CN113724216B

CN113724216B - Method and system for detecting wave crest welding spot defects

Info

Publication number: CN113724216B
Application number: CN202110977751.9A
Authority: CN
Inventors: 曹江中; 孔树荫
Original assignee: Guangdong University of Technology
Current assignee: Guangdong University of Technology
Priority date: 2021-08-24
Filing date: 2021-08-24
Publication date: 2023-03-21
Anticipated expiration: 2041-08-24
Also published as: CN113724216A

Abstract

The invention provides a wave crest welding spot defect detection method and a system, which solve the problem of unsatisfactory detection accuracy of the current wave crest welding spot defect detection method based on automatic optical detection.

Description

Method and system for detecting wave crest welding spot defects

Technical Field

The invention relates to the technical field of machine vision and image processing, in particular to a method and a system for detecting wave crest welding spot defects.

Background

With the rapid development of national economy, the manufacturing industry as a national support is rapidly developing, the yield of global printed circuit boards is increasing day by day, the PCBA is required to be used as the basic material of many high-tech products nowadays, the PCBA is an integrated circuit board formed by welding various components on a blank PCB, and the main components include a chip component, a DIP component, an SIP component and the like. The performance of PCBA board can receive welding quality's direct influence, because the task of present welding is realized by automatic soldering tin machine mostly, consequently because the systematic error of soldering tin machine, the failure of failure etc. reason in welding process, different kinds of defect can appear in soldering tin, if with this kind of PCBA board that has bad soldering tin use in actual science and technology product, will make the life-span greatly reduced of product, can arouse the potential safety hazard even, these are all unacceptable product quality problems. In order to guarantee the production benefit of factories, improve the yield of factory production and reduce the missing detection and error detection of defective products, circuit board manufacturers have been increasing the solder detection requirements of PCBA boards year by year.

At present, many PCBA manufacturers mainly adopt modes such as manual detection, on-line detection, automatic X-ray detection, automatic optical detection and the like for detecting the solder defects. The manual detection mainly comprises that a worker searches for defects on the PCBA by using the aid of tools such as a magnifying glass, however, the manual detection has the disadvantage that the worker is limited in energy, the worker can be in a fatigue state when working for a long time, the PCBA with the defects is easily overlooked, or the correct PCBA is mistakenly detected, so that the false alarm rate and the false alarm rate are increased. On-line detection is a common means for detecting the quality of soldering tin at present, and has the advantages of low detection cost of single plates and strong detection performance, but has the disadvantages of low detection speed and long programming time. The automatic X-ray detection can detect defects with the size of more than 0.1 mu m, but the X-ray has harm to human bodies, and the physical health of workers can be influenced if the X-ray is adopted for detection due to huge yield of production fields.

The automatic optical detection is a method commonly used at present, which is a method for obtaining the surface state of a finished product in an industrial process by using an optical mode and detecting defects such as foreign matters or abnormal patterns by image processing, for example, in 2015, 12 months and 9 days, an image processing method for quickly and automatically detecting unqualified welding spots of an electrical connector is disclosed in a Chinese invention patent (publication No. CN 105139386A), the scheme provided in the patent is to analyze and obtain the defective parts of the welding spots by image processing, so that the detection effect is achieved, the subjective error caused by manual inspection can be eliminated, the detection consistency can be improved, the efficiency and the precision can be improved, the safety of personnel is guaranteed, the color image threshold segmentation is involved in the process, when the color image threshold segmentation is performed, the original connector circumscribed rectangle image is converted into HSV, an H channel image is segmented, the color threshold of the welding spots is used, the H channel binary image is morphologically filtered, the welding spot image is obtained, the requirements on section circle and gray level interpolation of the electrical connector can be reduced, but the welding spots are relatively limited by a single channel image mode, the extraction of the welding spots is not beneficial to the accurate peak ratio detection in a certain binary detection.

Disclosure of Invention

In order to solve the problem that the detection accuracy of the existing wave crest welding spot defect detection method based on automatic optical detection is not ideal, the invention provides a wave crest welding spot defect detection method and system, which can improve the accuracy of welding spot defect detection and optimize the welding spot defect detection effect.

In order to achieve the technical effects, the technical scheme of the invention is as follows:

the application provides a wave crest solder joint detection method, which comprises the following steps:

s1, collecting a template image of the PCBA;

s2, preprocessing a template image, acquiring character outline characteristics, a template welding spot image of a welding spot detection area, a template soldering tin image of a soldering tin detection area and a template welding pad image of a welding pad detection area, and storing the welding spot detection area, the soldering tin detection area and the welding pad detection area;

s3, manually selecting segmentation parameters according with the colors of normal welding spots based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model, performing RGB and HSV color image threshold segmentation on the template soldering tin image and the template pad image, and storing the segmentation parameters;

s4, collecting a real-time detection image of the PCBA, roughly positioning the real-time detection image, and extracting a real-time welding spot image based on the position of a welding spot detection area after the real-time detection image is roughly positioned;

s5, accurately positioning the real-time welding spot image based on a circular template matching method to obtain a welding spot main area and a soldering tin main area of the real-time welding spot image;

s6, reading the segmentation parameters stored in the S3, respectively extracting a real-time pad image and a real-time soldering tin image based on the positions of the pad main area and the soldering tin main area in the real-time soldering point image, respectively carrying out RGB and HSV color image threshold segmentation on the real-time pad image and the real-time soldering tin image, and obtaining the characteristic parameters of the real-time soldering point image according to the segmentation result;

and S7, setting characteristic parameters of the template welding spot image, comparing the obtained characteristic parameters of the real-time welding spot image with the image characteristic parameters of the template welding spot, and classifying the welding spot.

In the technical scheme, the steps S1 to S3 can be used as a stage for constructing the welding spot model, and the steps S4 to S7 can be used as a stage for searching for welding spot defects.

Preferably, the preprocessing of step S2 includes:

carrying out gray level conversion on the template image, and scaling the gray level value of the template image by a linear transformation method to make the contrast of a character area in the template image prominent so as to obtain character outline characteristics; selecting welding spot detection area Reg in template image based on circular tool _check Manually adjust Reg _check Radius of the circular pad detection area Reg _{model_out} And annular solder detection area Reg _{model_in} Extracting a template soldering tin image Img _{model_in} And a stencil pad image Img _{model_out} ；

And generating a template file based on the character outline characteristics, and generating three detection area files based on the welding spot detection area, the soldering tin detection area and the welding pad detection area.

The template image is originally a color image, is converted into a gray image after gray conversion, and then is increased in contrast by a linear transformation method, so that the contrast of a character area is highlighted, the outline of the character area is conveniently obtained, and a template file is generated.

