Image method intelligent detection method and system for welding circuit board

Abstract

The invention discloses an image method intelligent detection method and system for a welding circuit board. The method comprises the steps that CCD images of electronic elements of the circuit board are collected; performing image enhancement identification on the CCD image to obtain an enhanced feature image; based on the enhanced feature image, establishing a standard library query set model, and based on the standard library query set model, identifying a to-be-detected electronic component; the method comprises the following steps: acquiring an infrared image of a to-be-detected electronic component, separating a reflection component from an incident component by adopting a Retinex method, performing high-frequency-band sharpening processing on an image representing inherent properties of a circuit board, performing image enhancement on an image of a foreground of the electronic component, and identifying defect characteristics of the circuit board; and based on the identified circuit board defect characteristics, establishing a unique mapping relation by establishing a comparison table of an element database and the acquired circuit board elements, and performing corresponding inspection on element types, shape sizes, connecting wire diameters, via hole types, and pad types and sizes.

Description

Technical Field

[0001] The invention belongs to the field of welding circuit board detection, and in particular relates to an image-based intelligent detection method and system for welding circuit boards. Background Art

[0002] Welded circuit boards are commonly used in electronic engineering and automated manufacturing. The quality inspection of a welded circuit board is very important, especially for some discrete electronic components. The inspection of the circuit board lacks certain rules, and even the PCB diagram cannot be generated from the schematic diagram, so it cannot be automatically inspected. If the welding is not firm, or there is false welding, leakage welding will cause major failures. Too thin line width will cause heat, and too wide line width will cause waste. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention proposes an image-based intelligent detection method and system for welded circuit boards, which realizes micro-short circuit detection and quantitative evaluation of energy storage lithium-ion battery packs based on an unsupervised clustering algorithm and the maximum charging voltage difference between adjacent cycles without establishing a battery model or setting a diagnostic threshold.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] An image-based intelligent detection method for a welded circuit board comprises the following steps:

[0006] Collect CCD images of electronic components on circuit boards;

[0007] Perform image enhancement and recognition on the CCD image to obtain an enhanced feature image;

[0008] Based on the enhanced feature image, a standard library query set model is established, and based on the standard library query set model, the electronic components to be tested are identified;

[0009] By collecting infrared images of the electronic components to be tested, the Retinex method is used to separate the reflected component from the incident component to obtain an image representing the inherent properties of the circuit board. The image representing the inherent properties of the circuit board is sharpened in the high frequency band to obtain an image of the foreground of the electronic components. The image of the foreground of the electronic components is enhanced to identify the defect characteristics of the circuit board.

[0010] Based on the identified circuit board defect features, a unique mapping relationship is established by building a component database and a comparison table of the collected circuit board components, and corresponding checks are performed on component type, shape and size, connection wire diameter, via type, pad type and size.

[0011] Preferably, the method for collecting CCD images of electronic components of a PCB circuit board includes:

[0012] The circuit board to be inspected is placed on the stage, and the vertical motor automatically adjusts the focal length so that the clarity of the electronic components on the circuit board meets the preset requirements. The CCD camera is driven by the horizontal motor X direction from the upper left corner of the circuit board to scan along the horizontal X axis direction, and driven by the horizontal motor Y direction to scan along the horizontal Y axis direction. The scanning path starts from the origin (0,0) and scans line by line in the preset arrow direction until it ends at the end point (X,Y), and is repeated three times.

[0013] Preferably, the method of performing image enhancement recognition on the CCD image to obtain the enhanced feature image includes:

[0014] The CCD image is stored in picture format, represented as P 1 (x1,y1), where P 1 (x1, y1) represents the image of electronic components on the soldered circuit board scanned by the CCD camera;

[0015] The electronic components PCB library components are retrieved from the drawing application software to form library components in the standard PCB mode, and the shapes and sizes of PCB library components are established. The standard PCB component database is dynamically established. s (x s ,y s ), where P s (x s ,y s ) represents all PCB component information in the drawing software;

[0016] Based on the standard PCB component database P s (x s ,y s ), P 1 (x1, y1) performs feature clustering data processing to obtain each feature P 1 (x'1,y'1);

[0017] For each characteristic P 1 The various eigenvectors of (x'1, y'1) are restored to obtain the corresponding eigenimages;

[0018] The corresponding feature image is denoised and enhanced to obtain the enhanced feature image.

