IC pin defect detection method based on image processing

Abstract

The present invention proposes an IC pin defect detection method based on image processing, which includes: step S1, obtaining IC pin color image; step S2, automatic positioning and segmentation of pin area; step S3, mean filtering and denoising; step S4, based on the local dynamic extreme difference threshold algorithm 2, binary processing is performed on each filtered pin area image; step S5, establishing a low tin feature selection model and a curled pin feature selection model; step S6, automatic positioning and segmentation of non-pin area; step S7, mean filtering and denoising; step S8, based on the local dynamic extreme difference threshold algorithm 2, binary processing is performed on each filtered pin area image; step S9, establishing a tin-connected feature selection model; step S10, IC pin defect determination. The present invention can avoid the drawbacks of traditional algorithm detection and improve the accuracy of IC pin tin-connected, low tin, and curled pin detection.

Description

Technical Field

[0001] The present invention belongs to the field of PCB board production and solder defect detection, and in particular to an IC pin defect detection method based on image processing. Background Art

[0002] During the PCB production process, defects such as lead bridging, insufficient soldering, and curled leads may occur after soldering due to factors such as solder materials and processes, or the IC chip itself. IC chips with these defects are considered substandard PCB products. To prevent substandard PCB products from entering the market, to control PCB production quality, and to improve and optimize PCB production processes, IC chips with these defects must be detected during the production process.

[0003] As a high-tech industry, PCB board production has always had high requirements for production speed and quality. However, due to the influence of raw material quality, placement process and welding process, traditional IC pin detection algorithms cannot specifically solve the adaptability and misjudgment problems caused by the large variety and size differences of IC chips, and are not universally applicable to IC pin detection.

[0004] Currently, IC pin detection still relies mainly on traditional algorithm detection methods, such as: connected domain pin division method, deep learning judgment method, manual threshold adjustment method, etc., among which traditional algorithm detection has many problems:

[0005] 1. The deep learning judgment method requires separate training for each IC chip and is not compatible.

[0006] 2. The IC chip sizes vary greatly, and the connected domain pin division method has the problem that the division conditions are not universal, resulting in the inability to detect.

[0007] 3. The manual threshold adjustment method has too many parameters and is too complicated, requiring the operator to have basic knowledge of optical processing, which increases the difficulty of getting started.

[0008] At present, there is an urgent need for an IC pin defect detection method that can replace traditional algorithm detection, is suitable for the detection of chips of various models and sizes, and has high robustness and high accuracy. Summary of the Invention

[0009] The purpose of this invention is to provide an IC pin defect detection method based on image processing, which can avoid the drawbacks of traditional algorithm detection and improve the accuracy of detecting bridging, insufficient tin, and lifted leads on IC pins. To achieve the above purpose, the technical solution adopted is as follows:

[0010] An IC pin defect detection method based on image processing comprises the following steps:

[0011] Step S1, obtaining a color image of IC pins;

[0012] Step S2: Automatically locate and segment the pin area, specifically including the following steps:

[0013] Step S2.1, mean filtering and denoising;

[0014] Step S2.2, get the grayscale value of all pixels AllValue:

[0015]

[0016] Among them, R N C M Represents the grayscale value of the pixel in the Nth column and the Mth row;

[0017] Step S2.3, calculate the horizontal grayscale average value:

[0018] Using the data in the array of step S2.2, calculate the average grayscale value of a row in the image:

[0019]

[0020] Step S2.4, calculate the vertical grayscale average: use the data in the array of step S2.2 to calculate the average grayscale value of a column in the image;

[0021]

[0022] Step S2.5: Mark the connected domains of the pin area based on the local dynamic threshold algorithm 1, specifically including the following steps:

[0023] Step S2.51, calculating the grayscale range, specifically including:

[0024] Step S2.51A: Create a one-dimensional discrete function, grayscale range f(x), and smooth f(x) using the Gaussian filter model exp(x);

[0025] Among them, the grayscale range f(x) is:

[0026] f(x)=HorProjections(N);x=1~M;

[0027] f(x)=VertProjections(M);x=1~N;

[0028] Among them, the Gaussian filter model exp(x) is:

[0029]

[0030] Step S2.51B, calculate the first-order derivative of f(x) and take the maximum value of the derivative, that is, max[g(x)];

[0031] At this time, the independent variable corresponding to max[g(x)] is x=x0;