Preferably, in step S3, the process of threshold segmentation of the stencil solder image based on the R, G, B three color components of the RGB color model is as follows:

will Img _{model_in} Dividing the image into RGB three color characteristics, taking the RGB three color characteristics as the index of color segmentation, and manually selecting a first segmentation parameter pair Img which accords with the color of a normal welding spot _modelin Performing threshold segmentation on the RGB color image and taking intersection of connected domains after threshold segmentation to obtain a dark color component;

selecting a first segmentation parameter pair Img _{model_in} Table for performing threshold segmentation of RGB color imageThe expression is as follows:

wherein, (i, j) represents pixel, R (i, j), G (i, j), B (i, j) respectively represent red, green, blue components of pixel, img _RGB (i, j) represents the result after the threshold segmentation processing, R _h 、G _h And B _h Maximum threshold segmentation parameters respectively representing red, green and blue components of RGB; r is _l 、G _l And B _l Respectively representing the minimum threshold segmentation parameters of red, green and blue components of RGB;

when in segmentation, a maximum threshold segmentation parameter R of RGB red, green and blue components is introduced simultaneously _h 、G _h And B _h Minimum threshold segmentation parameter R _l 、G _l And B _l Setting maximum threshold segmentation parameter R of red, green and blue components of RGB _h 、G _h And B _h And a minimum threshold segmentation parameter R _l 、G _l And B _l Realizing the template soldering image Img _{model_in} Obtaining a dark color component based on threshold segmentation of an RGB color model;

the process of simultaneously performing threshold segmentation on the template soldering tin image and the template pad image by using H, S, V three color components of the HSV color model is as follows:

will Img _{model_in} And Img _{model_out} Converting the RGB color model into HSV color model, selecting a second segmentation parameter pair Img by using H, S, V as the index of color segmentation _{model_in} And Img _{model_out} Performing threshold segmentation on the HSV color image and taking intersection of connected domains after threshold segmentation to obtain a blue component and an orange component;

selecting a second segmentation parameter pair Img _{model_in} And Img _{model_out} The expression for performing the threshold segmentation of the HSV color image is as follows:

wherein, (i, j) represents a pixel point, R (i, j), G (i, j), B (i, j) are red, green and blue components of the pixel point (i, j), img _HSV (i, j) is the result of the threshold segmentation process, H _h 、S _h And V _h Maximum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV; h _l 、S _l And V _l Minimum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV;

when in segmentation, maximum threshold segmentation parameters H of hue, saturation and brightness of HSV are introduced simultaneously _h 、S _h And V _h And minimum threshold segmentation parameter H _l 、S _l And V _l Setting the maximum threshold segmentation parameters H of hue, saturation and brightness of HSV _h 、S _h And V _h Minimum threshold segmentation parameter H _l 、S _l And V _l And realizing threshold segmentation of the template pad image and the template soldering tin image based on the HSV color model, and acquiring a blue component and an orange component.

The method is different from the conventional single or two-color component segmentation limit, when the first segmentation parameter is subjected to threshold segmentation, the three color components R, G, B based on an RGB color model are simultaneously segmented, and then the obtained dark color components are combined; when the second segmentation parameter is subjected to threshold segmentation, three color components of hue, saturation and lightness based on the HSV color model are simultaneously segmented to obtain a blue component and an orange component, and the blue component and the orange component are obtained as basic preparation for detecting welding spot defects (blue and orange sensitivity), so that the effective extraction of characteristic parameters of a defect area and the improvement of subsequent detection precision are facilitated.

Preferably, for the coarse positioning of the real-time detection image, based on the position of the welding spot detection area after the coarse positioning of the real-time detection image, the process of extracting the real-time welding spot image is as follows:

reading a template file, acquiring a rotation center and a rotation angle of a real-time detection image character based on a shape matching principle, generating an affine matrix, and thenPerforming affine transformation, detecting the position of the image in real time, rotating the image to the same position as the template image, and reading the welding spot detection area Reg _check Based on Reg _check Cutting the real-time detection image at the positioned position of the real-time detection image, and extracting a real-time welding spot image Img _cur 。

Preferably, in step S5, the real-time solder joint image is accurately positioned based on a circular template matching method, and before obtaining the pad main area and the solder main area of the real-time solder joint image, the method further includes: performing circle fitting on the real-time welding spot image to obtain circle sets C1 and C2;

the specific process is as follows:

s501, pair Img _cur Filtering to obtain filtered image Img _filter Calculating Img _filter Gradient and gradient direction of each pixel of (1), synthesizing a gradient image Img _gradient (ii) a For Img _filter Performing edge processing to obtain an edge image Img _edge ；

S502, calculating Img _gradient The central coordinate of (1) is used as an origin, and n rays and Img with different directions are established in an angle bisection mode _gradient Taking the intersection to form n intersection point sets, respectively solving the extreme points of the n intersection point sets to form n extreme point sets, classifying the extreme points of the n extreme point sets to obtain n beta ₊ Extreme point set of directions E1 _n+ And n beta _- Direction extreme point set E1 _n- (ii) a The classification rules are: calculating the horizontal included angle theta between the nth ray and the image _n At θ _n Is within +/-90 DEG as the positive gradient direction beta of the nth extreme point set ₊ The other angles are taken as the negative direction beta of the gradient of the nth extreme point set _- ；

S503, mixing E1 _n+ And E1 _n- And n rays are mapped to Img _edge At Img _edge To obtain Img _gradient In E1 _n+ 、E1 _n- Extreme point set E1 'of positions corresponding to n rays' _n+ 、E1′ _n- And n' rays, forming an edge map Img _edge Dividing the arc into multiple segments, and determining if there are multiple poles on the segmentThe value points belong to adjacent m rays and the part of the extreme points belong to E1' _n+ Or E1' _n- Wherein m is less than or equal to n', and the central angle theta of the circular arc _c If the angle is more than 45 degrees, the arc of the section is reserved, otherwise, the arc of the section is abandoned; the remaining circular arc forms beta ₊ Sets of arcs of direction L1 and beta _- A directional arc set L2;

s504, arcs in the sets L1 and L2 are sequentially taken, circle fitting is carried out on each arc by using a RANSAC random sampling consistency method, a plurality of groups of circles with different radiuses and circle center coordinates are obtained, the gradient directions of the arcs are mapped to the gradient directions of corresponding circles, and beta is formed ₊ Sets of circles C1 and β of directions _- A set of circles of direction C2.

After the real-time welding spot image of the coarse positioning is obtained, if the welding spot is normal, a circle of white thin circular ring-shaped welding disc is arranged on the periphery, and a white thick circular ring-shaped welding disc is arranged on the outermost side of the welding spot, edges of the places can be obtained according to the color of the welding spot and the gradient change of the color of the white ring, a plurality of circles can be obtained by performing circle fitting on each edge, the edge is obtained according to the gradient change of the real-time welding spot image, and circle sets C1 and C2 are fitted out according to the edges meeting the conditions, so that a plurality of circles meeting the conditions are provided, and a foundation is laid for matching of circular templates.