[0019] Preferably, each characteristic is classified into 1 Methods for restoring various eigenvectors of (x'1, y'1) and obtaining corresponding eigenimages include:

[0020] For each characteristic P 1 (x'1, y'1), use convolution to extract and establish a feature vector group, the expression is:

[0021] X={X1,X2,···X n )}

[0022] Among them, X is the feature vector of the feature area, X1,X2,···X n is the eigenvector of each item;

[0023] Restore various eigenvectors and find the corresponding eigenimages. The expression is:

[0024] P={P1,P2,···P n )}

[0025] Among them, P is the feature map corresponding to the feature vector, P1,P2,···P n is the feature image corresponding to each sub-item feature vector, and the expression is: in, Represents all components on the circuit board. Indicates the image information of each component.

[0026] Preferably, the method of denoising and enhancing the corresponding feature image to obtain the enhanced feature image includes:

[0027] right Perform wavelet transform to extract the high-frequency part. The expression is:

[0028]

[0029] Among them, D H is the high-frequency part of the image, which contains detail information. It means that the wavelet transform takes the high frequency part;

[0030] Sharpen the high-frequency image and get D' H ;

[0031] right Perform wavelet transform to extract the low-frequency part. The expression is:

[0032]

[0033] Among them, D L is the low-frequency part of the feature image, which contains background information. It means that the wavelet transform takes the low-frequency part;

[0034] Right D L Processing, Canny operator product, the expression is:

[0035] D' L =D L *g

[0036] Where D' is the processed low-frequency image, and g is the Canny operator;

[0037] Reconstruct the processed results, the expression is:

[0038] D,=D, L +D, H

[0039] Then D' is the enhanced feature image, which is expressed as D' in the form of a two-dimensional array. i (x,y).

[0040] Preferably, the query search set model is established, and the expression is:

[0041]

[0042] Among them, O tr Represents the comparison search set, D' i (x, y) represents the enhanced component images, and a comparison search is performed in the standard component library by collecting the images. K1 represents the number of searches.

[0043] Preferably, the method for identifying the electronic component to be tested based on the standard library query set model includes:

[0044] Based on the standard library query set model, a rough search is first performed to determine the type of electronic components, and then a detailed search is performed to determine the size and specifications of the electronic components to obtain the final query results; the expression is:

[0045]

[0046] Among them, O c Indicates the result of a rough search, O s,m Indicates the final search result, M indicates the identifier that meets the conditions, and D' i (,) represents the set of components on the circuit board. Represents a standard library component, k c ,k' c Indicates the number of searches, O size Indicates the output component size, O type Indicates the output component type;

[0047] The accuracy of the electronic components on the tested circuit board is determined based on the standard library component parameters, that is, D'(,) and P s (x s ,y s ) is within the preset error range, the identification is correct. Specifically, the minimum difference method is used for judgment, and the formula is:

[0048] Δ=|D'(,)-P s (,)|

[0049] When the error Δ<10 -2 When , the image automatic recognition program ends;

[0050] The vias, pads and connections of the identified electronic components are inspected to determine whether they meet the requirements. If they meet the requirements, they are judged to be correct, otherwise they are wrong.

[0051] Preferably, the method of collecting infrared images of the electronic components to be tested, separating the reflected component from the incident component by using the Retinex method, and obtaining an image representing the inherent properties of the circuit board; performing high-frequency sharpening processing on the image representing the inherent properties of the circuit board to obtain an image of the foreground of the electronic components; and performing image enhancement on the image of the foreground of the electronic components to identify the defect characteristics of the circuit board includes:

[0052] The collected infrared image is represented by P H (x, y), using the Retinex image enhancement method, the component representing the inherent properties of the circuit board is obtained, and the expression is:

[0053] R(x,y)=lnP H (x,y)-ln[G(x,y)*P H (x,y)]

[0054] Where R(x,y) is the image of the inherent properties of the circuit board; G(x,y) is the Gaussian function;

[0055] The image of the inherent properties of the circuit board is decomposed by wavelet into high-frequency components and low-frequency components. The expression is:

[0056]

[0057] in, Represents the high frequency part, Represents the low frequency part, represents a high-frequency image, Represents a low-frequency image;

[0058] Will After sharpening, it is reconstructed with the original low-frequency part to form an enhanced thermal image, and then synthesized with the incident component to form the final thermal image P' H , then P' H is the enhanced image;

[0059] The enhanced component infrared image is used to inspect various defect features of the circuit board.