[0032] g(x)=f′(x);

[0033] Step S2.52, binary segmentation:

[0034] If f(x)≥f(x0), the pixel corresponding to f(x) is the foreground, and its pixel value is set to 1, otherwise it is set to 0;

[0035]

[0036] Among them, the pixel value after applying the local dynamic extreme difference threshold algorithm is Thres (m,n) ;

[0037] The threshold P of the current pixel (m,n) in the current filtering window (m,n) ;

[0038] Step S2.53, connected domain labeling: For each region with a pixel value of 1 in the grayscale binary image, a connected domain labeling is performed, and morphological processing is performed to obtain a processed connected domain 1;

[0039] Step S2.6: Automatically locate and segment the pin area: The connected domain 1 obtained in step S2.5 is the automatically segmented pin area;

[0040] Step S2.7: Take a minimum bounding rectangle for all the pin areas obtained. This rectangle is the set of all pin areas and non-pin areas, that is, the IC chip pin area.

[0041] Step S3, mean filtering and denoising;

[0042] Step S4: Based on the local dynamic range threshold algorithm 2, each filtered pin area image is segmented on the H, S, and V channels, divided into three images and binary segmented, and the pixels whose grayscale range is not less than the set threshold are set to 1, otherwise they are set to 0, thereby obtaining a grayscale range binary image;

[0043] Step S5: Establishing a low-tin feature selection model and a curled-leg feature selection model:

[0044] Step S6: Automatically locate and segment the non-pin area:

[0045] Step S6.1, removing the pin area obtained in step S2.6 from the IC chip pin area in step S2.7;

[0046] Step S6.2: Connected domain marking;

[0047] Step S7, mean filtering and denoising;

[0048] Step S8: Based on the local dynamic range threshold algorithm 2, each filtered pin area image is segmented on the H, S, and V channels, divided into three images and binary segmented, and the pixels whose grayscale range is not less than the set threshold are set to 1, otherwise set to 0, thereby obtaining a grayscale range binary image;

[0049] Step S9, establishing a tin-linking feature selection model;

[0050] Step S10, IC pin defect determination: determine whether the IC pin to be tested is a defective IC pin based on the insufficient tin feature selection model, the lifted pin feature selection model, and the continuous tin feature selection model.

[0051] Preferably, step S4 includes the following steps:

[0052] Step S4.1, calculate the grayscale range:

[0053] Create a sliding window and slide it on the filtered current pin area image. Count the maximum and minimum values ​​of all pixels in the current sliding window, and use the average of the maximum and minimum values ​​as the grayscale range of the current pin area image.

[0054] Step S4.2, binary segmentation:

[0055] Compare the grayscale range of the current pin area image with the set threshold T. If the grayscale range is not lower than the set threshold T, the pixel corresponding to the grayscale range is the foreground and its value is set to 1, otherwise it is set to 0.

[0056] Step S4.3: Connected domain marking:

[0057] For each area with a pixel of 1 in the grayscale extreme difference binary image, a connected region is marked and morphologically processed to obtain the processed connected region 2.

[0058] Preferably, the low tin feature selection model is as follows: low tin feature selection is performed on each connected area in each threshold segmentation image, thereby detecting whether the threshold segmentation image has low tin; wherein the low tin feature satisfies the following two characteristics at the same time: in the B channel, the area of ​​the climbing area between the pin and the pad is less than the set value; in the R channel, the area of ​​the pad position is greater than the set value;

[0059] The lift-off feature selection model is as follows: for each connected area in each threshold segmentation image, the low-tin feature is selected separately, so as to detect whether the threshold segmentation image has low-tin; wherein, the lift-off feature satisfies the following four features at the same time: the blue area of ​​the climbing area in the B channel is 0; the area of ​​the pad centerline position in the R channel is greater than the set value; the positions on both sides of the pad centerline exist in the B channel or the G channel; the area of ​​the pin end in the R channel is less than the set value.

[0060] Preferably, the Lianxi feature selection model is:

[0061] Connected features are selected for each connected region in each threshold segmentation image to detect whether the threshold segmentation image has connected features.

[0062] Among them, the tin connection feature is to meet the following three characteristics at the same time: the ratio of the solder width to the width of the non-pin area is too large; the ratio of the solder position to the area of ​​the non-pin area is too large; and a single non-pin area is divided by solder.

