CN107945184B - Surface-mounted component detection method based on color image segmentation and gradient projection positioning - Google Patents
Surface-mounted component detection method based on color image segmentation and gradient projection positioning Download PDFInfo
- Publication number
- CN107945184B CN107945184B CN201711161398.7A CN201711161398A CN107945184B CN 107945184 B CN107945184 B CN 107945184B CN 201711161398 A CN201711161398 A CN 201711161398A CN 107945184 B CN107945184 B CN 107945184B
- Authority
- CN
- China
- Prior art keywords
- image
- color
- positioning
- area
- detection
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30148—Semiconductor; IC; Wafer
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Supply And Installment Of Electrical Components (AREA)
Abstract
The invention discloses a surface-mounted component detection method based on color image segmentation and gradient projection positioning, and belongs to the technical field of automatic optical detection of surface-mounted components. The invention firstly collects the color image through the designed image collecting system, then converts the red, green and blue three-color image into the HSI color model composed of three parameters of hue H (hue), saturation S (saturation) and intensity I (intensity), because the HSI model directly uses the hue and saturation which are irrelevant with the brightness when extracting the color information, the method has the characteristics of accurate and high-efficient division, and carries out the positioning and detection of the element on the basis, the positioning mainly adopts the color and the geometric characteristic, and provides the positioning method based on the gradient projection and the main component characteristic value detecting method based on the area, and the provided method has the advantages of high positioning precision, high efficiency and accuracy in detection.
Description
Technical Field
The invention relates to the technical field of mounted component detection, in particular to a mounted component detection method based on color image segmentation and gradient projection positioning.
Background
Electronic products are widely applied to various industrial fields, and surface-mounted electronic components are used as core components of the electronic products, so that the performance of the electronic products is directly influenced by the performance of the electronic components. At present, electronic components are mainly mounted on a circuit board of an electronic product through soldering tin by adopting a surface mounting technology, so that quality detection of welding spots is very critical. Common types of solder joint defects are solder skip, solder starvation, component misalignment, misplacement and tombstoning. The welding spot quality is detected by means of manual visual inspection traditionally, but the density becomes high each other along with the size of an electronic element becomes small, so that accurate judgment is difficult to make by human eyes, and visual fatigue is easy to cause and detection efficiency is not high due to the fact that the manual visual inspection depends on subjective experience. After the lead-free production mode is implemented, the surface color of the welding spot is darker, the manual detection is not facilitated, and the defects are more. Therefore, it is increasingly necessary to detect the quality of the welding spot in an automatic manner.
The automatic optical detection adopts an image acquisition system to acquire images of welding spots on a circuit board, and judges whether the quality of the welding spots is qualified or not by performing characteristic extraction, processing and analysis on the images. The detection method can detect the welding points of the circuit board with high mounting density and the tiny electronic components. Because of the characteristics of standard detection, effective product quality management, high efficiency and the like, automatic optical detection is increasingly used in surface mount production lines.
The common methods for detecting the mounted components of the printed circuit board by automatic optical detection include a reference comparison method and a non-reference comparison method, the reference comparison method compares an image to be detected with a reference image called a gold sample, and because the reference image provides comparison, the method looks simple in principle, and a defect can be calculated by finding a place different from the standard image. For example, the Chinese patent No. CN201510174839.1, the application date is 2015, 4 and 14, the invention and creation name is: a method for detecting the mounting precision of an electronic element; the method comprises the steps of comparing an image of a pre-detection circuit board, which is shot in a conversion mode, with an ideal image preset in a computer processing system, and determining the comparison between the positioning coordinates of the electronic element in the pre-detection circuit board and standard coordinate parameters preset in the computer processing system by analyzing the image of the pre-detection circuit board, so that the accurate and inaccurate corner point positioning of the electronic element on the circuit board and the wrong result of corner point detection are obtained. Compared with the traditional manual packaging detection method, the reference comparison method reduces the labor cost and greatly reduces the detection error rate. The main disadvantage of this method is however that the positioning must be accurate and that the illumination is uniform, which is the case of false positives for images that are not perfectly uniform but that are still acceptable with "gold samples".