[0060] Preferably, the method of using the enhanced component infrared image to inspect various defect characteristics of the circuit board includes:

[0061] When the circuit board is powered on, current will be generated in the pads, vias, and wires, and the fault point will be identified based on the pixel value of the enhanced component infrared image.

[0062] Extract the gray value of the background color as the threshold value θ, and judge with the qualified band α. The expression is:

[0063] λ<θ-α indicates a cold solder joint or a false solder joint;

[0064] λ>θ+α means the wire is too thin and the solder joint is too small;

[0065] θ-α<λ<θ+α indicates passing;

[0066] Among them, θ is the threshold value, which represents the pixel value of the background, α represents the qualified band, and λ represents the pixel value of the component.

[0067] The present invention also provides an image-based intelligent detection system for welding circuit boards, comprising: an acquisition module, an enhancement module, a component inspection module, a defect recognition module and a type detection module;

[0068] The acquisition module is used to acquire CCD images of electronic components of the circuit board;

[0069] The enhancement module is used to perform image enhancement recognition on the CCD image to obtain an enhanced feature image;

[0070] The component inspection module is used to establish a standard library query set model based on the enhanced feature image, and identify the electronic component to be tested based on the standard library query set model;

[0071] The defect recognition module is used to collect infrared images of the electronic components to be tested, separate the reflected component from the incident component by using the Retinex method, obtain an image representing the inherent properties of the circuit board, perform high-frequency sharpening processing on the image representing the inherent properties of the circuit board, obtain an image of the foreground of the electronic components, perform image enhancement on the image of the foreground of the electronic components, and identify the defect characteristics of the circuit board;

[0072] The type detection module is used to establish a unique mapping relationship based on the identified circuit board defect characteristics by establishing a component database and a comparison table of the collected circuit board components, and perform corresponding checks on component types, shape dimensions, connection wire diameters, via types, pad types and dimensions.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] The present invention proposes an intelligent welding circuit board detection method and system. Through the coordinated movement of four-axis servo motors, a CCD camera obtains visible light images of electronic components and obtains initial information of the electronic components on the circuit board. Then, an infrared camera obtains infrared images of the electronic components and obtains dynamic information of the electronic components. Image enhancement and recognition are performed on the collected CCD images. High-frequency components and low-frequency components are decomposed through wavelet transform, and high-frequency components are sharpened to highlight component features, separate components and background, and achieve the purpose of recognition.

[0075] By collecting infrared images of components and using the Retinex method, the reflected component and the incident component are separated, the image representing the inherent properties of the circuit board is sharpened in the high frequency band, the boundary features are enhanced, the components are separated from the background, and the recognition features of the image representing the foreground of the components are enhanced, that is, the features of pads that are too small and wires that are too thin are identified, and the features of cold solder joints and false solder joints are identified.

[0076] By establishing a comparison table between the component database and the collected circuit board components, a unique mapping relationship is established to perform corresponding checks on component type, shape size, connection wire diameter, via type, pad type and size, as shown in Table 2. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0078] Figure 1 A schematic diagram of a CCD scanning direction according to an embodiment of the present invention;

[0079] Figure 2 A schematic diagram of the scanning direction of an infrared camera according to an embodiment of the present invention;

[0080] Figure 3 The present invention is a flowchart of an intelligent detection method for welding circuit boards using an image method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0081] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0082] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0083] Embodiment 1

[0084] like Figure 3 As shown, the present invention provides an image-based intelligent detection method for a welded circuit board, comprising the following steps:

[0085] Collect CCD images of electronic components on circuit boards;

[0086] Perform image enhancement and recognition on the CCD image to obtain an enhanced feature image;

[0087] Based on the enhanced feature image, a standard library query set model is established, and based on the standard library query set model, the electronic components to be tested are identified;

[0088] By collecting infrared images of the electronic components to be tested, the Retinex method is used to separate the reflected component from the incident component to obtain an image representing the inherent properties of the circuit board. The image representing the inherent properties of the circuit board is sharpened in the high frequency band to obtain an image of the foreground of the electronic components. The image of the foreground of the electronic components is enhanced to identify the defect characteristics of the circuit board.

[0089] Based on the identified circuit board defect features, a unique mapping relationship is established by establishing a comparison table between the component database and the collected circuit board components, and corresponding checks are performed on component type, shape size, connection wire diameter, via type, pad type and size, as shown in Table 2.