[0063] Compared with the prior art, the advantages of the present invention are:

[0064] 1. The present invention calculates the average horizontal and vertical grayscale values, and then realizes automatic positioning and segmentation of the pin area based on a local dynamic threshold algorithm. By using mixed processing of multi-channel images, each channel image is used as an information source for determining defects, thereby increasing the credibility and accuracy of defect detection.

[0065] 2. The present invention uses the average vertical and horizontal grayscale values ​​as the condition for dividing pins, eliminating the influence of grayscale and total area on the division caused by different pin sizes. Therefore, it is suitable for the detection of chips of various models and sizes and has strong universality.

[0066] 3. The present invention uses a local dynamic numerical algorithm to segment the image into connected domains instead of using manually fixed numerical values, so it is less affected by the environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Flowchart of the IC pin defect detection method based on image processing;

[0068] Figure 2 is a one-dimensional discrete function graph;

[0069] Figure 3 This is the original pin diagram;

[0070] Figure 4 Pin area separated for automation;

[0071] Figure 5 It is a collection of pins and non-pin areas. DETAILED DESCRIPTION

[0072] The following diagrams illustrate a more detailed description of the IC pin defect detection method based on image processing. These diagrams illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art may modify the present invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as a general guideline for those skilled in the art and not as a limitation of the present invention.

[0073] like Figure 1 As shown, an IC pin defect detection method based on image processing includes the following steps:

[0074] Step S1, obtain IC pin color image, such as Figure 1 shown.

[0075] An image acquisition device including a color camera, a telecentric lens, and an AOI four-color light source is used to obtain images of the IC chip body and all pins on the top, bottom, left, and right (if any) on the PCB.

[0076] The telecentric lens is mounted on the color camera, and the AOI light source is coaxially installed below the lens. The axis of the light source, lens, and camera is perpendicular to the plane of the IC chip, and the IC chip and pins are within the depth of field of the camera lens to ensure that the captured image is clear and usable.

[0077] Step S2: Automatically locate and segment the pin area, specifically including the following steps:

[0078] Step S2.1, mean filtering denoising:

[0079] Using the selective median filtering algorithm, the midpoint value of each IC chip area image obtained in step S1 is replaced by the average of all pixel values ​​above, below, left, right, upper left, lower left, upper right, and lower right. This removes background noise and pin edge noise, thereby generating an enhanced image. This results in a filtered IC chip area image.

[0080] Step S2.2: Get the grayscale values ​​of all pixels:

[0081] The enhanced images obtained in step S2.1 are all subjected to image grayscale processing, and the pixel grayscale value is used to represent the pixel color.

[0082] The grayscale method preferably performs weighted addition of the components of the R (red), G (green), and B (blue) channels of the three-channel color image according to a color formula to obtain the grayscale value of the image and convert it into a single-channel grayscale image. Of course, as an alternative, other grayscale methods in the prior art may also be used.

[0083] A single-channel image is selected according to actual conditions, preferably a color channel that does not exist on the PCB substrate.

[0084] Save the grayscale value of each pixel of the single-channel image in a two-dimensional array in the following format, where R N C M Represents the grayscale value of the pixel in the Nth column and the Mth row:

[0085]

[0086] Step S2.3, calculate the horizontal grayscale average value:

[0087] Using the data in the array of step S2.2, calculate the average grayscale value of a row in the image:

[0088]

[0089] Step S2.4, calculate the vertical grayscale average: using the data in the array of step S2.2, calculate the average grayscale value of a column in the image:

[0090]

[0091] Step S2.5: Mark the connected domains of the pin area based on the local dynamic threshold algorithm 1, specifically including the following steps:

[0092] Step S2.51, calculating the grayscale range, specifically including:

[0093] Step S2.51A, create a one-dimensional discrete function, grayscale range f(x), such as Figure 2 As shown in the figure, a peak in the one-dimensional function represents a pin. The Gaussian filter model exp(x) is used to smooth f(x);

[0094] Among them, the grayscale range f(x) is:

[0095] f(x)=HorProjections(N);x=1~M;

[0096] f(x)=VertProjections(M);x=1~N;

[0097] Among them, the Gaussian filter model exp(x) is:

[0098]

[0099] Step S2.51B, calculate the first-order derivative of f(x) and take the maximum value of the derivative, that is, max[g(x)];