Preferably, in step S5, the real-time solder joint image is accurately positioned based on the circular template matching method, and the process of obtaining the pad main area and the solder main area of the real-time solder joint image is as follows:

s51, calculating the center coordinates and the radius of each circle in the circle set C1 and the circle set C2;

s52, preprocessing each circle in the circle set C1 and the circle set C2 to further obtain beta ₊ Sets of circles C11 and β of directions _- A set of directional circles C21;

s53, calculating the radius r of each circle in the circle set C11 and the circle set C21 _t1 And r _t2 Taking a circle Q of the circle set C11 according to the rule of descending radius _t1 And a circle Q of the circle set C21 _t2 Matching is carried out, firstly, the distance D1 between the two circle centers is calculated by adopting the following formula:

wherein (x) _t1 ，x _t2 ) And (y) _t1 ，y _t2 ) Are respectively Q _t1 And Q _t2 D1 represents the euclidean distance of the two circle center coordinates, r _t1 And r _t2 Represents the radius of two circles; if D1 is less than r _t1 And r _t2 One third of the absolute value of the difference, the matching is successful, the matching is stopped, and the Q of the successful matching is recorded _t1 And Q _t2 Q1 and Q2, respectively; otherwise, continuing to match until all circles are matched;

s54, selecting all beta in the range of Q2 ₊ The circles in the direction form a circle set C3, and the circles Q of the circle set C3 are sequentially taken according to the rule of descending radius _t3 Matching with the circle Q2, firstly, calculating the distance D2 between the two circle center coordinates by adopting the following formula:

wherein (x) _t2 ，y _t2 ) And (x) _t3 ，y _t3 ) Are circles Q2 and Q, respectively _t3 D2 represents the Euclidean distance of two circle center coordinates, r _t2 And r _t3 Are Q2 and Q, respectively _t3 If D2 is less than r _t2 And r _t3 One third of the absolute value of the difference, if the matching is successful, stopping the matching, and recording the Q of the successful matching _t3 Is Q3; otherwise, continuing to match until all circles are matched;

s55, if Q1, Q2 and Q3 are all matched, the matching is successful and reserved, Q1, Q2 and Q3 are an outer circle, a middle circle and an inner circle in sequence, and a difference set is obtained between Q1 and Q2 to obtain an annular region Reg _{cur_out} As Img _cur The difference between Q3 and a circle Q4 of half the radius of Q3 is used to obtain an annular region Reg _{cur_in} As Img _cur Step S6 is executed in the solder main area of (1); if all the circle sets C1 and C2 have no match after traversalIf a proper circle is matched, matching fails, and Reg is adopted _{model_out} As Img _cur Main area Reg of bonding pad _{cur_out} ，Reg _{model_in} As Img _cur Solder main region Reg of _{cur_in} Step S6 is executed.

The positioning of soldering tin and a welding disc in the wave crest welding spot is optimized through the matching process of the circular templates, so that the positioning accuracy is improved, and more accurate defect areas can be conveniently extracted subsequently.

Preferably, in step S52, the process of preprocessing each circle in the set C1 and the set C2 is as follows:

let Reg _check Has a maximum circumscribed circle radius of R _max Calculating the radius of each circle in the circle sets C1 and C2, taking two circles in the circle sets C1 or C2 according to the rule of radius descending order, and firstly calculating the distance D between the two circle centers by adopting the following formula:

wherein (x) ₁ ，y ₁ ) And (x) ₂ ，y ₂ ) Is the coordinates of the centers of the two circles, D is the Euclidean distance of the coordinates of the centers of the two circles, if the Euclidean distance D is less than R _max One tenth of the above, the two circles are judged to be the same circle. Calculating the radius r of the two circles ₁₁ And r ₁₂ Comparison of r ₁₁ And r ₁₂ Size, only the circle with the largest radius is reserved. After each circle is pretreated, the remaining circles form beta ₊ Sets of circles C11 and β of directions _- A set of circles of direction C21.

Preferably, in step S6, based on Reg _{cur_out} And Reg _{cur_in} At Img _cur Position of (2) to Img _cur Respectively extracting real-time pad images Img _{cur_out} And real-time solder image Img _{cur_in} . The stored segmentation parameters comprise a first segmentation parameter and a second segmentation parameter, and the first segmentation parameter and the second segmentation parameter are respectively used for Img _{cur_in} Performing RGB and HSV color image threshold segmentation, and extracting dark color component regionReg _Deep Blue component region Reg _{Blue_in} And orange component region Reg _orange To get Reg _Deep And Reg _Blue Intersection is made to obtain new solder region Reg _solder Major area Reg of solder _in And Reg _solder Make the complement to obtain the characteristic region Reg _feature ，Reg _feature And Reg _Orange Intersection forming characteristic region Reg _feature2 Calculating Reg _feature2 Area characteristic A of ₁ As characteristic parameters of the real-time soldering tin image;

for Img _{cur_out} Performing HSV color image threshold segmentation, and extracting blue component region Reg _{Blue_out} For Reg _{Blue_out} Performing polar coordinate transformation to obtain region Reg _polar Calculating Reg _polar Area characteristic A of ₂ And height characteristic H ₁ As a characteristic parameter of the real-time pad image.

The characteristic parameters of the real-time soldering tin image and the characteristic parameters of the real-time welding pad image jointly form the characteristic parameters of the real-time welding spot image.

Preferably, the solder joint types include normal, low tin, tin clad, and bridging; wherein, the unqualified types comprise: tin is less, tin is wrapped and bridging is carried out;

the characteristic parameters of the template welding spot image comprise: minimum threshold MIN for tin-poor area, maximum threshold MAX for tin-coated area, and minimum threshold VALUE for bridging area, reg in S5 _out Radius r of the inner and outer circles of ₁ And r ₂ (ii) a Comparing the characteristic parameters of the real-time welding spot image obtained in the step S6 with the characteristic parameters of the template welding spot image, and classifying the welding spots S, wherein the classification standard is as follows:

wherein a1 represents that the welding spot is normal, a2 represents that the welding spot is low in tin, a3 represents that the welding spot is wrapped with tin, a4 represents that the welding spot is bridged, MIN represents a minimum threshold VALUE of a low-tin area, MAX represents a maximum threshold VALUE of a wrapped-tin area, VALUE represents a minimum threshold VALUE of a bridging area, and r represents ₁ Is a main area Reg of a circular bonding pad _out Inner circle radius of (d) ₂ Is a circular ring-shaped bonding pad main region Reg _out The outer radius of the circle.