[0090] In this embodiment, the method for collecting CCD images of electronic components of a PCB circuit board includes:

[0091] like Figure 1 As shown, the circuit board to be inspected is placed on the stage, and the vertical motor automatically adjusts the focal length so that the clarity of the electronic components on the circuit board meets the preset requirements. The CCD camera is driven by the horizontal motor X direction from the upper left corner of the circuit board to scan along the horizontal X axis direction, and driven by the horizontal motor Y direction to scan along the horizontal Y axis direction. The scanning path starts from the origin (0,0) and scans line by line in the preset arrow direction until it ends at the end point (X,Y), and is repeated three times.

[0092] In this embodiment, the method of performing image enhancement recognition on the CCD image to obtain the enhanced feature image includes:

[0093] The CCD image is stored in picture format, represented as P 1 (x1,y1), where P 1(x1, y1) represents the image of electronic components on the soldered circuit board scanned by the CCD camera;

[0094] The electronic components PCB library components are retrieved from the drawing application software to form library components in the standard PCB mode, and the shapes and sizes of PCB library components are established. The standard PCB component database is dynamically established. s (x s ,y s ), where P s (x s ,y s ) represents all PCB component information in the drawing software;

[0095] Based on the standard PCB component database P s (x s ,y s ), P 1 (x1, y1) performs feature clustering data processing to obtain each feature P 1 (x'1,y'1);

[0096] For each characteristic P 1 The various eigenvectors of (x'1, y'1) are restored to obtain the corresponding eigenimages;

[0097] The corresponding feature image is denoised and enhanced to obtain the enhanced feature image.

[0098] Among them, for each characteristic classification P 1 Methods for restoring various eigenvectors of (x'1, y'1) and obtaining corresponding eigenimages include:

[0099] For each characteristic P 1 (x'1, y'1), use convolution to extract and establish a feature vector group, the expression is:

[0100] X={X1,X2,···X n )}

[0101] Among them, X is the feature vector of the feature area, X1,X2,···X n is the eigenvector of each item;

[0102] Restore various eigenvectors and find the corresponding eigenimages. The expression is:

[0103] P={P1,P2,···P n )}

[0104] Among them, P is the feature map corresponding to the feature vector, P1,P2,···P n is the feature image corresponding to each sub-item feature vector, and the expression is: in, Represents all components on the circuit board. Indicates the image information of each component.

[0105] The method of denoising and enhancing the corresponding feature image to obtain the enhanced feature image includes: Each component image in the P s (x s ,y s ) to compare with the standard component image in the image. Wavelet transform threshold technology is used to separate the electronic components of the welded circuit board from the background. The operation is as follows:

[0106] right Perform wavelet transform to extract the high-frequency part. The expression is:

[0107]

[0108] Among them, D H is the high-frequency part of the image, which contains detail information. It means that the wavelet transform takes the high frequency part;

[0109] Sharpen the high-frequency image and get D' H ;

[0110] right Perform wavelet transform to extract the low-frequency part. The expression is:

[0111]

[0112] Among them, D L is the low-frequency part of the feature image, which contains background information. It means that the wavelet transform takes the low-frequency part;

[0113] Right D L Processing, Canny operator product, the expression is:

[0114] D' L =D L *g

[0115] Where D' is the processed low-frequency image, and g is the Canny operator;

[0116] Reconstruct the processed results, the expression is:

[0117] D'=D' L +D' H

[0118] Then D' is the enhanced feature image, which is expressed as D' in the form of a two-dimensional array. i (x,y).

[0119] In this embodiment, the query search set model is established, and the expression is:

[0120]

[0121] Among them, O tr Represents the comparison search set, D' i (x, y) represents the enhanced component images. The collected images are searched in the standard component library. K1 represents the number of searches. D'(,) represents D' i This model means that only the search set is established and no operation is performed.

[0122] In this embodiment, based on the standard library query set model, the method for identifying the electronic component to be tested includes:

[0123] Based on the standard library query set model, a rough search is first performed to determine the type of electronic components, and then a detailed search is performed to determine the size and specifications of the electronic components to obtain the final query results; the expression is:

[0124]

[0125] Among them, the formula Indicates that the captured image is searched in the standard library to determine the type of component. Indicates further search for the determined size, O c Indicates the result of a rough search, O s,m Indicates the final search result, M indicates the identifier that meets the conditions, and D' i (,) represents the set of components on the circuit board. Represents a standard library component, k c ,k' c Indicates the number of searches, O size Indicates the output component size, O type Indicates the output element type; It means that when the component image meets the rough conditions of the components in the standard library, the output logo is: Expressed as The result is further searched, and when it fully meets the conditions of the components in the standard library, the component size and type are output. s,m Indicates the final query result.