[0100] At this time, the independent variable corresponding to max[g(x)] is x=x0;

[0101] g(x)=f′(x);

[0102] Step S2.52, binary segmentation:

[0103] If f(x)≥f(x0), the pixel corresponding to f(x) is the foreground, and its pixel value is set to 1, otherwise it is set to 0;

[0104]

[0105] Among them, the pixel value after applying the local dynamic extreme difference threshold algorithm is Thres (m,n) ;

[0106] The threshold P of the current pixel (m,n) in the current filtering window (m,n) ;

[0107] Step S2.53, connected domain marking: For each region with a pixel value of 1 in the grayscale binary image, a connected domain marking is performed, and morphological processing is performed to obtain a processed connected domain 1.

[0108] Among them, morphological processing is an existing technology, and preferably: performing a grayscale opening operation on the connected area to obtain the connected area after the opening operation; the opening operation includes using a structural element to perform an erosion operation and a dilation operation on the connected area in sequence.

[0109] A connected domain refers to an area in an image that consists of pixels with the same pixel value and adjacent positions.

[0110] Step S2.6, automatic positioning and segmentation of the pin area: The connected domain 1 obtained in step S2.5 is the automatically separated pin area; Figure 4 shown.

[0111] Step S2.7, take a minimum circumscribed rectangle for all the pin areas obtained, which is the set of all pin areas and non-pin areas, that is, the IC chip pin area; Figure 5 shown.

[0112] Step S3: mean filtering and denoising.

[0113] Based on the selective mean filtering algorithm, each pin area image obtained in step S2.6 is subjected to selective mean filtering and denoising to obtain a filtered pin area image.

[0114] The mean filter mentioned above uses a convolution kernel of size 3x3. Suppose the pixel value of the current pixel (m, n) in the current filter window before applying the mean filter algorithm is A (m,n) The pixel value after is M(m,n) ,but:

[0115]

[0116] Step S4: Based on the local dynamic extreme difference threshold algorithm 2, each filtered pin area image is segmented on the H, S, and V channels, divided into three images and binary segmented. The pixel points whose grayscale extreme difference is not less than the set threshold are set to 1, otherwise they are set to 0, thereby obtaining a grayscale extreme difference binary image, that is, a threshold segmented image.

[0117] The above method of binary segmentation of the pin area image using the local dynamic range threshold algorithm specifically includes the following steps:

[0118] Step S4.1, calculate the grayscale range:

[0119] Create a sliding window and slide it on the filtered current pin area image, count the maximum and minimum values ​​of all pixels in the current sliding window, and calculate the average of the maximum and minimum values. The grayscale extremes of the current pin area image are represented as follows: For example, if the sliding window moves 3 positions along the current image, the current image will have 3 maximum values ​​and 3 minimum values.

[0120] Step S4.2, binary segmentation:

[0121] Compare the grayscale range of the current pin area image with the set threshold T. If the grayscale range is not lower than the set threshold T, the pixel corresponding to the grayscale range is the foreground and its value is set to 1, otherwise it is set to 0.

[0122] Step S4.3: Connected domain marking:

[0123] For each area with a pixel of 1 in the grayscale extreme difference binary image, a connected region is marked and morphologically processed to obtain the processed connected region 2.

[0124] Among them, morphological processing is an existing technology, and preferably: performing a grayscale opening operation on the second connected area to obtain the connected area after the opening operation; the opening operation includes using a structural element to perform an erosion operation and a dilation operation on the second connected area in sequence.

[0125] A connected domain is an area in an image that consists of pixels with the same pixel value and adjacent positions.

[0126] Step S5: Establishing a low-tin feature selection model and a curled-leg feature selection model:

[0127] Among them, the low tin feature selection model: select the low tin feature for each connected area in each threshold segmentation image, so as to detect whether the threshold segmentation image has low tin; the low tin feature is to meet the following two characteristics at the same time: a) in the B channel, the area of ​​the climbing area between the pin and the pad is less than the set value; b) in the R channel, the area of ​​the pad position is greater than the set value;

[0128] The lead feature selection model performs low-tin feature selection on each connected region in each threshold segmentation image, thereby detecting whether the threshold segmentation image has low-tin. The lead feature satisfies the following four characteristics simultaneously: a) In the B channel, the blue area of ​​the climbing area is 0; b) In the R channel, the area of ​​the pad centerline is greater than the set value; c) In the B channel or the G channel, the pad exists on both sides of the centerline; d) In the R channel, the area of ​​the pin end is less than the set value.