The non-reference comparison method can realize the defect detection without reference to a standard image. The method mainly has the advantages that accurate positioning is not needed, and standard images occupying a large amount of storage space are not needed, such as patent number CN 201510357803.7, the application date is 2015, 6 and 25, the invention is named as: a visual-based TR element positioning and defect detection method; the application is a scheme for defect detection by adopting a non-reference comparison method, and comprises the steps of extracting an outer boundary point set through a binarization region image, searching an effective boundary point set, searching a minimum external rectangle, classifying the effective boundary point set, carrying out affine transformation on a binary image, checking the type of a TR element, classifying and numbering the effective boundary point set again, fitting a pin straight line, determining the detail information of the TR element, and checking the defects of the pins of the TR element. The application can solve the problems of personal errors, low precision, poor real-time performance and sensitivity of a calculation result to illumination of surface-mounted element detection, but the application has a complex detection process, uses the technologies of 4-field rapid contour tracing, a biaxial rotation method for searching the minimum external rectangle and the like, and is not easy to realize.
Disclosure of Invention
1. Technical problem to be solved by the invention
The invention aims to solve the problems of low positioning precision and high false detection rate in the existing pasted component defect detection scheme, and provides a pasted component detection method based on color image segmentation and gradient projection positioning; the invention relates to a regular detection method for extracting characteristics of a mounted electronic component in different regions, which is not easily interfered by noise and has good robustness for different illumination conditions.
2. Technical scheme
In order to achieve the purpose, the technical scheme provided by the invention is as follows:
the invention discloses a surface-mounted component detection method based on color image segmentation and gradient projection positioning, which comprises the following steps:
firstly, adopting a camera and an annular red, green and blue LED structure light source to acquire images of welding spots on a circuit board;
step two, converting the image obtained in the step one from an RGB color model into an HSI color model, and carrying out color image segmentation on the basis;
step three, setting MARK points for the image obtained in the step two, and compensating the errors in the xy two directions in the PCB image by using the set MARK points;
fourthly, positioning a pad body area and a mounting character area;
and step five, detecting welding spots of the bonding pad and the main body area.
Furthermore, the defect detection of the surface-mounted component comprises a training stage and a testing stage, wherein the training stage and the testing stage respectively comprise the processes of image acquisition, color model conversion, MARK point matching, color image segmentation, welding point positioning and characteristic parameter detection; in the training stage, image acquisition is carried out on welding spots on a circuit board, MARK points are manufactured, color image segmentation is carried out, and welding spot image characteristic parameters are determined in advance and stored; in the testing stage, a MARK point recognition mode is adopted to compensate the positioning error of the circuit board, then the welding spot image of the circuit board is collected in real time, color image segmentation is carried out, the characteristic parameters of the welding spot image in the detection area are extracted and compared with the characteristic parameters of the welding spot image stored in the training stage, and whether the welding spot is qualified or not is judged.
Furthermore, after the image is converted from the RGB color model to the HSI color model, the two characteristic indexes of hue and saturation are used as the basis of color segmentation, and the segmentation is performed according to the following formula:
in the formula, H (i, j) and S (i, j) are hue and saturation of the pixel point (i, j), img (i, j) is a binarization processing result, and the highest value H of hue H and saturation S is sethAnd ShAnd a minimum value HlAnd SlI.e. capable of doing HSI-based colorAnd (5) color segmentation of the model to obtain the specified color.
Furthermore, errors in two xy directions of the whole circuit board are compensated by two MARK points, the area shape characteristics of the MARK points are extracted in a search region, the MARK MARK points are subjected to threshold segmentation, and then the image is subjected to binarization, specifically as follows:
the gray value f (x, y) of a pixel point corresponding to a coordinate point (x, y) of an image with the length and the width of M × N is according to a color threshold value [ c ]min,cmax]The binarization process is performed, and r (x, y) is the result of the process
Calculating the area of the MARK Point
And comparing the result of the area calculation with the setting, and finishing the setting of the MARK point.