[0126] The accuracy of the electronic components on the tested circuit board is determined based on the standard library component parameters, that is, D'(,) and P s (x s ,ys ) is within the preset error range, the identification is correct. Specifically, the minimum difference method is used for judgment, and the formula is:

[0127] Δ=|D'(,)-Ps(,)|

[0128] Where D'(,) represents D' i (,) set, when the error Δ<10 -2 When , the image automatic recognition program ends;

[0129] The identified electronic components’ vias, pads, and connections are inspected to determine whether they meet the requirements. If they meet the requirements, they are considered correct, otherwise they are considered wrong.

[0130] By O s,m Find P in s Compare the PCB electronic component information in (,) to determine the size of each PCB electronic component.

[0131] Create query table 1, O s,m With P s (,) for comparison

[0132] Table 1

[0133]

[0134] Among them, O s,m | i (i=1,2...7) represents the size of each type of each circuit element of the collected circuit board image, P s (,)| i (i=1,2...7) indicates the standard size of the circuit components in the standard library, subscript 1 indicates the size of the pin, subscript 2 indicates the type of pad, subscript 3 indicates the size of the pad, subscript 4 indicates the type of via, subscript 5 indicates the size of the via, subscript 6 indicates the type of wire, and subscript 7 indicates the size of the wire. They are compared with the corresponding types of the standard library component diagram. If the comparison result is consistent, "Y" is output, otherwise "N" is output.

[0135] In this embodiment, by collecting infrared images of electronic components to be tested, using the Retinex method to separate the reflected component from the incident component, an image representing the inherent properties of the circuit board is obtained; the image representing the inherent properties of the circuit board is sharpened in the high frequency band to obtain an image of the foreground of the electronic components; the image of the foreground of the electronic components is enhanced to identify the defect characteristics of the circuit board, including:

[0136] First, the circuit board power supply is powered on, and then infrared detection is performed. The infrared camera is properly heated before collecting images to identify the circuit board, and then the thermal image is collected and scanned by the infrared camera. Figure 2 As shown,

[0137] The collected infrared image is represented by P H (x, y), using the Retinex image enhancement method, the component representing the inherent properties of the circuit board is obtained, and the expression is:

[0138] R(x,y)=lnP H (x,y)-ln[G(x,y)*P H (x,y)]

[0139] Among them, R(x,y) is the image of the inherent properties of the circuit board, and G(x,y) is a Gaussian function that can adjust the effect of image enhancement;

[0140] The image of the inherent properties of the circuit board is decomposed by wavelet into high-frequency components and low-frequency components. The expression is:

[0141]

[0142] in, Represents the high frequency part, Represents the low frequency part, represents a high-frequency image, Represents a low-frequency image;

[0143] Will After sharpening, it is reconstructed with the original low-frequency part to form an enhanced thermal image, and then synthesized with the incident component to form the final thermal image P' H , then P' H is the enhanced image;

[0144] The enhanced component infrared image is used to inspect various defect features of the circuit board.

[0145] Among them, the method of using the enhanced component infrared image to inspect various defect characteristics of the circuit board includes:

[0146] When the circuit board is powered on, current will be generated in the pads, vias, and wires. The location of the cold solder joint will appear dark, the solder joint will appear shiny if it is too small, and the wire will appear shiny if it is too thin. The fault point is determined based on the value of the image pixels.

[0147] Extract the gray value of the background color as the threshold value θ, and judge with the qualified band α. The pixel value of the component image is higher than the threshold value θ, but not too much higher. The pixel value of the part where the solder joint is too small, the connection line is too thin, and the connection line spacing is too small will be much higher than the threshold value θ, generally 2 to 3 orders of magnitude higher. The pixel value of the virtual solder joint and false solder joint position will be much lower than the threshold value θ. In this way, the expression is:

[0148] λ<θ-α indicates a cold solder joint or a false solder joint;

[0149] λ>θ+α means the wire is too thin and the solder joint is too small;

[0150] θ-α<λ<θ+α indicates passing;

[0151] Among them, θ is the threshold, which represents the pixel value of the background, α represents the qualified band, and λ represents the pixel value of the component. A defect query table is established, as shown in Table 2.