[0129] Step S6: Automatically locate and segment the non-pin area:

[0130] Step S6.1: Remove the pin area obtained in step S2.6 from the IC chip pin area in step S2.7.

[0131] Let the pixel value of the current pixel (a, b) in the IC chip pin area minus the pin area be Thres (a,b) , the pixel value of the pin area is ThresY (a,b) ,but:

[0132]

[0133] Step S6.2: Connected domain marking:

[0134] For each area with a pixel of 1 in the grayscale binary image, a connected domain is marked and morphological processing is performed to obtain the processed connected area three, and the connected area three is the divided non-pin area.

[0135] Among them, morphological processing is an existing technology, and preferably: performing a grayscale opening operation on the connected area to obtain the connected area after the opening operation; the opening operation includes using a structural element to perform an erosion operation and a dilation operation on the connected area in sequence.

[0136] Step S7: mean filtering and denoising:

[0137] Median filtering is performed on each non-pin image to remove background noise points and pin edge noise points to obtain an enhanced image.

[0138] Step S8: Based on the local dynamic extreme difference threshold algorithm 2, each filtered pin area image is segmented on the H, S, and V channels, divided into three images and binary segmented, and the pixel points whose grayscale extreme difference is not less than the set threshold are set to 1, otherwise they are set to 0, thereby obtaining a grayscale extreme difference binary image.

[0139] Step S9: Establishing a tin-linking feature selection model:

[0140] For each connected area in each threshold segmentation image, a connection feature (between two pins) is selected to detect whether the threshold segmentation image has a connection.

[0141] Among them, the tin connection feature is to meet the following three characteristics at the same time: a) the ratio of the solder width to the width of the non-pin area is too large; b) the ratio of the solder position to the area of ​​the non-pin area is too large; c) a single non-pin area is divided by solder.

[0142] Step S10, IC pin defect determination: When the IC pin to be tested has any one of the following three items A, B and C, the IC pin to be tested is determined to be a defective IC pin.

[0143] A. In the low tin feature selection model, the product is considered low tin if both conditions are met at the same time;

[0144] B. In the feature selection model of the raised foot, the foot is considered raised if it meets the four conditions.

[0145] C. In the Lianxi feature selection model, the three conditions are met to be Lianxi.

[0146] The above description is merely a preferred embodiment of the present invention and does not limit the present invention in any way. Any person skilled in the art who, without departing from the scope of the present invention, makes any equivalent substitution, modification, or other changes to the technical solution and technical content disclosed in the present invention shall be deemed to be within the scope of the present invention and still fall within the scope of protection of the present invention.