Furthermore, the red, green and blue colors of the pad region are sequentially extracted in the positioning of the pad body region, and the calculation formula is as follows:
Xkis a three-dimensional feature vector, Tr(x, y) is the percentage of the r-th color frame color image, r is 1, 2 or 3, 1 represents red, 2 represents green, 3 represents blue, and M × N is the number of pixels in the pad area.
Furthermore, the positioning of the surface-mounted character area adopts a gradient integral projection-based method, firstly, a Sobel operator gradient image is adopted to perform projection positioning on the gradient image; wherein, after gradient processing is performed on the image, f (x, y) represents a gradient value at a point (x, y), w and h represent widths and heights of the mounting character regions, and a density function f is obtained by projecting the function f (x, y) onto the x-axis and the y-axisx(x, y) and fy(x, y), and then positioning;
furthermore, the detection of the pad area and the body area is realized by extracting the area, the gravity center characteristic and the principal component characteristic value under red, green and blue colors in the area and thinning the pad area into a plurality of sub-areas.
Further, when extracting principal component eigenvalue, a small window with fixed size is selected in the image, then covariance in two directions of x-axis and y-axis is calculated in the window with gray-scale value as density to form a covariance matrix, two eigenvalues of the covariance matrix are calculated by principal component analysis, the two eigenvalues represent two maximum variance values orthogonal to each other in two-dimensional space, and the difference between the eigenvalue of normal image and abnormal image is used as detection index.
3. Advantageous effects
Compared with the prior art, the technical scheme provided by the invention has the following remarkable effects:
(1) the invention relates to a mounting component detection method based on color image segmentation and gradient projection positioning, which adopts an HSI color model to carry out image segmentation, wherein two HS characteristics in three basic characteristic quantities of hue H (hue), saturation S (saturation) and intensity I (intensity) of the model are factors irrelevant to brightness, two characteristic indexes of hue and saturation are used as bases for color segmentation, and a color image segmentation method combining hue and saturation is more suitable for the situation of frequently encountered light source brightness changeability compared with the color segmentation based on an RGB model because the method is irrelevant to the brightness factor;
(2) the invention relates to a surface-mounted component detection method based on color image segmentation and gradient projection positioning, which adopts a gradient image projection positioning method, is not easy to be interfered by noise, and has good robustness for different illumination conditions;
(3) the invention discloses a mounted Component detection method based on color image segmentation and gradient projection positioning, which adopts a characteristic value detection method based on Principal Component Analysis (PCA), selects a small window with a fixed size in an image, then calculates covariance in two directions of an x axis and a y axis by taking a gray-scale value as density in the window to form a covariance matrix, calculates two characteristic values (Eigenvalue) of the covariance matrix by using the Principal Component Analysis method, the two characteristic values represent two maximum variance values which are orthogonal in a two-dimensional space, can design a characteristic index of consistency by using the difference of the characteristic values of a normal image and an abnormal image, and has the characteristics of high calculation efficiency and good robustness.
Drawings
FIG. 1 is a schematic diagram of a surface mount component inspection flow of the present invention;
FIG. 2 is a schematic diagram of the location of the MARK point of the present invention;
FIG. 3 is a schematic diagram of pad color extraction according to the present invention;
fig. 4 is a schematic illustration of a sub-window of a pad region of the present invention.
Detailed Description
For a further understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings and examples.
Example 1
The embodiment provides a method for detecting a surface-mounted electronic component as a component of a circuit board, which specifically comprises the following processing procedures based on color image segmentation and gradient projection positioning, with reference to fig. 1, because the conventional manual visual inspection has the defects of high false alarm and low efficiency, and automatic optical detection is adopted to automatically detect the solder joint of the surface-mounted electronic component, which is the main current trend:
component defect detection
A training stage: and acquiring a PCB (printed Circuit Board) picture, and acquiring an image of a welding spot on the circuit board by adopting a device of a 3CCD (charge coupled device) camera, an annular red, green and blue LED (light emitting diode) structure light source and a lens. And then, manufacturing a MARK point, determining well characteristic parameters of the welding spot image in advance according to the color image segmentation principle, and storing.