Because the colors of the tin-poor solder and the tin-coated solder are different from the normal colors, the position of the solder-connected pad is blue, and the maximum radius of the blue is the same as the radius of the pad detection area.

The application also provides a wave soldering point detection system, which is used for realizing the wave soldering point detection method and comprises the following steps:

the image acquisition module is used for acquiring a template image of the PCBA board;

the preprocessing module is used for preprocessing the template image, acquiring the character outline characteristics and the template welding spot image of the welding spot detection area, the template soldering tin image of the soldering tin detection area and the template pad image of the pad detection area, and storing the welding spot detection area, the soldering tin detection area and the pad detection area;

the first threshold segmentation module is used for manually selecting segmentation parameters according with the colors of normal welding spots based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model respectively, performing RGB and HSV color image threshold segmentation on the template soldering tin image and the template pad image, and storing the segmentation parameters;

the acquisition coarse positioning module is used for acquiring a real-time detection image of the PCBA, coarsely positioning the real-time detection image, and extracting a real-time welding spot image based on the position of a welding spot detection area after the coarse positioning of the real-time detection image;

the accurate positioning module is used for accurately positioning the real-time welding spot image based on a circular template matching method to obtain a welding spot main area and a soldering tin main area of the real-time welding spot image;

the second threshold segmentation module is used for reading the stored segmentation parameters, respectively extracting a real-time pad image and a real-time soldering tin image based on the positions of the pad main area and the soldering tin main area in the real-time soldering point image, respectively performing RGB (red, green, blue) and HSV (hue, saturation and value) color image threshold segmentation on the real-time pad image and the real-time soldering tin image, and obtaining the characteristic parameters of the real-time soldering point image according to the segmentation result;

and the comparison and classification module is used for setting the characteristic parameters of the template welding spot image, comparing the obtained characteristic parameters of the real-time welding spot image with the image characteristic parameters of the template welding spot, and classifying the welding spot.

Compared with the prior art, the technical scheme of the invention has the beneficial effects that:

the invention provides a method and a system for detecting defects of wave crest welding spots, wherein the method comprises the steps of firstly carrying out the processes of PCBA template image acquisition, preprocessing and color image threshold segmentation, and is different from the traditional single color component threshold segmentation limit, wherein the processes are respectively based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model, and simultaneously carrying out color image threshold segmentation, so that the threshold segmentation process is optimized, the segmentation result is more suitable for the requirements of welding spot detection, the effective extraction of welding spot defect areas is facilitated, a circular template matching optimization method is introduced, the positioning of soldering tin and a welding pad in the wave crest welding spots is optimized, more accurate defect areas are provided for the subsequent second threshold segmentation, finally, the obtained welding spot image characteristic parameters are compared with the template image characteristic parameters, a defect sample is detected, the accuracy of welding spot defect detection is improved, the yield of PCBA plates produced in factories can be improved, and the use effect of scientific and technological products is finally improved.

Drawings

Fig. 1 is a schematic flow chart illustrating a method for detecting a solder bump defect according to embodiment 1 of the present invention;

FIG. 2 is a diagram showing the positioning effect of the matching of the wave solder point circular template obtained by applying the method proposed in embodiment 1 of the present invention;

FIG. 3 is a schematic view showing a normal solder joint proposed in embodiment 1 of the present invention;

FIG. 4 is a schematic view showing solder joint tin coating proposed in embodiment 1 of the present invention;

FIG. 5 is a schematic view showing solder joint tin reduction proposed in embodiment 1 of the present invention;

FIG. 6 is a schematic view showing a solder joint bridge proposed in embodiment 1 of the present invention;

FIG. 7 is a diagram showing another practical example of solder joint with less tin according to embodiment 1 of the present invention;

FIG. 8 is a diagram showing the detection result of the solder joint defect in the case of performing color threshold segmentation by using only G color component in the conventional RGB color model in embodiment 1 of the present invention;

FIG. 9 is a diagram showing the detection result of the solder joint defect in the embodiment 1 of the present invention, which is obtained by performing color threshold segmentation using the combination of the R color component and the B color component in the conventional RGB color model;

fig. 10 is a diagram illustrating a detection result of a solder joint defect in embodiment 1 of the present invention when color threshold segmentation is performed by using R, G, B color components of an RGB color model and H, S, V color components of an HSV color model;

fig. 11 is a structural diagram of a system for detecting a solder bump defect according to embodiment 2 of the present invention.

Detailed Description

The drawings are for illustrative purposes only and are not to be construed as limiting the patent;

for better illustration of the present embodiment, certain parts of the drawings may be omitted, enlarged or reduced, and do not represent actual dimensions;

it will be understood by those skilled in the art that certain well-known descriptions of the figures may be omitted.

The positional relationships depicted in the drawings are for illustrative purposes only and should not be construed as limiting the present patent;

the technical solution of the present invention is further described below with reference to the accompanying drawings and examples.

Example 1

Referring to fig. 1, the present application provides a method for detecting a solder bump, which includes the following steps:

s1, collecting a template image of the PCBA;

s2, preprocessing a template image, acquiring character outline characteristics, a template welding spot image of a welding spot detection area, a template soldering tin image of a soldering tin detection area and a template pad image of a pad detection area, and storing the welding spot detection area, the soldering tin detection area and the pad detection area;

s6, reading the segmentation parameters stored in the S3, respectively extracting a real-time pad image and a real-time soldering tin image based on the positions of the pad main area and the soldering tin main area in the real-time soldering point image, respectively carrying out RGB and HSV color image threshold segmentation on the real-time pad image and the real-time soldering tin image, and obtaining characteristic parameters of the real-time soldering point image according to a segmentation result;

In combination with the above steps, steps S1 to S3 of the wave solder joint detection method provided in this embodiment can be used as a phase of constructing a solder joint model, the phase of constructing the solder joint model includes processes of PCBA template image acquisition, preprocessing and color image threshold segmentation, and steps S4 to S7 can be used as a phase of searching for solder joint defects, the phase of searching for solder joint defects includes acquiring a real-time detection image of the PCBA board, coarsely positioning the real-time detection image, accurately positioning a real-time detection image of a main area of solder and a main area of a pad based on a principle of circular template matching, reading stored segmentation parameters, performing color image threshold segmentation on the real-time detection images of the main areas of the solder and the pad respectively, acquiring characteristic parameters of the real-time solder joint image, comparing whether the solder joint is qualified or not according to set characteristic parameter rules of the template solder joint image, and classifying the solder joints.