Claims

1. An IC pin defect detection method based on image processing, characterized in that: The following steps are involved: Step S1, obtaining a color image of an IC pin; Step S2, automatic positioning and segmentation of the pin area, specifically includes the following steps: Step S2.1, mean filtering and denoising; Step S2.2, get all pixel grayscale values ​​AllValue: Among them, R N C M It represents the gray value of the pixel in the Nth row and the Mth column; Step S2.3, calculate the horizontal grayscale average value: Using the data in the array of step S2.2, calculate the average grayscale value of a row in the image: Step S2.4, calculating the vertical grayscale average value: using the data in the array of step S2.2, calculating the average grayscale value of a column in the image; Step S2.5: Based on the local dynamic threshold algorithm 1, mark the connected domain of the pin area, specifically including the following steps: Step S2.51, calculating the grayscale range, specifically including: Step S2.51A, create a one-dimensional discrete function, grayscale range f(x), and use the Gaussian filter model exp(x) to smooth f(x); Among them, the grayscale range f(x) is: f(x)=HorProjections(N), x=1~N; when calculating the local dynamic range threshold of the column; f(x)=VertProjections(M), x=1~M; when calculating the local dynamic range threshold of the row; Among them, the Gaussian filter model exp(x) is: Step S2.51B, obtain the first-order derivative of f(x), and take the maximum value of the derivative, that is, take max[g(x)]; At this time, the independent variable corresponding to max[g(x)] is x=x0; g(x)=f ′ (x); Step S2.52, binary segmentation: If f(x)≥f(x0), the pixel corresponding to f(x) is the foreground, and its pixel value is set to 1, otherwise it is set to 0; Among them, the pixel value after applying the local dynamic extreme difference threshold algorithm is Thres (m,n) ; The threshold P of the current pixel (m,n) in the current filter window (m,n ); Step S2.53, connected domain marking: for each region with a pixel value of 1 in the grayscale binary image, a connected domain marking is performed, and morphological processing is performed to obtain a processed connected region 1; Step S2.6, automatic positioning and segmentation of the pin area: the connected domain 1 obtained in step S2.5 is the automatically separated pin area; Step S2.7, make a minimum circumscribed rectangle for all the pin areas obtained, which is the set of all pin areas and non-pin areas, that is, the IC chip pin area; Step S3, mean filtering and denoising; Step S4, based on the local dynamic extreme difference threshold algorithm 2, each filtered pin area image is segmented on the three channels of H, S, and V, and divided into three images and binary segmented, and the pixel points whose grayscale extreme difference is not less than the set threshold are set to 1, otherwise they are set to 0, so as to obtain a grayscale extreme difference binary image; Step S4 includes the following steps: Step S4.1, calculate the grayscale range: Create a sliding window to slide on the filtered current pin area image, count the maximum and minimum values ​​of all pixels in the current sliding window, and take the average of the maximum and minimum values ​​as the grayscale range of the current pin area image; Step S4.2, binary segmentation: Compare the grayscale range of the current pin area image with the set threshold T. If the grayscale range is not lower than the set threshold T, the pixel corresponding to the grayscale range is the foreground and its value is set to 1, otherwise it is set to 0; Step S4.3, connected domain marking: For each area with a pixel of 1 in the grayscale extreme difference binary image, a connected domain is marked and morphologically processed to obtain a processed connected region 2; Step S5, establishing a low-tin feature selection model and a curled-leg feature selection model: Step S6: Automatically locate and segment the non-pin area: Step S6.1, removing the pin area obtained in step S2.6 from the IC chip pin area in step S2.7; Step S6.2, connecting domain labeling, specifically includes: For each area with a pixel of 1 in the gray value binary image, a connected domain is marked and morphologically processed to obtain a processed connected area three, where the connected area three is the divided non-pin area; Step S7, mean filtering and denoising, specifically includes: Median filtering is performed on each non-pin area image; Step S8, based on the local dynamic extreme difference threshold algorithm 2, each filtered non-pin area image is segmented on the three channels of H, S, and V, and divided into three images and binary segmented, and the pixel points whose grayscale extreme difference is not less than the set threshold are set to 1, otherwise they are set to 0, so as to obtain a grayscale extreme difference binary image; Step S9, establishing a tin-linking feature selection model; Step S10, IC pin defect determination: determine whether the IC pin is a defective IC pin based on the insufficient tin feature selection model, the lifted pin feature selection model, and the continuous tin feature selection model.

2. The IC pin defect detection method based on image processing according to claim 1, characterized in that: The low tin feature selection model is: low tin feature selection is performed for each connected area in each threshold segmentation image, so as to detect whether the threshold segmentation image has low tin; wherein, the low tin feature satisfies the following two characteristics at the same time: in the B channel, the area of ​​the climbing area between the pin and the pad is less than the set value; in the R channel, the area of ​​the pad position is greater than the set value; The model for selecting the tip-up feature is as follows: a tip-up feature is selected for each connected area in each threshold segmentation image, so as to detect whether the threshold segmentation image has a tip-up; wherein, the tip-up feature satisfies the following four features at the same time: the blue area of ​​the climbing area in the B channel is 0; the area of ​​the pad centerline position in the R channel is greater than the set value; the positions on both sides of the pad centerline exist in the B channel or the G channel; the area of ​​the pin end in the R channel is less than the set value.

3. The IC pin defect detection method based on image processing according to claim 1, characterized in that: Lianxi feature selection model is: For each connected region in each threshold segmentation image, a connection feature selection is performed to detect whether the threshold segmentation image has a connection; Among them, the tin connection feature is to simultaneously meet the following three characteristics: the ratio of the solder width to the width of the non-pin area is too large; the ratio of the solder position to the area of ​​the non-pin area is too large; and a single non-pin area is divided by solder.

Images