(II) a testing stage: firstly, positioning compensation is carried out, and the positioning error of the circuit board is compensated by adopting a MARK point identification mode. And then collecting the welding spot image of the circuit board in real time, segmenting according to the proposed color image segmentation principle, extracting the characteristic parameters of the welding spot image in the detection area, comparing the characteristic parameters with the characteristic parameters of the welding spot image stored in the training stage, and judging whether the welding spot is qualified.
Therefore, the steps of the training stage and the testing stage comprise image acquisition, MARK point matching, color image segmentation, welding point positioning and characteristic parameter detection.
In the embodiment, image acquisition is carried out, and the surface of the welding spot keeps the property of a plane mirror, which follows the law of reflection. After the light strikes the solder joint surface, it is reflected to the red, green, and blue light of the camera, representing flat, slowly sloping, and rapidly sloping portions, respectively. In a 3-color illumination and color camera system, the three-dimensional shape information of the weld spot can be depicted by a two-dimensional color image.
After image acquisition, the color of the image must be set during printed circuit board inspection, and the RGB color model is often susceptible to interference. Since lead is not present, the saturation of the solder joint is reduced, and it is necessary to consider using a color model that minimizes the brightness factor, and this embodiment employs the HSI color model.
Converting the RGB model to the HSI model:
for the range RGB values [0, 1], the corresponding H, S and I are calculated in the HSI model components:
in the formula (1), if G is larger than or equal to B, H value is calculated in the interval [0 degrees and 180 degrees ], and if G is smaller than B, H is set to 360-H, so that H can be converted to [180 degrees and 360 degrees ].
In the HSI color model, two features HS of three basic feature quantities, namely hue H (hue), saturation S (saturation) and intensity i (intensity), are factors unrelated to brightness, so that two feature indexes of hue and saturation are used as bases for color segmentation, a color image segmentation method combining hue and saturation is shown in formula 4, where (i, j) is a corresponding pixel point in an image, H (i, j) and S (i, j) are hue and saturation of the pixel point (i, j), and img (i, j) is a binarization processing result. Since the method is independent of brightness factors, the method is more suitable for the situation that the brightness of the light source is variable compared with the color segmentation based on the RGB model.
By setting the maximum value H of hue H and saturation ShAnd ShAnd a minimum value HlAnd SlThe color segmentation based on the HSI color model can be performed to segment the designated color.
Positioning of mounted electronic components
The positioning of the surface-mounted electronic component is a precondition for detection, and the positioning effect directly affects the quality of the detection.
MARK Point positioning As shown in FIG. 2, the xy two-directional error of the entire board can be compensated by two MARK points. The method is used for positioning compensation, the area shape characteristics of the MARK points are extracted from the search region, the threshold segmentation is firstly carried out on the MARK MARK points, and then the image binarization is carried out. The method comprises the following specific steps:
gray values f (x, y) of pixel points corresponding to coordinate points (x, y) of an image with the length and the width of M × N respectively are according toColor threshold value cmin,cmax]The binarization process is performed, and r (x, y) is the result of the process
Calculating the area of the MARK Point
The MARK point can be set by comparing the area calculation result with the setting, and the positioning principle is shown in fig. 2.
The pad body region is positioned by sequentially extracting red, green and blue colors of the pad region, and a pad positioning color extraction schematic diagram is shown in fig. 3, wherein: a is an automatic drawing window, B is a bonding pad window, C is a mounting window, and D is a main body window. The calculation formula is shown in the following formula (7)
XkIs a three-dimensional feature vector, Tr(x, y) is the percentage of the r-th color frame (1 ═ red, 2 ═ green, 3 ═ blue) color image, and M × N is the number of pixels in the pad area.