In this embodiment, a template image of the PCBA board is collected using a camera and an annular red, green, and blue LED structure light source, and the preprocessing of the template image in step S2 includes:

carrying out gray level conversion on the template image, and scaling the gray level value of the template image by a linear transformation method to ensure that the contrast of a character area in the template image is prominent; selecting welding spot detection area Reg in template image based on circular tool _check Manually adjusting Reg _check Radius of the annular bonding pad _{model_out} And annular solder detection area Reg _{model_in} Extracting a template solder image Img _{model_in} And a stencil pad image Img _{model_out} The manual adjustment refers to manual operation based on software, and the sliding bar control refers to sliding bar setting of a software interface; and generating a template file based on the character outline characteristics, and respectively generating three detection area files in a welding spot detection area, a soldering tin detection area and a welding pad detection area.

The template image is originally a color image, is converted into a gray image after gray conversion, and then the contrast is increased through a linear transformation method, so that the contrast of a character area is highlighted, the outline of the character area is conveniently obtained, and a template file is generated.

The expression formula of the specific linear transformation is as follows:

GMax is the maximum gray value of the image, GMin is the minimum gray value of the image, mult is a scaling factor, and Add is scaling compensation; extracting outline information of partial characters in the template image as a template file and storing the template file in a PCBA characteristic library, and acquiring a template welding spot image of a welding spot detection area, a template soldering tin image of a soldering tin detection area and a template pad image of a pad detection area in a manual selection mode.

In step S3, based on R, G, B three color components of the RGB color model and H, S, V three color components of the HSV color model, the segmentation parameters conforming to the color of the normal solder joint are manually selected, and the process of performing RGB and HSV color image threshold segmentation on the template solder image and the template pad image at the same time is as follows:

will Img _{model_in} Dividing the image into RGB three color characteristics, taking the RGB three color characteristics as the index of color segmentation, and manually selecting a first segmentation parameter pair Img which accords with the color of a normal welding spot _{model_in} Performing threshold segmentation on the RGB color image and taking intersection of connected domains after threshold segmentation to obtain a dark color component;

selecting a first segmentation parameter pair Img _{model_in} The expression for performing threshold segmentation of an RGB color image is:

wherein, (i, j) represents pixel, R (i, j), G (i, j), B (i, j) respectively represent red, green, blue components of pixel, img _RGB (i, j) represents the result after the threshold segmentation process, R _h 、G _h And B _h Maximum threshold segmentation parameters respectively representing red, green and blue components of RGB; r _l 、G _l And B _l Respectively representing the minimum threshold segmentation parameters of red, green and blue components of RGB;

will Img _{model_in} And Img _{model_out} Conversion from RGB color model to HSV color modelSelecting a second segmentation parameter pair Img by using H, S, V as an index of color segmentation _{model_in} And Img _{model_out} Performing HSV color image threshold segmentation and taking intersection of connected domains after threshold segmentation to obtain a blue component and an orange component;

wherein, (i, j) represents a pixel point, R (i, j), G (i, j), B (i, j) are red, green and blue components of the pixel point (i, j), img _HSV (i, j) is the result after the threshold segmentation process, H _h 、S _h And V _h Maximum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV; h _l 、S _l And V _l Minimum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV;

when in segmentation, maximum threshold segmentation parameters H of hue, saturation and brightness of HSV are introduced simultaneously _h 、S _h And V _h And minimum threshold segmentation parameter H _l 、S _l And V _l Setting maximum threshold segmentation parameters H of hue, saturation and brightness of HSV _h 、S _h And V _h Minimum threshold segmentation parameter H _l 、S _l And V _l And realizing threshold segmentation of the template pad image and the template soldering tin image based on an HSV color model, and acquiring a blue component and an orange component.

In summary, the deep color component, the blue color component and the orange color component are basic color preparations for welding spot detection, are single and are different from the traditional single or two color component segmentation limit, and when the first segmentation parameter is subjected to threshold segmentation, the three color components R, G, B based on an RGB color model are simultaneously segmented, and then the obtained deep color components are combined; when the second segmentation parameter is subjected to threshold segmentation, three color components of hue, saturation and lightness based on the HSV color model are simultaneously segmented to obtain a blue component and an orange component, and the blue component and the orange component are obtained as basic preparation for detecting welding spot defects (blue and orange sensitivity), so that the effective extraction of characteristic parameters of a defect area and the improvement of subsequent detection precision are facilitated. In this embodiment, according to the mapping effect of the ring red, green and blue LED structure light source, the HSV values of the extracted blue component are 145, 155, 69, 255, 116, 255, and the HSV values of the extracted orange component are 0, 61, 118, 255, 132, 255.

In this embodiment, the tools used for acquiring the real-time detection image of the PCBA board in step S4 are the same as the tools in step S1, and are both a camera and an annular red, green and blue LED structure light source, and the real-time detection image of the PCBA board is acquired, and the real-time detection image is coarsely positioned, and based on the position of the solder joint detection area after the real-time detection image is coarsely positioned, the process of extracting the real-time solder joint image is as follows:

reading a template file, acquiring a rotation center and a rotation angle of characters of a real-time detection image based on a shape matching principle, generating an affine matrix, then carrying out affine transformation, rotating the real-time detection image to the same position as the template image, and extracting a welding spot detection real-time image Img of the positioned real-time detection image _cur The specific method comprises the following steps: based on Reg _check Cutting the real-time welding spot image Img at the position positioned by the real-time detection image _cur 。

Wherein, the expression of the affine matrix is:

whereinHomMat2 Dedentity is the identity matrix, homMat2 DRote is the rotation matrix, phi is the rotation angle, homMat2DTranslate is the translation matrix, T is the rotation angle _x Is the horizontal translation distance, T _y Is the vertical translation distance. Affine transformation is carried out on the real-time detection image, namely the real-time detection image is rotated to the same position as the template image, and a welding spot detection area Reg is read _check This is a region file, then based on Reg _check Cutting the real-time welding spot image Img at the position positioned by the real-time detection image _cur 。

In this embodiment, in step S5, the real-time solder joint image is accurately positioned based on a circular template matching method, and the method further includes, before obtaining the pad main area and the solder main area of the real-time solder joint image: performing circle fitting on the real-time welding spot image to obtain circle sets C1 and C2;

the specific process is as follows:

S502, calculating Img _gradient The central coordinate of the X-ray source is used as an original point, and n rays and Img with different directions are established in an angle bisection mode _gradient And taking the intersection to form n intersection point sets, respectively solving extreme points from the n intersection point sets to form n extreme point sets, and classifying the extreme points of the n extreme point sets. The classification rule is as follows: calculating the horizontal included angle theta between the nth ray and the image _n At θ _n Is within +/-90 DEG as the positive gradient direction beta of the nth extreme point set ₊ The other angles are taken as the negative direction beta of the gradient of the nth extreme point set _- Obtaining n beta ₊ Extreme point set of directions E1 _n+ And n beta _- Direction extreme point set E1 _n- (ii) a In this example, img is used _gradient The central coordinate of the ray source is used as an origin, and the better effect can be achieved by establishing 16 rays with different directions in an angle bisection mode, wherein each ray and the Img _gradient All have intersection points, and 16 groups of intersection points form 16 groups of point sets U _n Where n is not more than 16, for U respectively _n Obtaining extreme points to form n extreme point sets Z _n And to Z _n And (3) classifying, wherein the specific method of classification is as follows: after the gradient and gradient direction of each pixel have been previously determined, the nth ray and Img are calculated _gradient Included angle theta _n At θ _n Is within +/-90 DEG as the positive gradient direction beta of the nth extreme point set ₊ The other angles are taken as the negative gradient direction beta of the nth extreme point set _- . Obtaining n beta as a result of the classification ₊ Set of extreme points Z of direction _n+ And n beta _- Set of extreme points Z of direction _n- 。

Z _n+ And Z _n- The gradient values respectively form a one-dimensional array A _n+ And A _n- For the one-dimensional array A _n+ And A _n- Conversion to a one-dimensional function F _n+ And F _n- Search out F _n+ And F _n- All extreme points of (2) form a new extreme point set E _n+ And E _n- . Pair of gradient values E represented by extreme points _n+ And E _n- Sorting in descending order to obtain E _n+ The first six extreme point sets E1 _n+ And E _n- The first three extreme points E1 _n- If the number of extreme points is not enough, all the extreme points are taken, and each ray beta is obtained according to the method ₊ Extreme point set E1 in gradient direction _n+ ，β _- Extreme point set E1 in gradient direction _n- ；

S503, mixing E1 _n+ And E1 _n- And n rays are mapped to Img _edge At Img _edge In order to obtain Img _gradient In E1 _n+ 、E1 _n- Extreme point set E1 'of positions corresponding to n rays' _n+ 、E1′ _n- And n' rays, forming an edge map Img _edge Dividing the arc into a plurality of segments, and judging if a plurality of extreme points exist on the segment of arc and belong to m adjacent rays, and the extreme points belong to E1' _n+ Or E1' _n- Wherein m is less than or equal to n', and the central angle theta of the circular arc _c Greater than 45 deg., then remainAnd if not, discarding the arc. The remaining circular arc forms beta ₊ Sets of arcs of direction L1 and beta _- A set L2 of arcs of direction;

s504, sequentially taking the arcs in the sets L1 and L2, performing circle fitting on each arc by using a RANSAC random sampling consensus method to obtain a plurality of groups of circles with different radiuses and center coordinates, and mapping the gradient directions of the arcs to the gradient directions of the corresponding circles to form beta ₊ Sets of circles C1 and β of directions _- A set of circles of direction C2.

After fitting the circle sets C1 and C2, the step S5 of accurately positioning the real-time welding spot image based on the circle template matching method, wherein the process of obtaining the pad main area and the soldering tin main area of the real-time welding spot image comprises the following steps:

wherein (x) _t2 ，y _t2 ) And (x) _t3 ，y _t3 ) Are circles Q2 and Q, respectively _t3 D2 represents the Euclidean distance of two circle center coordinates, r _t2 And r _t3 Are Q2 and Q, respectively _t3 If D2 is less than r _t2 And r _t3 If the absolute value of the difference is one third, the matching is successful, the matching is stopped, and the Q of the successful matching is recorded _t3 Is Q3; otherwise, continuing to match until all circles are matched;

s55, if Q1, Q2 and Q3 are all matched, the matching is successful and reserved, Q1, Q2 and Q3 are an outer circle, a middle circle and an inner circle in sequence, and a difference set is obtained between Q1 and Q2 to obtain an annular region Reg _{cur_out} As Img _cur Q3 and a circle Q4 with a half of the radius of Q3 to obtain an annular region Reg _{cur_in} As Img _cur Step S6 is executed in the solder main area of (1); if no proper circle is matched after all the circle sets C1 and C2 are traversed, the matching is failed, and Reg is adopted _{model_out} As Img _cur Main region Reg of bonding pad _{cur_out} ，Reg _{model_in} As Img _cur Solder main region Reg of _{cur_in} Step S6 is executed.

The positioning of soldering tin and pad in the wave soldering spot is optimized through the process of matching the circular templates, the positioning accuracy is improved, more accurate defect regions can be conveniently extracted subsequently, the positioning effect diagram of matching the wave soldering spot circular templates obtained by applying the circular template matching method is shown in fig. 2, three matched circles are an outer circle Q1, a middle circle Q2 and an inner circle Q3 from outside to inside in sequence, the difference set is obtained by taking the Q1 and the Q2 to obtain a circular ring-shaped region as a main region of the pad, and the difference set is obtained by taking the Q3 and a circle Q4 with the half of the radius of the Q3 to obtain the circular ring-shaped region as a main region of the soldering tin.

In the present embodiment, in step S6, based on Reg _{cur_out} And Reg _{cur_in} At Img _cur For Img, for _cur Respectively extracting real-time pad images Img _{cur_out} And real-time solder image Img _{cur_in} . The stored segmentation parameters comprise a first segmentation parameter and a second segmentation parameter, and the Img is respectively paired by using the first segmentation parameter and the second segmentation parameter _{cur_in} Performing threshold segmentation on RGB and HSV color images, and extracting a dark color component region Reg _Deep Blue component region Reg _{Blue_in} And orange component region Reg _Orange To get Reg _Deep And Reg _Blue Intersection is made to obtain new solder region Reg _solder Soldering the main area Reg _in And Reg _soldec Making a complementary set to obtain a characteristic region Reg _feature ，Reg _feature And Reg _orange Intersection forming characteristic region Reg _feature2 Calculating Reg _feature2 Area characteristic A of ₁ As characteristic parameters of the real-time soldering tin image;

here, "calculating" means for Reg _feature2 The area of the irregular area can be calculated by the existing mature technology.

For Img _{cur_out} Performing HSV color image threshold segmentation, and extracting blue component region Reg _{Blue_out} For Reg _{Blue_out} Performing polar coordinate transformation to obtain region Reg _polar Calculating Reg _polar Specific surface area ofSymbol A ₂ And height characteristic H ₁ As a characteristic parameter of the real-time pad image. Here, "calculating" means for Reg _polar The area of the irregular region is calculated, and the height of the region is obtained by taking the bottom line of the region as a vertical line, which can be realized by the existing relatively mature technology.