[0152] Table 2

[0153]

[0154] in, The pixel value of the standard image representing the size of the circuit board component, that is, the qualified pixel value, The pixel value representing the obtained component size, Indicates the pixel value of the positive and negative poles of the power supply, (1.1~1.5)θ, indicates 1.1~1.5 times the background color value, and θ indicates the background pixel value. V| + Indicates the positive pole of the power supply, V| - Indicates the negative pole of the power supply, Cu indicates copper. When the pixel value of the component is equal to the standard pixel value, it is qualified, otherwise it is unqualified. When the positive and negative poles of the power supply are greater than (1.1-1.5)θ, it is unqualified, otherwise it is qualified. When the pixel value of the copper is equal to the background pixel value θ, it is qualified.

[0155] Embodiment 2

[0156] The present invention also provides an image-based intelligent detection system for welding circuit boards, comprising: an acquisition module, an enhancement module, a component inspection module, a defect recognition module and a type detection module;

[0157] The acquisition module is used to acquire CCD images of electronic components of the circuit board;

[0158] The enhancement module is used to perform image enhancement recognition on the CCD image to obtain an enhanced feature image;

[0159] The component inspection module is used to establish a standard library query set model based on the enhanced feature image, and identify the electronic components to be tested based on the standard library query set model;

[0160] The defect recognition module is used to collect infrared images of the electronic components to be tested, separate the reflected component from the incident component using the Retinex method, obtain an image representing the inherent properties of the circuit board, perform high-frequency sharpening processing on the image representing the inherent properties of the circuit board, obtain an image of the foreground of the electronic components, perform image enhancement on the image of the foreground of the electronic components, and identify the defect characteristics of the circuit board;

[0161] The type detection module is used to establish a unique mapping relationship based on the identified circuit board defect characteristics by establishing a component database and a comparison table of the collected circuit board components, and perform corresponding checks on component types, shape dimensions, connection wire diameters, via types, pad types and dimensions.

[0162] Embodiment 3

[0163] The present invention also provides an image method intelligent detection device for welding circuit boards, comprising: an upper crossbeam support, a horizontal guide rail, a focus knob, a CCD camera, an infrared camera, an upper light source, a support column, a stage, a lower light source, a heater, a data line, a computer, a linear motor and circuit board inspection software;

[0164] The upper crossbeam bracket is connected to the bracket column by screws, and the bracket column connects the upper crossbeam bracket to the stage; the bracket column is connected to the stage by screws; a lower light source and a heater are installed in the stage, and the lower light source is fixed to the bottom plate under the stage by bolts; the heater is fixed to the bottom plate under the stage by bolts; the upper light source is fixed to the bracket column by screws; the CCD camera is connected to coupling 1 by a coupling joint; the infrared thermal camera is connected to coupling 2 by a coupling joint; coupling 1 and coupling 2 are connected and fixed to the horizontal guide rail by bolts; the horizontal guide rail is fixed to the upper crossbeam bracket by screws; the data line is the output signal line of the CCD camera and the infrared camera, and is connected to the computer; a linear motor connects the upper crossbeam bracket to the stage, and the linear motor guide rail is installed on the bracket column;

[0165] The upper crossbeam bracket is used to install CCD cameras and infrared cameras;

[0166] The horizontal guide rail is used to move the CCD camera and infrared camera in the horizontal direction;

[0167] The focus knob is used to adjust the focal length of the CCD camera;

[0168] The CCD camera is used to capture images of electronic components on the PCB circuit board;

[0169] Infrared cameras are used to identify pads, vias, wires, and ground wires on PCBs when power is applied;

[0170] The upper light source is used to shine the light onto the PCB circuit board;

[0171] The support column is used to connect the upper and lower parts;

[0172] The stage is used to place the PCB circuit board;

[0173] The lower light source is used to provide light source for the CCD camera;

[0174] The heater is used for infrared camera to identify the target;

[0175] The data cable is used to connect the output cable of the CCD camera and the infrared camera to the computer;

[0176] Computers are used to collect electronic components on PCB circuit boards, identify errors in circuit board soldering, and areas that need to be optimized;

[0177] The linear motor is used to connect the upper crossbeam support and the stage, and is installed on the support column. It automatically adjusts the height according to different focal lengths to achieve the purpose of imaging.

[0178] The circuit board inspection software is used to operate the CCD camera movement, CCD image recognition electronic components, and capture infrared images for error recognition.

[0179] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An image-based intelligent detection method for welding circuit boards, characterized in that: The following steps are involved: Collect CCD images of electronic components on circuit boards; Perform image enhancement and recognition on the CCD image to obtain an enhanced feature image; Based on the enhanced feature image, a standard library query set model is established, and based on the standard library query set model, the electronic components to be tested are identified; By collecting infrared images of the electronic components to be tested, the Retinex method is used to separate the reflected component from the incident component to obtain an image representing the inherent properties of the circuit board. The image representing the inherent properties of the circuit board is sharpened in the high frequency band to obtain an image of the foreground of the electronic components. The image of the foreground of the electronic components is enhanced to identify the defect characteristics of the circuit board. Based on the identified circuit board defect features, a unique mapping relationship is established by building a component database and a comparison table of the collected circuit board components, and corresponding checks are performed on component type, shape and size, connection wire diameter, via type, pad type and size.

2. The image-based intelligent detection method for welding circuit boards according to claim 1 is characterized in that: The method for collecting CCD images of electronic components of a PCB circuit board includes: The circuit board to be inspected is placed on the stage, and the vertical motor automatically adjusts the focal length so that the clarity of the electronic components on the circuit board meets the preset requirements. The CCD camera is driven by the horizontal motor X direction from the upper left corner of the circuit board to scan along the horizontal X axis direction, and driven by the horizontal motor Y direction to scan along the horizontal Y axis direction. The scanning path starts from the origin (0,0) and scans line by line in the preset arrow direction until it ends at the end point (X,Y), and is repeated three times.

3. The image-based intelligent detection method for welding circuit boards according to claim 2 is characterized in that: The method of performing image enhancement recognition on the CCD image and obtaining the enhanced feature image includes: The CCD image is stored in picture format, represented as P 1 (x1,y1), where P 1 (x1, y1) represents the image of electronic components on the soldered circuit board scanned by the CCD camera; The electronic components PCB library components are retrieved from the drawing application software to form library components in the standard PCB mode, and the shapes and sizes of PCB library components are established. The standard PCB component database is dynamically established. s (x s ,y s ), where P s (x s ,y s ) represents all PCB component information in the drawing software; Based on the standard PCB component database P s (x s ,y s ), P 1 (x1, y1) performs feature clustering data processing to obtain each feature P 1 (x'1,y'1); For each characteristic P 1 The various eigenvectors of (x'1, y'1) are restored to obtain the corresponding eigenimages; The corresponding feature image is denoised and enhanced to obtain the enhanced feature image.

4. The image-based intelligent detection method for welded circuit boards according to claim 3 is characterized in that: Classify each characteristic P 1 Methods for restoring various eigenvectors of (x'1, y'1) and obtaining corresponding eigenimages include: For each characteristic P 1 (x'1, y'1), use convolution to extract and establish a feature vector group, the expression is: X={X1,X2,···X n )} Among them, X is the feature vector of the feature area, X1,X2,···X n is the eigenvector of each item; Restore various eigenvectors and find the corresponding eigenimages. The expression is: P={P1,P2,···P n )} Among them, P is the feature map corresponding to the feature vector, P1,P2,···P n is the feature image corresponding to each sub-item feature vector, and the expression is: in, Represents all components on the circuit board. Indicates the image information of each component.

5. The image-based intelligent detection method for welded circuit boards according to claim 4 is characterized in that: The method of denoising and enhancing the corresponding feature image to obtain the enhanced feature image includes: right Perform wavelet transform to extract the high-frequency part. The expression is: Among them, D H is the high-frequency part of the image, which contains detail information. It means that the wavelet transform takes the high frequency part; Sharpen the high-frequency image to get D, H ; right Perform wavelet transform to extract the low-frequency part. The expression is: Among them, D L is the low-frequency part of the feature image, which contains background information. It means that the wavelet transform takes the low-frequency part; Right D L Processing, Canny operator product, the expression is: D, L =D L *g Among them, D′ L is the processed low-frequency image, g is the Canny operator; Reconstruct the processed results, the expression is: D,=D, L +D, H Then D' is the enhanced feature image, which is expressed in the form of a two-dimensional array as D'(x, y).