Since the positioning of the mounting text area is easily interfered by the outside, the embodiment provides a method based on gradient integral projection, which specifically includes the following steps:
1) gradient image solving by Sobel operator
The Sobel operator is an edge detection operator based on first order differential, and examines the weighted difference of the gray values of the upper, lower, left and right adjacent points of each pixel f (i, j) in the image, and the weight of the adjacent points close to the weighted difference, and is defined as follows:
Sobel(i,j)=|Δxf|+|Δyf|
=|f(i-1,j-1)+2f(i-1,j)+f(i-1,j+1)+f(i+1,j-1)+f(i+1,j)+f(i+1,j+1)|+|f(i-1,j-1)+2f(i,j-1)+f(i+1,j-1)+f(i-1,j+1)+f(i,j+1)+f(i+1,j+1)|
the convolution operator is:
and carrying out convolution summation operation on the image by adopting the convolution operator to obtain the gradient image based on the Sobel operator.
2) Projection positioning of gradient images
After the image is processed by gradient, the character area can be characterized by f (x, y) which represents the gradient value at point (x, y), w and h correspond to the width and height of the character area, and the density function f is obtained by projecting the function f (x, y) to the x-axis and the y-axisx(x, y) and fy(x, y), and then positioning.
Detection of pad and body regions
The detection of the pad region and the body region is mainly performed by extracting features of the region, such as shapes, colors and the like under three colors of red, green and blue, and the pad region can be subdivided into a plurality of sub-regions, and the region schematic diagram is shown in fig. 4 and table 1.
TABLE 1 detection Window
The features extracted are:
1) normalized area
r (x, y) is the gray value f (x, y) of the pixel point corresponding to the coordinate point (x, y) of the image with the length and the width of M × N according to the color threshold value [ c [min,cmax]The result of the binarization processing is the same as the MARK point feature extraction method described above.
2) Normalized center of gravity
Let the color center of gravity of the object be (x)m,ym) Normalized center of gravity of (G)R,X,GR,Y) Then G isR,X=xm/M,GR,Y=ymand/N. Thus, GR,XAnd GR,YThe definition is as follows:
3) principal component eigenvalue
Principal component analysis is a powerful tool for extracting the structure of a data set from characteristic value analysis. The square root of the two eigenvalues is geometrically related to the length of the major and minor axes of an ellipse, corresponding to the two-dimensional case. They statistically correspond to the two largest variances in the orthogonal principal directions.
Principal component analysis is applied to one gray-scale image, F is the sum of all gray-scale values in the (2w +1) × (2w +1) size image,
the center of gravity of all pixels of the image is given by
Wherein the content of the first and second substances,
the weighted covariance matrix M of the eigenvalues λ is obtained by the following equation
M-λ·I=0 (17)
The covariance matrix is 2 × 2, a symmetric semi-positive definite matrix, thus, the covariance matrix M of the two eigenvalues is given by2≤λ1
In the embodiment, a small window with a fixed size is selected in an image, then covariance in two directions of an x axis and a y axis is calculated in the window by taking a gray-scale value as a density, so that a covariance matrix can be formed, two eigenvalues (eigenvalues) of the covariance matrix are calculated by using a principal component analysis method, the two eigenvalues represent two maximum variance values which are orthogonal to each other in a two-dimensional space, a characteristic index of consistency can be designed by using the difference of characteristic values of a normal image and an abnormal image, and the method has the characteristics of high calculation efficiency and good robustness.
The present invention and its embodiments have been described above schematically, without limitation, and what is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Therefore, if the person skilled in the art receives the teaching, without departing from the spirit of the invention, the person skilled in the art shall not inventively design the similar structural modes and embodiments to the technical solution, but shall fall within the scope of the invention.