The welding spot types comprise normal, less tin, tin coating and bridging; wherein, the unqualified types comprise: tin is less, tin is wrapped and bridging is carried out;

wherein a1 represents that the welding spot is normal, a2 represents that the welding spot is low in tin, a3 represents that the welding spot is wrapped with tin, a4 represents that the welding spot is bridged, MIN represents a minimum threshold VALUE of a low-tin area, MAX represents a maximum threshold VALUE of a wrapped-tin area, VALUE represents a minimum threshold VALUE of a bridging area, and r represents ₁ Is a circular ring-shaped bonding pad main region Reg _out Inner circle radius of (d) ₂ Is a circular ring-shaped bonding pad main region Reg _out The outer radius of the circle.

More specifically, to further verify the effectiveness of color image threshold segmentation on an image of a solder joint detection area based on three color components R, G, B of an RGB color model and H, S, V of an HSV color model, respectively, compared to the existing segmentation based on a single or two color components, the following description will be further made by taking as an example a situation where a solder joint is tin-poor as shown in fig. 7, an actual schematic view of such a situation refers to fig. 7, a "bright white" in a white circle represents a tin-poor area and is also an area to be detected for a defect, fig. 8 shows a result diagram of a tin-poor defect under color threshold segmentation based on a single G color component of an RGB color model, a defect area matched under the threshold segmentation is a white mark line (black inside), fig. 9 shows a result diagram of a defect area under color threshold segmentation performed based on a R, B color component combination of an RGB color model, a defect area matched under a white mark line (black inside is a white mark line) under the threshold segmentation condition, fig. 10 shows that a defect area matched under the threshold segmentation condition is found by using 3245 of an HSV color model, and a defect area can be detected by a color combination of a white mark line, and the method of a defect detection method of a defect can be performed by using a white mark line (black color component) in a white mark line) when a single G color threshold value, and a defect area under a defect area matching defect area under a threshold value of a defect area matched condition, and a defect detection method of a defect area matched under a defect is found by the invention, sometimes, the color threshold value is poor, and therefore, the color threshold value is unstable, the variation of the welding point cannot be effectively identified by the segmentation parameters of the color threshold value segmentation of the single color component and the combined component of the color threshold value, the false alarm rate is high, and the method provided by the invention is more stable.

Example 2

Referring to fig. 7, the present application further provides a solder bump inspection system, where the system is used to implement the solder bump inspection method described in embodiment 1, and the solder bump inspection system includes:

In summary, by combining the above methods, firstly, the processes of PCBA template image acquisition, preprocessing and color image threshold segmentation are performed, and different from the traditional single color component threshold segmentation limit, the color image threshold segmentation is performed simultaneously based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model, so that the threshold segmentation process is optimized, and effective extraction of a defect area is facilitated, a circular template matching optimization method is introduced, the positioning of soldering tin and a soldering pad in a wave crest soldering point is optimized, a more accurate defect area is provided for the subsequent second threshold segmentation, finally, the obtained real-time image characteristic parameters of the soldering point are compared with the template image characteristic parameters, a defect sample is detected, the accuracy of soldering point defect detection is improved, the yield of the PCBA produced in a factory can be improved, and the use effect of scientific and technological products is finally improved.

It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for detecting wave crest welding spot defects is characterized by comprising the following steps:

s1, collecting a template image of the PCBA;

the preprocessing of step S2 includes: carrying out gray level conversion on the template image, and scaling the gray level value of the template image by a linear transformation method to make the contrast of a character area in the template image prominent so as to obtain character outline characteristics; selecting welding spot detection area Reg in template image based on circular tool _check Manually adjusting Reg _check Radius of the circular pad detection area Reg _{model_out} And an annular solder detection area Reg _{model_in} Extracting a template soldering tin image Img _{model_in} And stencil pad image Img _{model_out} ；

Generating a template file based on the character outline characteristics, and generating three detection area files based on a welding spot detection area, a soldering tin detection area and a bonding pad detection area;

s3, manually selecting segmentation parameters according with the colors of normal welding spots based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model, performing RGB and HSV color image threshold segmentation on the template solder image and the template pad image, and storing the segmentation parameters;

in step S3, based on R, G, B three color components of the RGB color model and H, S, V three color components of the HSV color model, the segmentation parameters conforming to the color of the normal solder joint are manually selected, and the process of performing RGB and HSV color image threshold segmentation on the template solder image and the template pad image is as follows:

wherein, (i, j) represents pixel, R (i, j), G (i, j), B (i, j) respectively represent red, green, blue components of pixel, img _RGB (i, j) represents the result after the threshold segmentation processing, R _h 、G _h And B _h Maximum threshold segmentation parameters respectively representing red, green and blue components of RGB; r _l 、G _l And B _l Respectively representing the minimum threshold segmentation parameters of red, green and blue components of RGB;

will Img _{model_in} And Img _{model_out} Converting the RGB color model into HSV color model, taking H, S, V as the index of color segmentation, selecting a second segmentation parameter pair Img _{model_in} And Img _{model_out} Performing threshold segmentation on the HSV color image and taking intersection of connected domains after threshold segmentation to obtain a blue component and an orange component;

wherein, (i, j) represents a pixel, R (i, j), G (i, j), B (i, j) are red, green and blue components of the pixel, img _HSV (i, j) is the result of the threshold segmentation process, H _h 、S _h And V _h Maximum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV; h _l 、S _l And V _l Minimum threshold segmentation parameters respectively representing hue, saturation and brightness of HSV;

when in segmentation, the maximum threshold segmentation parameter H of hue, saturation and brightness of HSV is introduced simultaneously _h 、S _h And V _h And minimum threshold segmentation parameter H _l 、S _l And V _l Setting the maximum threshold segmentation parameters H of hue, saturation and brightness of HSV _h 、S _h And V _h Minimum threshold segmentation parameter H _l 、S _l And V _l Threshold segmentation of the template pad image and the template soldering tin image based on the HSV color model is achieved, and a blue component and an orange component are obtained;

2. The method for detecting the solder bump defect of a wave crest of claim 1, wherein for the coarse positioning of the real-time detection image, based on the position of the solder bump detection area after the coarse positioning of the real-time detection image, the process of extracting the real-time solder bump image is as follows:

reading a template file, acquiring a rotation center and a rotation angle of a character outline in a real-time detection image based on a shape matching principle, generating an affine matrix, then carrying out affine transformation, and rotating the real-time detection image to the same position as the template image; reading a pad detection area Reg _check Based on Reg _check Cutting the real-time detection image at the position of the positioned real-time detection image, and extracting a real-time welding spot image Img _cur 。

3. The method for detecting the defect of the wave soldering point according to claim 2, wherein before the step S5 of accurately positioning the real-time soldering point image based on the circular template matching method, obtaining the main area of the pad and the main area of the soldering tin of the real-time soldering point image, the method further comprises: performing circle fitting on the real-time welding spot image to obtain circle sets C1 and C2;

the specific process is as follows:

S502, calculating Img _gradient The central coordinate of (1) is used as an origin, and n rays and Img with different directions are established in an angle bisection mode _gradient Taking intersection to form n intersection point sets, respectively solving extreme points from the n intersection point sets to form n extreme point sets, and classifying the extreme points of the n extreme point sets to obtain n beta ₊ Extreme point set of directions E1 _n+ And n beta _- Direction extreme point set E1 _n- (ii) a The classification rules are: calculating the horizontal included angle theta between the nth ray and the image _n At θ _n Is taken as the positive gradient direction beta of the nth extreme point set within the range of +90 DEG ₊ The other angles are taken as the negative direction beta of the gradient of the nth extreme point set _- ；

S503, mixing E1 _n+ And E1 _n- And n rays map to Img _edge At Img _edge To obtain Img _gradient In E1 _n+ 、E1 _n- Extreme point set E1 'of positions corresponding to n rays' _n+ 、E1′ _n- And n' rays, forming an edge map Img _edge Dividing the arc into a plurality of segments, judging if a plurality of extreme points exist on the segment of arc and belong to m adjacent rays, and all the partial extreme points belong to E1' _n+ Or E1' _n- Wherein m is less than or equal to n', and the central angle theta of the circular arc _c If the angle is more than 45 degrees, the arc of the section is reserved, otherwise, the arc of the section is abandoned; the remaining circular arc forms beta ₊ Sets of arcs of direction L1 and beta _- A set L2 of arcs of direction;

4. The method for detecting the wave soldering spot defect of claim 3, wherein the step S5 is to precisely position the real-time soldering spot image based on a circular template matching method, and the process of obtaining the main area of the pad and the main area of the solder of the real-time soldering spot image is as follows:

s52, preprocessing each circle in the circle set C1 and the circle set C2 to further obtain beta ₊ Sets of circles C11 and β of directions _- A set of circles of direction C21;

wherein (x) _t1 ，x _t2 ) And (y) _t1 ，y _t2 ) Are respectively Q _t1 And Q _t2 D1 represents the Euclidean distance of two circle center coordinates, r _t1 And r _t2 Represents the radius of two circles; if D1 is less than r _t1 And r _t2 One third of the absolute value of the difference, the matching is successful, the matching is stopped, and the Q of the successful matching is recorded _t1 And Q _t2 Q1 and Q2, respectively; otherwise, continuing to match until all circles are matched;

wherein (x) _t2 ，y _t2 ) And (x) _t3 ，y _t3 ) Are circles Q2 and Q, respectively _t3 D2 represents the euclidean distance of the two circle center coordinates, r _t2 And r _t3 Are Q2 and Q, respectively _t3 If D2 is less than r _t2 And r _t3 One third of the absolute value of the difference, if the matching is successful, stopping the matching, and recording the Q of the successful matching _t3 Is Q3; otherwise, continuing to match until all circles are matched;

s55, if Q1, Q2 and Q3 are all matched, the matching is successful and reserved, Q1, Q2 and Q3 are an outer circle, a middle circle and an inner circle in sequence, and a difference set is obtained between Q1 and Q2 to obtain an annular region Reg _{cur_out} As Img _cur The difference between Q3 and a circle Q4 of half the radius of Q3 is used to obtain an annular region Reg _{cur_in} As Img _cur Step S6 is executed in the solder main area of (1); if no proper circle is matched after all the circle sets C1 and C2 are traversed, the matching is failed, and Reg is adopted _{model_out} As Img _cur Main area Reg of bonding pad _{cur_out} ，Reg _{model_in} As Img _cur Solder main region Reg of _{cur_in} Step S6 is executed.

5. The method for detecting solder bump defects according to claim 4, wherein in step S52, the preprocessing of each circle in the set C1 and the set C2 comprises:

let Reg _check Has a maximum circumscribed circle radius of R _max Calculating the radius of each circle of the circle sets C1 and C2, taking two circles of the circle set C1 or C2 according to the rule of descending the radius, and firstly calculating the distance D between the two circle centers by adopting the following formula:

wherein (x) ₁ ，y ₁ ) And (x) ₂ ，y ₂ ) Are the coordinates of the centers of two circles, D is the coordinates of two circlesIf Euclidean distance D is smaller than R _max One tenth of the first, the two circles are judged to be the same circle; calculating the radius r of two circles ₁₁ And r ₁₂ Comparison of r ₁₁ And r ₁₂ Size, only the circle with the largest radius is reserved; after each circle is pretreated, the remaining circles form beta ₊ Sets of circles C11 and β of directions _- A set of circles of direction C21.

6. The method for detecting solder bump defects of claim 5, wherein in step S6, the method is based on Reg _{cur_out} And Reg _{cur_in} At Img _cur For Img, for _cur Respectively extracting real-time pad images Img _{cur_out} And real-time solder image Img _{cur_in} (ii) a The stored segmentation parameters comprise a first segmentation parameter and a second segmentation parameter, and the Img is respectively paired by using the first segmentation parameter and the second segmentation parameter _{cur_in} Performing RGB and HSV color image threshold segmentation, and extracting dark color component region Reg _Deep Blue component region Reg _{Blue_in} And orange component region Reg _Orange To get Reg _Deep And Reg _Blue Intersection is made to obtain new solder region Reg _solder Soldering the main area Reg _in And Reg _solder Make the complement to obtain the characteristic region Reg _feature ，Reg _feature And Reg _Orange Intersection forming characteristic region Reg _feature2 Calculating Reg _feature2 Area characteristic A of ₁ As the characteristic parameters of the real-time soldering tin image;

for Img _{cur_out} Performing HSV color image threshold segmentation, and extracting blue component region Reg _{Blue_out} For Reg _{Blue_out} Performing polar coordinate transformation to obtain region Reg _polar Calculating Reg _polar Area characteristic A of ₂ And height characteristic H ₁ As a characteristic parameter of the real-time pad image;

7. The method according to claim 6, wherein the solder joint types include normal, low tin, tin-clad, and bridging; wherein, the unqualified types comprise: tin is less, tin is wrapped and bridging is carried out;

8. A system for detecting solder bump defects, the system being used for implementing the method of any one of claims 1 to 7, the system comprising:

the first threshold segmentation module is used for manually selecting segmentation parameters which accord with the colors of normal welding spots based on R, G, B three color components of an RGB color model and H, S, V three color components of an HSV color model respectively, performing RGB and HSV color image threshold segmentation on the template soldering tin image and the template pad image, and storing the segmentation parameters;

the acquisition coarse positioning module is used for acquiring a real-time detection image of the PCBA, coarsely positioning the real-time detection image and extracting a real-time welding spot image based on the position of a welding spot detection area after the coarse positioning of the real-time detection image;