6. The image-based intelligent detection method for welded circuit boards according to claim 5, characterized in that: The query search set model established is expressed as: Among them, O tr Denotes the comparison search set, D′ i (x, y) represents the enhanced component images. The collected images are searched in the standard component library. K1 represents the number of searches. D'(,) represents D' i The set of (,).

7. The image-based intelligent detection method for welded circuit boards according to claim 6, characterized in that: Based on the standard library query set model, the method of identifying the electronic components to be tested includes: Based on the standard library query set model, a rough search is first performed to determine the type of electronic components, and then a detailed search is performed to determine the size and specifications of the electronic components to obtain the final query results; the expression is: Among them, O c Indicates the result of a rough search, O s,m represents the final search result, M represents the identifier that meets the conditions, and D′ i (,) represents the set of components on the circuit board. Represents a standard library component, k c ,k′ c Indicates the number of searches, O size Indicates the output component size, O type Indicates the output component type; The accuracy of the electronic components on the tested circuit board is determined based on the standard library component parameters, that is, D'(,) and P s (x s ,y s ) is within the preset error range, the identification is correct. Specifically, the minimum difference method is used for judgment, and the formula is: Δ=|D'(,)-Ps(,)| When the error Δ<10 -2 When , the image automatic recognition program ends; The vias, pads and connections of the identified electronic components are inspected to determine whether they meet the requirements. If they meet the requirements, they are judged to be correct, otherwise they are wrong.

8. The image-based intelligent detection method for welded circuit boards according to claim 1, characterized in that: By collecting infrared images of the electronic components to be tested, the Retinex method is used to separate the reflected component from the incident component to obtain an image representing the inherent properties of the circuit board; the image representing the inherent properties of the circuit board is sharpened in the high frequency band to obtain an image of the foreground of the electronic components; The method of performing image enhancement on the image of the foreground of the electronic component and identifying the defect features of the circuit board includes: The collected infrared image is represented by P H (x, y), using the Retinex image enhancement method, the component representing the inherent properties of the circuit board is obtained, and the expression is: R(x,y)=lnP H (x,y)-ln[G(x,y)*P H (x,y)] Among them, R(x,y) is the image of the inherent properties of the circuit board, and G(x,y) is the Gaussian function; The image of the inherent properties of the circuit board is decomposed by wavelet into high-frequency components and low-frequency components. The expression is: in, Represents the high frequency part, Represents the low frequency part, represents a high-frequency image, Represents a low-frequency image; Will After sharpening, it is reconstructed with the original low-frequency part to form an enhanced thermal image, and then synthesized with the incident component to form the final thermal image P' H , then P' H is the enhanced image; The enhanced component infrared image is used to inspect various defect features of the circuit board.

9. The image-based intelligent detection method for welded circuit boards according to claim 8, characterized in that: Methods for inspecting various defect characteristics of circuit boards using enhanced component infrared images include: When the circuit board is powered on, current will be generated in the pads, vias, and wires, and the fault point will be identified based on the pixel value of the enhanced component infrared image. Extract the gray value of the background color as the threshold value θ, and judge with the qualified band α. The expression is: λ<θ-α indicates a cold solder joint or a false solder joint; λ>θ+α means the wire is too thin and the solder joint is too small; θ-α<λ<θ+α indicates passing; Among them, θ is the threshold value, which represents the pixel value of the background, α represents the qualified band, and λ represents the pixel value of the component.

10. An image-based intelligent detection system for welding circuit boards, characterized in that: include: Acquisition module, enhancement module, component inspection module, defect identification module and type detection module; The acquisition module is used to acquire CCD images of electronic components of the circuit board; The enhancement module is used to perform image enhancement recognition on the CCD image to obtain an enhanced feature image; The component inspection module is used to establish a standard library query set model based on the enhanced feature image, and identify the electronic component to be tested based on the standard library query set model; The defect recognition module is used to collect infrared images of the electronic components to be tested, separate the reflected component from the incident component by using the Retinex method, obtain an image representing the inherent properties of the circuit board, perform high-frequency sharpening processing on the image representing the inherent properties of the circuit board, obtain an image of the foreground of the electronic components, perform image enhancement on the image of the foreground of the electronic components, and identify the defect characteristics of the circuit board; The type detection module is used to establish a unique mapping relationship based on the identified circuit board defect characteristics by establishing a component database and a comparison table of the collected circuit board components, and perform corresponding checks on component types, shape dimensions, connection wire diameters, via types, pad types and dimensions.

Images