Claims (1)
1. A surface-mounted component detection method based on color image segmentation and gradient projection positioning comprises the following steps:
firstly, adopting a camera and an annular red, green and blue LED structure light source to acquire images of welding spots on a circuit board;
step two, converting the image obtained in the step one from an RGB color model into an HSI color model, and carrying out color image segmentation on the basis; after an image is converted into an HSI color model from an RGB color model, two characteristic indexes of hue and saturation are used as the basis of color segmentation, and the segmentation is carried out by adopting the following formula:
in the formula, H (i, j) and S (i, j) are hue and saturation of the pixel point (i, j), img (i, j) is a binarization processing result, and the highest value H of hue H and saturation S is sethAnd ShAnd a minimum value HlAnd SlNamely, the color segmentation based on the HSI color model can be carried out, and the specified color can be segmented;
thirdly, setting a MARK point for the image obtained in the second step, and compensating the error of the PCB image in the x direction and the y direction by using the set MARK point; the errors of the whole circuit board in the x direction and the y direction are compensated by two MARK points, the area shape characteristics of the MARK points are extracted in a search region, the MARK MARK points are subjected to threshold segmentation at first, and then the image is subjected to binarization, specifically as follows:
the gray value f (x, y) of a pixel point corresponding to a coordinate point (x, y) of an image with the length and the width of M × N is according to a color threshold value [ c ]min,cmax]The binarization process is performed, and r (x, y) is the result of the process
Calculating the area of the MARK Point
The setting of the MARK point can be completed by comparing the area calculation result with the setting;
fourthly, positioning a pad body area and a mounting character area; sequentially extracting red, green and blue colors of the pad region in the positioning of the pad body region, wherein the calculation formula is as follows:
Xkis a three-dimensional feature vector, Tr(x, y) is the percentage of the color image of the r-th color frame, r is 1, 2 or 3, 1 represents red, 2 represents green, 3 represents blue, M × N is the number of pixels in the pad area, the positioning of the mounting character area adopts a gradient integral projection-based method, firstly, a Sobel operator is used for graduating the image to perform projection positioning on the gradient image, after the image is subjected to gradient processing, g (x, y) represents the gradient value of a point (x, y), w and h represent the width and height of the mounting character area, and a density function g is obtained by projecting a function g (x, y) onto the x axis and the y axisx(x, y) and gy(x, y), and then positioning;
detecting the pad area and the main body area by extracting the area, the gravity center characteristic and the main component characteristic value under red, green and blue colors in the area and thinning the pad area into a plurality of sub-areas;
fifthly, detecting welding spots of the bonding pad and the main body area;
the defect detection of the surface-mounted element comprises a training stage and a testing stage, wherein the training stage and the testing stage respectively comprise the processes of image acquisition, color model conversion, MARK point matching, color image segmentation, welding point positioning and characteristic parameter detection; in the training stage, image acquisition is carried out on welding spots on a circuit board, MARK points are manufactured, color image segmentation is carried out, and welding spot image characteristic parameters are determined in advance and stored; in the testing stage, a MARK point recognition mode is adopted to compensate the positioning error of the circuit board, then the welding spot image of the circuit board is collected in real time, color image segmentation is carried out, the characteristic parameters of the welding spot image in the detection area are extracted and compared with the characteristic parameters of the welding spot image stored in the training stage, and whether the welding spot is qualified or not is judged; wherein the content of the first and second substances,
