectively, as shown in Figure 2 As shown, the positions of area A and area D are swapped, and the positions of area B and area C are swapped. Based on the adjustment of the above area positions, the low-frequency part of the energy is moved to the center of the Fourier spectrum diagram, thereby completing the centering of the Fourier spectrum diagram.

[0099] In order to more clearly describe the centering process of the Fourier spectrum graph, the following is a further example of the centering of the two-dimensional Fourier spectrum graph. Figure 3-1 and Figure 3-2 As shown, Figure 3-1 is the original two-dimensional Fourier spectrum, Figure 3-2is a two-dimensional Fourier spectrum after centralization, u represents the vertical direction, and v represents the horizontal direction; Figure 3-1 The darker part in the middle is the high-frequency area of ​​the Fourier spectrum, and the lighter part around the edges is the low-frequency area of ​​the Fourier spectrum. Figure 3-1 After centering the original two-dimensional Fourier spectrum shown in Figure 3-2 The two-dimensional Fourier spectrum shown in Figure 3-2 It can be seen that in the two-dimensional Fourier spectrum after centralization, the lighter part in the middle is the low-frequency area of ​​the Fourier spectrum, and the darker part at the edges is the high-frequency area of ​​the Fourier spectrum. In other words, the low-frequency area of ​​the original Fourier spectrum has been transferred to the center of the Fourier spectrum.

[0100] S3. Based on the high-pass filter and the band-pass filter, a high-pass-band-pass fusion filter is constructed.

[0101] Based on the above step S3, it should be noted that, for the construction of the high-pass-bandpass fusion filter, an optional processing process is:

[0102] S3-1. Obtain a second high-pass filter based on the first high-pass filter.

[0103] Based on the aforementioned step S3-1, for the first high-pass filter, it should be noted that:

[0104] Since the background of the metal strip surface is single, except for a very small amount of noise and defects, the grayscale values ​​of the image are concentrated in a smaller range. In the Fourier spectrum diagram, the frequency of the background area is basically at the center point. Therefore, in this embodiment, the ideal high-pass filter can be used as the first high-pass filter. The image obtained based on the ideal high-pass filter is characterized in that the image edge gradient is steep and there is no smooth transition area. Therefore, the ideal high-pass filter is suitable for segmenting the edges of defect information such as scratches and black spots on the surface of the metal strip.

[0105] Furthermore, the ideal high-pass filter is a high-pass filter whose flux is only 0 or 1. Specifically, the ideal high-pass filter can be expressed by the following formula (2):

[0106]

[0107] In formula (2), u corresponds to the horizontal coordinate of the Fourier spectrum graph, v corresponds to the vertical coordinate of the Fourier spectrum graph, dis(u,v) represents the distance from the coordinate point (u,v) in the Fourier spectrum graph to the center point of the Fourier spectrum graph, and D0 represents the radius of the high-pass filtered image.

[0108] In this embodiment, the ideal high-pass filter is used to completely suppress low-frequency energy and completely release high-frequency energy according to the flux corresponding to the ideal high-pass filter.

[0109] Based on the above formula (2), it should be noted that: for the point (u, v) with a radius greater than D0 in the Fourier spectrum graph, H(u, v) takes a value of 1; for the point (u, v) with a radius less than D0 in the Fourier spectrum graph, H(u, v) takes a value of 0; in the filtering process, the point with a value of 0 in H(u, v) is directly filtered out, thereby retaining only the high-frequency part of the Fourier spectrum graph;

[0110] In addition, it should be noted that for the strip surface with a relatively simple background, the low-frequency area is quite concentrated, and D0 can be set to a constant of 1. Specifically, an ideal high-pass filter with D0 of 1 is as follows: Figure 4 As shown, the effect diagram obtained based on the ideal high-pass filter processing with D0 as 1 is as follows Figure 5-1 As shown; due to Figure 5-1 The effect picture shown is a grayscale picture. In order to further extract the surface defects of the metal sheet, Figure 5-1 Perform binary threshold processing to obtain Figure 5-2 The schematic diagram of the binarization effect obtained based on the ideal high-pass filter is shown;

[0111] Through the aforementioned Figure 5-2 It can be seen that the black spot defect parts in the figure can be identified and extracted. Even if D0 is set to the minimum unit value 1, pits and bulges are still not detected, and the filter needs to be further optimized and adjusted.

[0112] Based on the aforementioned step S3-1, an optional processing procedure for obtaining the second high-pass filter is as follows:

[0113] The radius of the Fourier spectrum image passing through the first high-pass filter is set to 10 pixels, and the first high-pass filter is subjected to channel attenuation processing to obtain a second high-pass filter. The second high-pass filter can be expressed by the following formula (3):

[0114]

[0115] In formula (3), C is the sum of the long side and the wide side of the original image on the metal strip surface.

[0116] In this embodiment, the Fourier spectrum obtained by Fourier transform is consistent with the size of the original image of the metal strip surface, and may be a rectangular image with a long side and a wide side.

[0117] S3-2, superimposing the second high-pass filter and the Butterworth band-pass filter to construct a high-pass-band-pass fusion filter.

[0118] Based on the aforementioned step S3-2, it should be noted that the curve in the frequency band corresponding to the Butterworth bandpass filter is maximally flat and has no ripples, which is conducive to extracting the defect contour. Furthermore, in this embodiment, the Butterworth bandpass filter is used to process defects with relatively large scales, fuzzy edges and low frequencies.

[0119] Based on the aforementioned step S3-2, it should be noted that the Butterworth bandpass filter can be expressed by the following formula (4):

[0120]

[0121] In formula (4), W represents the passband width, and n represents the edge gradient smoothness.

[0122] Based on the above formula (4), it should be noted that the larger the value of n is, the steeper the gradient is. When the value of n is the minimum value 1, the gradient is the smoothest. Since the edges of defects at medium and low frequencies are relatively blurred, smoothing processing helps to extract the defect contour. Therefore, in this embodiment, n is taken as 1.

[0123] Based on the above step S3-2, when D0 is 10 and W is 8, the Butterworth bandpass filter is as follows: Figure 6 As shown in the figure, and the effect diagram obtained based on Butterworth bandpass filter processing is as follows Figure 7-1 As shown; due to Figure 7-1 The effect picture shown is a grayscale picture. In order to further extract the surface defects of the metal sheet, Figure 7-1 Perform binary threshold processing to obtain Figure 7-2 Schematic diagram of the binarization effect obtained based on the bandpass filter.

[0124] Based on the aforementioned Figure 7-2 It can be seen that defects such as pits and bulges can be detected based on the Butterworth bandpass filter, but most of the small black bar defect information is still unrecognizable, and the filter needs to be further adjusted and improved.

[0125] Based on the above step S3-2, for the construction of the high-pass-bandpass fusion filter, it should be noted that:

[0126] In this embodiment, the numerical values ​​corresponding to the coordinate positions in the second high-pass filter and the Butterworth band-pass filter are used as fluxes; based on the fluxes of the second high-pass filter and the Butterworth band-pass filter, a high-pass-band-pass fusion filter is constructed; specifically, based on the fluxes corresponding to the coordinate positions in the second high-pass filter and the Butterworth band-pass filter, the fluxes of the two are multiplied by each other in pairs, and based on the new fluxes obtained by the multiplication process, a high-pass-band-pass fusion filter is constructed.

[0127] In this embodiment, for the numerical value corresponding to the coordinate position, 0 indicates that the energy is blocked, 1 indicates that all the energy passes, and the interval between (0, 1) indicates that part of the energy passes.

[0128] Based on the above step S3-2, the high-pass-bandpass fusion filter constructed is as follows: Figure 8 As shown, and the effect diagram of the high-pass-bandpass fusion filter is as follows Figure 9-1 As shown; due to Figure 9-1 The effect picture shown is a grayscale picture. In order to further extract the surface defects of the metal sheet, Figure 9-1 Perform binary threshold processing to obtain Figure 9-2 The diagram shows a schematic diagram of the binarization effect obtained based on a high-pass-bandpass fusion filter.

[0129] Based on the above Figure 9-2 , we can know that the pits, bulges and small black spots in the image are all detected.

[0130] S4. Enhance the centralized Fourier spectrum image through a high-pass-bandpass fusion filter.

[0131] S5. Based on the enhanced Fourier spectrum image, obtain a corresponding spatial domain image through inverse Fourier transform, and obtain surface defect information of the metal strip based on the spatial domain image.

[0132] In this embodiment, the surface defect information of the metal strip is obtained, such as Fig.10 As shown, the defect information on the surface of the metal strip includes pits, small black spots and bulges.

[0133] Based on the aforementioned step S5, it is necessary to explain the process of acquiring the spatial domain image as follows:

[0134] S5-1, performing inverse Fourier transform on the enhanced Fourier spectrum image to obtain a spatial domain image, specifically:

[0135]

[0136] In formula (5), f'(x, y) represents the inverse Fourier transform of F(u, v).

[0137] S5-2. Binarize the obtained spatial domain image to obtain a binary-spatial domain image.

[0138] In this embodiment, the binary-spatial domain image is a spatial domain image represented by numerical values ​​0 and 255.

[0139] S5-3. Obtaining surface defect information of metal strip based on binary-spatial domain images.

[0140] Based on the metal strip surface defect rapid detection method based on filter fusion in Example 2, by fusing the high-pass filter and the band-pass filter to construct a high-pass-band-pass fusion filter, it can be achieved that: all types of metal strip surface defect information can be obtained through only one filtering process; the processing steps of defect detection are further shortened, the processing time of defect detection is saved, and the processing efficiency of defect detection is improved.

[0141] It should be noted that in the claims, any figure marks between brackets should not be understood as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "one" or "an" preceding a component does not exclude the presence of multiple such components. In addition, it should be noted that in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" and the like refers to the specific features, structures, materials or characteristics described in conjunction with the embodiment or example included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0142] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments after knowing the basic creative concept. Therefore, the claims should be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0143] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention should also include these modifications and variations.

Claims

1. A rapid detection method for metal strip surface defects based on filter fusion, It is characterized in that The rapid detection method comprises: S1, obtaining an original image of the metal strip surface; S2. Converting the original image of the metal strip surface into a Fourier spectrum through Fourier transform, and centering the Fourier spectrum; The Fourier spectrum diagram is a two-dimensional matrix of the same size as the original image of the metal strip surface; S3, constructing a high-pass-bandpass fusion filter based on the high-pass filter and the band-pass filter; The high-pass filter and the band-pass filter are: a two-dimensional matrix of the same size as the Fourier spectrum obtained by setting the flux in the interval [0,1]; wherein the flux is a value corresponding to a coordinate position in the filter; The high-pass-band-pass fusion filter is: a two-dimensional matrix with the same size as the Fourier spectrum diagram constructed according to the fluxes corresponding to the high-pass filter and the band-pass filter; The high-pass-bandpass fusion filter is used to suppress low-frequency energy in the Fourier spectrum diagram based on the flux corresponding to the high-pass-bandpass fusion filter, and release intermediate frequency and high frequency energy in the Fourier spectrum diagram; S4, enhancing the centralized Fourier spectrum image through a high-pass-bandpass fusion filter; S5. Based on the enhanced Fourier spectrum image, obtain a corresponding spatial domain image through inverse Fourier transform, and obtain surface defect information of the metal strip based on the spatial domain image; The high-pass-band-pass fusion filter is constructed based on the high-pass filter and the band-pass filter in S3, including: S3-1, obtaining a second high-pass filter based on the first high-pass filter; The first high-pass filter is an ideal high-pass filter, and the ideal high-pass filter is a high-pass filter whose flux is only 0 or 1; The ideal high-pass filter is used to completely suppress low-frequency energy and completely release high-frequency energy according to the flux corresponding to the ideal high-pass filter; S3-2, superimposing the second high-pass filter and the Butterworth band-pass filter to construct a high-pass-band-pass fusion filter; The step of obtaining the second high-pass filter based on the first high-pass filter in S3-1 includes: The first high-pass filter is expressed as: Where u corresponds to the horizontal coordinate of the Fourier spectrum, v corresponds to the vertical coordinate of the Fourier spectrum, dis(u,v) represents the distance from the coordinate point (u,v) in the Fourier spectrum to the center point of the Fourier spectrum, and D 0 Represents the radius of the high-pass filtered image; The step of obtaining the second high-pass filter based on the first high-pass filter in S3-1 includes: The radius of the Fourier spectrum image passing through the first high-pass filter is set to 10 pixels, and the first high-pass filter is subjected to channel attenuation processing to obtain a second high-pass filter, which is expressed as: Where C is the sum of the long side and the wide side of the original image on the metal strip surface.

2. The rapid detection method according to claim 1, It is characterized in that The step S2 converts the original image of the metal strip surface into a Fourier spectrum diagram through Fourier transform, including: The original image of the metal strip surface is subjected to Fourier transform, specifically: Wherein, f(x,y) represents the M×N matrix corresponding to the original image of the metal strip surface, where x=0,1,2,…,M-1 and y=0,1,2,…,N-1; F(u,v) represents the Fourier transform of f(x,y).

3. The rapid detection method according to claim 1, It is characterized in that The step S2 converts the original image of the metal strip surface into a Fourier spectrum diagram through Fourier transformation, and centralizes the Fourier spectrum diagram, including: S2-1, taking the center point of the Fourier spectrum graph as the boundary, dividing the Fourier spectrum graph into four regions: upper left, upper right, lower left and lower right; S2-2. Swap the upper left area with the lower right area in the Fourier spectrum graph, and swap the upper right area with the lower left area in the Fourier spectrum graph to obtain a centralized Fourier spectrum graph.

4. The rapid detection method according to claim 1, It is characterized in that The step of superimposing the second high-pass filter and the Butterworth band-pass filter to construct a high-pass-band-pass fusion filter in S3-2 includes: The Butterworth bandpass filter is expressed as: Where W represents the passband width and n represents the edge gradient smoothness.

5. The rapid detection method according to claim 1, It is characterized in that The step of superimposing the second high-pass filter and the Butterworth band-pass filter to construct a high-pass-band-pass fusion filter in S3-2 includes: The values ​​corresponding to the coordinate positions in the second high-pass filter and the Butterworth band-pass filter are used as fluxes; based on the fluxes of the second high-pass filter and the Butterworth band-pass filter, a high-pass-band-pass fusion filter is constructed; Among them, for the value corresponding to the coordinate position, 0 means that the energy is blocked, 1 means that all the energy passes, and the interval between (0, 1) means that the energy partially passes.

6. The rapid detection method according to claim 1, It is characterized in that The step of acquiring a corresponding spatial domain image based on the enhanced Fourier spectrum image by inverse Fourier transform in S5, and acquiring surface defect information of the metal strip based on the spatial domain image, includes: S5-1, performing inverse Fourier transform on the enhanced Fourier spectrum image to obtain a spatial domain image, specifically: Where f'(x,y) represents the inverse Fourier transform of F(u,v); S5-2, performing binarization processing on the obtained spatial domain image to obtain a binary-spatial domain image; The binary-spatial domain image is a spatial domain image represented by values ​​0 and 255; S5-3. Obtaining surface defect information of metal strip based on binary-spatial domain images.

7. The rapid detection method according to claim 1, It is characterized in that The step of acquiring the surface defect information of the metal strip based on the Fourier spectrum image in S5 includes: The surface defect information of the metal strip includes roller marks, scratches, bulges, black stripes and holes.