when a principal component characteristic value is extracted, selecting a small window with a fixed size in an image, then calculating covariance in two directions of an x axis and a y axis by taking a gray-scale value as a density in the window to form a covariance matrix, calculating two characteristic values of the covariance matrix by using a principal component analysis method, wherein the two characteristic values represent two maximum variance values which are orthogonal in a two-dimensional space, and using the difference of a normal image and an abnormal image in the characteristic values as a detection index; the method specifically comprises the following steps:
principal component analysis is applied to one gray-scale image, F is the sum of all gray-scale values in the (2w +1) × (2w +1) size image,
the center of gravity of all pixels of the image is given by
Wherein the content of the first and second substances,
the weighted covariance matrix L of the eigenvalues λ is obtained by the following equation
L-λ·I=0
The covariance matrix is 2 × 2, a symmetric semi-positive definite matrix, so the covariance matrix L of the two eigenvalues is given by2≤λ1
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711161398.7A CN107945184B (en) | 2017-11-21 | 2017-11-21 | Surface-mounted component detection method based on color image segmentation and gradient projection positioning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711161398.7A CN107945184B (en) | 2017-11-21 | 2017-11-21 | Surface-mounted component detection method based on color image segmentation and gradient projection positioning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107945184A CN107945184A (en) | 2018-04-20 |
CN107945184B true CN107945184B (en) | 2020-10-09 |
Family
ID=61929301
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711161398.7A Active CN107945184B (en) | 2017-11-21 | 2017-11-21 | Surface-mounted component detection method based on color image segmentation and gradient projection positioning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107945184B (en) |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109283182A (en) * | 2018-08-03 | 2019-01-29 | 江苏理工学院 | A kind of detection method of battery welding point defect, apparatus and system |
CN109410212B (en) * | 2018-10-11 | 2021-06-22 | 陕西师范大学 | Interactive image segmentation method |
CN109697732B (en) * | 2018-11-27 | 2023-10-13 | 惠州市德赛电池有限公司 | Method for positioning resistance welding area |
CN109406519A (en) * | 2018-11-28 | 2019-03-01 | 广州番禺职业技术学院 | A kind of detection device and method of the special-shaped irregular solder joint of inductance element pin |
CN109725002A (en) * | 2019-01-23 | 2019-05-07 | 深圳市华星光电技术有限公司 | A kind of base board defect classification method of discrimination based on AOI |
CN109785324B (en) * | 2019-02-01 | 2020-11-27 | 佛山市南海区广工大数控装备协同创新研究院 | Large-format PCB positioning method |
CN109961067B (en) * | 2019-03-19 | 2021-05-28 | 上海望友信息科技有限公司 | Method and system for selecting optical reference point, computer storage medium and equipment |
CN110929795B (en) * | 2019-11-28 | 2022-09-13 | 桂林电子科技大学 | Method for quickly identifying and positioning welding spot of high-speed wire welding machine |
CN111553913A (en) * | 2020-05-08 | 2020-08-18 | 视睿(杭州)信息科技有限公司 | Image cutting method and device for industrial particle continuous product based on multi-channel characteristics |
CN112884838B (en) * | 2021-03-16 | 2022-11-15 | 重庆大学 | Robot autonomous positioning method |
CN113724216B (en) * | 2021-08-24 | 2023-03-21 | 广东工业大学 | Method and system for detecting wave crest welding spot defects |
CN114299086B (en) * | 2021-12-24 | 2023-05-26 | 深圳明锐理想科技有限公司 | Image segmentation processing method, electronic equipment and system for low-contrast imaging |
CN116797446A (en) * | 2022-03-17 | 2023-09-22 | 中国移动通信有限公司研究院 | Data processing method, device and equipment |
CN114418899B (en) * | 2022-03-28 | 2022-08-16 | 深圳市嘉年印务有限公司 | Self-adaptive repairing method and system for self-color printing and readable storage medium |
CN116774135B (en) * | 2023-08-22 | 2023-11-17 | 山东国研自动化有限公司 | Remote meter reading abnormity monitoring method and system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101915769A (en) * | 2010-06-29 | 2010-12-15 | 华南理工大学 | Automatic optical inspection method for printed circuit board comprising resistance element |
CN104899871A (en) * | 2015-05-15 | 2015-09-09 | 广东工业大学 | Missing solder detection method of IC element solder joints |
CN106093074A (en) * | 2016-06-16 | 2016-11-09 | 广东工业大学 | A kind of IC element solder joint detection method based on Robust Principal Component Analysis |
CN106651857A (en) * | 2017-01-04 | 2017-05-10 | 中国农业大学 | Printed circuit board patch defect detection method |
CN106780440A (en) * | 2016-11-29 | 2017-05-31 | 北京邮电大学 | Destruction circuit plate relic image automatic comparison recognition methods |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4327654B2 (en) * | 2004-05-10 | 2009-09-09 | シャープマニファクチャリングシステム株式会社 | Image processing method, image processing apparatus, image processing program, and recording medium on which image processing program is recorded |
JP2006038582A (en) * | 2004-07-26 | 2006-02-09 | Dainippon Screen Mfg Co Ltd | Detection of flaw due to regional division of image |
-
2017
- 2017-11-21 CN CN201711161398.7A patent/CN107945184B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101915769A (en) * | 2010-06-29 | 2010-12-15 | 华南理工大学 | Automatic optical inspection method for printed circuit board comprising resistance element |
CN104899871A (en) * | 2015-05-15 | 2015-09-09 | 广东工业大学 | Missing solder detection method of IC element solder joints |
CN106093074A (en) * | 2016-06-16 | 2016-11-09 | 广东工业大学 | A kind of IC element solder joint detection method based on Robust Principal Component Analysis |
CN106780440A (en) * | 2016-11-29 | 2017-05-31 | 北京邮电大学 | Destruction circuit plate relic image automatic comparison recognition methods |
CN106651857A (en) * | 2017-01-04 | 2017-05-10 | 中国农业大学 | Printed circuit board patch defect detection method |
Non-Patent Citations (4)
Title |
---|
Solder joint defect classification based on ensemble learning;Hao Wu;《Soldering and Surface Mount Technology》;20170630;1-7 * |
印刷电路板自动光学检测系统性能分析与优化;洪始良;《中国优秀硕士学位论文全文数据库信息科技辑》;20110315;第2011年卷(第3期);第2.3.4节 * |
基于模式识别技术的焊点自动光学检测算法研究;吴浩;《中国博士学位论文全文数据库信息科技辑》;20131115;第2013年卷(第11期);第1.2.1-1.2.2,2.1-2.2,3.2.2,4.2,6.2节 * |
综合边缘和颜色特征的IC类贴装器件的检测;吴晖辉 等;《光学精密工程》;20090930;第17卷(第9期);第3节 * |
Also Published As
Publication number | Publication date |
---|---|
CN107945184A (en) | 2018-04-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107945184B (en) | Surface-mounted component detection method based on color image segmentation and gradient projection positioning | |
US7680320B2 (en) | Image processing method, substrate inspection method, substrate inspection apparatus and method of generating substrate inspection data | |
CN113724216B (en) | Method and system for detecting wave crest welding spot defects | |
CN106501272B (en) | Machine vision soldering tin positioning detection system | |
CN108154502B (en) | Through hole welding spot identification method based on convolutional neural network | |
CN109472271B (en) | Printed circuit board image contour extraction method and device | |
CN106651857B (en) | A kind of printed circuit board patch defect inspection method | |
JP4736764B2 (en) | Substrate inspection device, inspection logic setting method and inspection logic setting device | |
WO2013136591A1 (en) | Image inspection method and inspection region setting method | |
US20060271226A1 (en) | Inspection standard setting device, inspection standard setting method and process inspection device | |
CN109520706B (en) | Screw hole coordinate extraction method of automobile fuse box | |
CN110473184A (en) | A kind of pcb board defect inspection method | |
CN113034488B (en) | Visual inspection method for ink-jet printed matter | |
CN107895362A (en) | A kind of machine vision method of miniature binding post quality testing | |
JP2004340832A (en) | Method and system for visual inspection of circuit board | |
CN114136975A (en) | Intelligent detection system and method for surface defects of microwave bare chip | |
CN112304957A (en) | Machine vision-based intelligent detection method and system for appearance defects | |
CN114119591A (en) | Display screen picture quality detection method | |
CN115170497A (en) | PCBA online detection platform based on AI visual detection technology | |
JP5045591B2 (en) | Method for creating area setting data for inspection area and board appearance inspection apparatus | |
CN114092448B (en) | Plug-in electrolytic capacitor mixed detection method based on deep learning | |
CN109870463B (en) | Electronic chip fault detection device | |
CN117589770A (en) | PCB patch board detection method, device, equipment and medium | |
CA2446259A1 (en) | Inspection system using dynamically obtained values and related techniques | |
JP2007033126A (en) | Substrate inspection device, parameter adjusting method thereof and parameter adjusting device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |