CN1318839C - Automated Optical Inspection Method for Defective Components on Printed Circuit Boards - Google Patents

Abstract

An automatic optical detection system for defective components on a printed circuit board comprises three units of an execution system architecture, a practical identification unit, a classification detection unit and the like; the system structure unit is established with software and hardware structures, automatically positions the printed circuit board by the hardware structure, captures images to a computer, establishes data such as a standard component, a virtual image reference template, a standard detection board, an online detection program, a component defect and the like by the software structure, can establish a component standard detection value and related environmental parameters during offline operation, and detects the component defect state on the printed circuit board during production online operation; the practical identification unit identifies and calibrates the automatically positioned printed circuit board and the reference template in the system architecture by a graph comparison method or a normalized correlation coefficient method; the classification detection unit is used for performing classification calculation on the components on the printed circuit board after identification and calibration, and accurately detecting and obtaining defective components such as missing components, skew, polarity reversal, bridging, excessive or insufficient soldering quantity and the like.

Description

Automated Optical Inspection Method for Defective Components on Printed Circuit Boards

technical field

The present invention is an automatic optical inspection system for defective components on printed circuit boards, especially suitable for assembly lines of printed circuit boards, to detect common or predictable component missing parts, skew, polarity reversal, bridging, soldering A detection system designed for defects such as too much or too little.

Background technique

In the market, whether it is a printed circuit board (Printed Circuit Board, referred to as PCB), surface mount design (Surface Mounting Design, referred to as SMD) or surface mount process (Surface Mounting Technology, referred to as SMT) testing machine, the degree of commercialization of its products is different. Already quite high. As far as the inspection machine for surface mount design (SMD) is concerned, the detectable items of each product are roughly the same, and the difference between the products lies in speed and some special functions (such as stereo vision inspection for solder joints). Here is a brief description of the characteristics of the testing machines currently used for surface mount design (SMD) in the market as follows:

1. Test items: The surface mount design (SMD) test in the market has been developed for a relatively short period of time, so so far, the test items that can be tested by various machines have not changed much; related tests in the component part Items include: missing pieces, skew, tombstoning, polarity, shifting, and more. Solder joint-related inspection items include: too much solder, too little solder, bridging, solder holes, solder pin lift, etc. In addition, in terms of IC text recognition, it is more inclined to use Optical Characteristic Verification (OCV) instead of traditional Optical Characteristic Recognition (OCR). On the one hand, the result of laser marking may be too fragmented. , On the other hand, the words printed on the IC are all predictable results, so it only needs to be able to judge whether the marking is correct or not.

2. Moving mechanism: The surface mount design (SMD) is designed with a dual-axis stage (X-YTable) on the inspection machine. It is called image visual device) that moves with the light source, or has a design load printed circuit board (PCB) mover, and also has a load charge-coupled device (CCD) that moves along the X-axis and simultaneously loads the printed circuit board (PCB) that moves along the Y-axis , but if the inspection machine focuses on speed, the movement of the printed circuit board (PCB) is the main thing.

3. Imaging mechanism: In order to increase the detection range of the surface mount design (SMD) detection device, the resolution of the charge-coupled device (CCD) used for detection is also continuously improved. In addition, the application of digital and color charge-coupled devices (CCDs) has increased significantly; color information has improved the detection effect to a certain extent, and the image quality obtained by digital charge-coupled devices (CCDs) is better than that of analog ones, and some industry players claim that The digital focus function of its products can make the detection results not be affected by the height of the components.

4. Light source: Some industry operators regard the light source system as a commercial secret and package the light source group and charge-coupled device (CCD). However, there are also very simple inspection machines that only use white ring fluorescent lamps. Light Emitting Diode (LED) light source has become the best choice for many inspection machines because of its stability, but the light source form has many variations such as ring, square, matrix and refractor. The main purpose of the light source of most inspection machines is uniform illumination, and it is rare to see switching between different light sources to obtain more various image information.

In terms of surface mount design (SMD) detectors using visible light, the development direction is mainly to detect errors that may occur in general, and requires higher speed and lower false positive rate. And SMD visual inspection does not use too complicated algorithms, but uses the commonly used basic algorithms to apply them properly.

We also know that there are existing manufacturers of SMT inspection machines and their product functions (as shown in Table 1). Among them, the automatic optical inspection system (AOI) designed by Israel is recognized as the strongest in the world:

manufacturer Product Features Orbotech (Israel technology agent) (Please refer to the description on page 4 for each detection function item) ■Function (a.b.c.d.e.f.h.i.) ■5 or 13 CCDs, Xenon flash lamp ring optical illumination ■ Position accuracy measurement of X, Y and θ of components, component displacement (OCV), Empty soldering, short circuit, empty soldering of IC feet, air bubbles (wave soldering) 2D solder paste printing disadvantages: unmelted tin, flatness of tin dots, lack of tin, loose solder paste stains, poor printing accuracy, cracks and gaps 2.5D Disadvantages of solder paste printing: solder paste printing thickness control standards TERADYNE (USA) ■Function (a.b.d.g.f.h.)

■5 CCDs, LED light sources ■Defects in plug-in pins Sony (Japan Jiantaifeng, Lindian International) The appearance inspection machine (Solder Paste Inspection Machine) has high resolution. High-speed inspection, resolution: 29um in length and 24um in width Use the image obtained by the upper side illumination and the image obtained by the lateral illumination to perform calculations to detect the printing state of the solder paste on the substrate Omron (Japan Wenhui agent) ■Function (a.d.f.) ■OCR: 32 characters can be read at a time, no need to log in the character code library ■Virtual soldering, ball soldering HIROX (the agent of Emperor Shigao in Japan) ■3D rotating inspection system (for QC laboratory) Samsung (South Korea Hongqi, Heisei agent) ■PCB detection MVP (American Huma agent) ■AOI/Automatic Optical Parts Inspection Machine CYBEROPTICS (America Reco agent) ■Function (a.b.c.d.e.f.g.h.i.) OCV Agilent (Dublin, Ireland, Taiwan Hong Kong construction agent) ■Function (a.c.d.e.f.g.h.) ■OCR, OCV ■2D tin detection: tin thickness, position, tin amount Delu (Taiwan businessman) ■Functions (a.b.c.d.e.f.g.h.) ■Anti-whitening, gold finger surface flaws, PCB surface scratches ■Board bending, board warping, software automatic correction ■Special light source and CCD Camera automatic compensation function Yutian New Technology Co., Ltd. (Taiwan businessman) ■PCB bare version detection, BGA detection Changyu Xinye Co., Ltd. ■LED brightness and wavelength detection Good Instair Electronics Industrial Co., Ltd. ■Detect assembly circuit board Dongjie Semiconductor Technology ■Check PCB, BGA defects ■Check TFT LCD panel defects

Berger Technologies ■PCB surface flaw detection, BGA detection ■IC broken pin detection system ■Micro-level precision positioning system Kaiyan Technology Co., Ltd. (Taiwan businessman) ■IN-LINE visual inspection machine, PCB testing system Functional bullet description on the right of the table: (e.) polarity (polarity) (a.) missing part (part missing) (f.) tin bridge (solder bridge) (b.) error (wrong) (g.) connection Leads floating/bend (c.) misalignment (h.) solder lack/excess (d.) tombstone (i.) seam quality ( joint quality)

Table 1: Manufacturers of SMT inspection machines and their product function comparison

Contents of the invention

The purpose of the present invention is to provide an automatic optical inspection system for defective components on a printed circuit board. The difference between the automatic optical inspection system (AOI for short) and the traditional technology is that the present invention has been planned in the initial stage of the software architecture of the system architecture. Novel and distinctive three-tier structure, including A program (A-Prog.), B program (B-Prog.) and C program (C-Prog.), this three-tier structure has greater flexibility in use, And all of them can be executed independently, and their main advantages are described as follows:

1. A program (A-Prog.): It is used to establish a database of standard components on the supply side, which can add new standard components or modify the detection parameters of standard components already established in the database. In this way, the component database maintainer on the design side can set detectable defect types for this standard component, and select detection methods for various defect types. At the same time, because of the high repeatability of the components on the printed circuit board (PCB), each component only needs to create a standard image once and store it in the standard component database of the A program (A-Prog.), and then it can be reused, saving training time .

2. B program (B-Prog.): It is a fairly intuitive operation method of component frame selection at the supply, distribution or user end to complete the setting of the detection component position and detection items, and the standard detection board can be established for each different order The test data for the online test program to test the whole batch of printed circuit boards (PCB).

3. C program (C-Prog.): It is for the supply end to operate on the production line, especially when it is often necessary to change orders to produce printed circuit boards (PCB) with different configurations (layout), the C program (C-Prog.) is in When changing the order, you only need to call the test data completed in the B program (B-Prog.) of the printed circuit board (PCB), and the whole batch test can be carried out immediately.

The present invention can adapt to changes in the industry, including:

1. In the component part; due to the continuous improvement of the semiconductor manufacturing process, the volume of the components is getting smaller and smaller, and the component placement density of the PCB is also increased. The present invention can be overcome with the improvement of the resolution of the CCD. In addition, some components use newly developed packaging technology (such as BGA, etc.), and the present invention can also be used with penetrating detection technology (such as X-ray) or a multi-lens stereo vision solution to solve the problem.

2. In the PCB industry: the current main development direction of the present invention is to detect motherboards, interface cards and other products used in PCs. If such products develop to a saturated stagnation state in the future, the key detection technology can still be applied to many other PCB-based products. Architecture of emerging products, such as mobile phones, PDAs and so on.

3. In the chip packaging industry: the extension direction of the present invention is the detection capability of a smaller scale. Using a high-magnification lens and a line-scanning camera to capture images, combined with the mature algorithm developed in the previous stage, can shorten the overall development time and quickly establish a suitable Inspection machine for chip packaging industry (BGA, wire bonding).

In addition, the development goal of the present invention is to replace manual visual inspection and enhance the quality and speed of inspection. The most important key to the success of an automatic optical inspection system lies in the development of detection algorithms. The purpose of the detection algorithm is to extract representative feature pointer values from the detection image and set appropriate judgment rules for the detection items of different components to determine whether the detection target is a good product or a defective product. A good detection algorithm should not only have a good detection effect, but also pursue the lowest computational complexity; the lower the complexity, the faster the calculation speed, and the industrial practicability of this algorithm is also greatly improved. The detection algorithm used in the present invention is to first observe the component features in the detection image, combine and apply them with simple algorithms, try to quantify the image features into pointer values, and then experiment with a certain number of detection images (including good and bad products) After that, determine the good/defective judgment rule.

In order to clearly obtain the image characteristics of the components, there must be an appropriate light source system to assist in imaging. The function of the light source system is not only to provide enough illumination to obtain images, but also to further highlight the characteristics of components. The types of light source systems are very diverse, and even a single type of light source can have many forms of changes; the AOI system needs a light source system that can cooperate with various detection algorithms in order to have a good detection effect. The present invention will simultaneously design a set of mutually matching and applicable detection algorithms and a program-controlled light source system to develop a detection machine suitable for detecting defects of various components.

The scope of application of the present invention and field include:

(1).CCD precise mobile positioning

(2). Detection algorithm and automatic matching control of light source system

(3). Various defect detection of components on PCB such as chip resistor (Resistor), chip capacitor (Capacitor), small external pin integrated circuit (SOP) and square flat package integrated circuit (QFP), as shown in Table 2 List:

(The empty part indicates that this defect does not occur in the manufacturing process)

Table 2: It is the component type and its defect items on the detectable PCB of the present invention

The present invention will be described in detail below in conjunction with the accompanying drawings.

Description of drawings

Fig. 1 is a schematic diagram of the hardware architecture of the detection system of the present invention;

Fig. 2 is a schematic diagram of the software architecture of the detection system of the present invention;

Fig. 3 is the flow chart of the present invention setting up the A program of standard component in software framework;

Fig. 4 is that the present invention sets up the B1 program flowchart of the reference template of PCB virtual CCD in software architecture;

Fig. 5 is the flowchart that the present invention sets up the B program of detecting PCB data in software architecture;

Fig. 6 is the flow chart of online detection C program in the software framework of the present invention;

Fig. 7 is the D program flow chart of inspecting PCB defect data in the software framework of the present invention;

Fig. 8 is a flow chart of online automatic positioning in the practice identification of the present invention;

Fig. 9 is a schematic diagram of an off-line virtual reference template image in the practice identification of the present invention;

Fig. 10 is a conceptual flow chart of setting standard characteristic values in the detection method of the present invention;

Fig. 11 is an orthographic projection view of IC pin bridging defects of the present invention;

Fig. 12 is a diagram of the processing result of orthographic projection of components according to the present invention.

Detailed ways

The present invention has developed an automatic optical inspection system suitable for defective components on the PCB assembly line. The design focus is on the identification of common or predictable defects on the production line, and the three units of the inspection system architecture, practical identification, and classification detection methods will be used here. , explained one by one as follows:

(1) Detection system architecture unit:

The detection system architecture design is divided into hardware architecture and software architecture control design, which will be discussed in two implementation parts: offline operation and online operation, among which:

The off-line operation refers to the fact that the detection system does not affect the productivity of the production line, and under cost considerations, only one PC can be used as the operating environment to provide the required information; and to establish the standard detection value of the component and the related environment Parameters are the main function, so it is also called training (Training) job.

The on-line operation means that the inspection system is based on assisting the quality control on the production line, and its main function is to detect the defect status of the components to be tested, so it is also called inspection (Inspection) operation.

The hardware architecture (as shown in Figure 1); includes a two-axis stage 10 (X-Y Table), which is a load image visual device 11 (CCD) and an LED ring light source 12 (Ring LED Light), and is controlled by a drive controller 13 (Driver Controller) control moves to the user-designated printed circuit board 16 (PCB) placement position; Wherein,

The image visual device 11 (CCD) is used for image acquisition, and converts analog image signals into digital image signals through an image capture card 14 (Frame Grabber).

The LED ring-shaped light source 12 (Ring LED Light): is controlled by a digital conversion analog signal controller 15 (Digital/Analog Converter). According to the requirements of different testing items, it provides an appropriate light source lighting method, and various light sources can be generated by program control. combination.

Above-mentioned drive controller 13, image acquisition card (14 and digital conversion analog signal controller 15 all can be controlled by a personal computer 17 as operating platform.

According to the above hardware structure, during the offline training phase, the standard printed circuit board (PCB) is manually loaded on the testing platform; during the online testing phase, when a PCB is tested, the system will issue a signal to replace the PCB and stop all Check the action until a PCB to be tested is replaced, continue to issue testing instructions, and then start testing.

The software architecture (as shown in Figure 2) is to consider the offline operation of practical identification in the actions of PCB inspection system, such as establishing standard components and establishing and inspecting a piece of PCB data. In order not to make the offline training action affect the operation of the production line, this Invented and designed the concept of virtual CCD (Virtual CCD) to assist and improve the operation of software offline operations. include:

I. A program (A-Prog.) for establishing standard components: the main function is to provide the design end with establishing a standard component database (A01 to A05), which can add new standard components or modify the detection parameters of established standard components in the database.

The maintainer of the component database on the design side can set the defect types 21 of the standard components in the above-mentioned standard component database [(A01 to A05], *.cop), select the appropriate detection algorithm 22 for each defect type, and establish and obtain Image data 23 of the standard component, and establish a data stream such as parameters 24 required for detection (as shown in FIG. 3 ), so as to store the standard component data 2 (*.cop) (that is, the characteristic value of the standard component).

II. B1 program (B1-Prog) for establishing PCB virtual CCD data B10 (as shown in Figure 2): the main function is to create a reference template (Reference Template) data B15 based on the concept of virtual CCD to store the entire standard PCB information , including setting the PCB information B11, setting the data of moving the biaxial stage 10 to the fixed position B12, selecting the data of the image combining method B13, etc. The combined image B14 can be supplied to the offline operation program to simulate the action of real CCD image capture (As shown in Figure 4).

III. Establish the B program for testing PCB data [as shown in Figure 2, (B-Prog.)]: the main function is to create a standard test board data B2 and produce training data B20 for online testing at the distribution or user end The program performs the entire batch test. The user can select the component to be tested B21 on the standard test board from the above-mentioned standard component database (A01 to A05) and read the virtual CCD data B10 (as shown in Figure 2), and establish automatic positioning data B23 and establish or select the standard As for the component B24, select the component detection item B22 at the same time, and move the biaxial stage to record the component position B25 (as shown in FIG. 5 ). With this program B, it is convenient to quickly adjust and test items for PCBs produced by new orders.

IV. C program for online detection [as shown in Figure 2, (C-Prog.)]: the main function is to use the training data B20 (*.trn) file of the B program (as shown in Figure 5) to process the entire batch of PCBs to be tested For the inspection of C10, obtain the inspection result data C11 (*.inp) and defect data C12 and C15 (*.fut) of each PCB.

Also refer to Figure 6, the implementation process of the C program can be further known, including first reading the above training data B20, and loading the entire batch of PCBs to be tested C10, so that each PCB is automatically positioned at C16, and then the PCBs to be tested are detected C17 , in order to read each of the above PCB inspection result data C11(*.inp) and defect data C12(*.fut).

V. Program D for viewing PCB defect data [as shown in Figure 2, (D-Prog,)]: the main function is to use the program D to point out the defect (D0) of the defect data C12 of each PCB generated by the above C program ), including defect component position D01 and defect category D02 for repair (as shown in FIG. 7 ).

In FIG. 7 , it can be seen that when program D is executed, the above-mentioned virtual CCD data B10 and defect data C12 files will be read, and the defective component position D01 and defect type D02 on each PCB will be revealed on the computer display 18 .

(2) Practice identification unit:

In the above-mentioned software architecture of the present invention, considering the reference templates established by the automatic positioning of the PCB on the production line and the offline virtual CCD respectively, both of them adopt the Pattern Matching method of graphic comparison or the Normalized correlation coefficient method (Normalized correlation coefficient), as follows:

(I) On-Line PCB automatic positioning;

When inspecting each PCB on the line, due to the influence of the external factors of the conveyor, the PCB cannot be correctly positioned every time, and its subsequent inspection operations may be affected by this, resulting in misjudgments. Therefore, this automatic positioning method is designed.

The timing and process of using PCB to automatically locate offline training operations and online detection operations (as shown in Figure 8) are explained as follows:

[a]. When using program B to establish a standard PCB in offline training (Training), first select and record the PCB positioning feature B3, and record the feature-related position B4 of the components on the PCB. The feature relative position (B4) includes the position of the image relative to the biaxial stage 10, and the position of the positioning feature B3 relative to the image.

[b]. Before using the C program to detect each PCB, it will automatically calculate the X-axis or (and) Y-axis offset of the dual-axis carrier 10 caused by the conveyor or positioning mechanism according to the previously set positioning feature B3. , and when the biaxial carrier 10 is moved, correction and calibration positioning is performed for the X or (and) Y axis offset of the PCB.

The above-mentioned X or (and) Y-axis offset is from the positioning feature B3 of the B program, using the Pattern Matching method to search C2, find out the possible position on the PCB to be tested, and compare the position coordinates C3 , that is to compare the possible position on the PCB to be tested with the feature-related position B4 recorded during the offline training operation (Training); when a deviation occurs, the difference between the two is the deviation of the X or (and) Y axis At this time, the CCD position C4 should be corrected to the correct position, so as to move the CCD C5 to the top of the PCB to detect C17.

(II) Build a PCB reference template with an Off-Line virtual CCD;

The virtual CCD provides a reference template B15 with no distortion and the same magnification as the real CCD. In the present invention, the CCD magnification is 640×480/23×17 (pixels/mm2), and the entire image of a PCB (23×20cm ² ) is about 6400×6300 pixels (pixels). The virtual CCD function is to simulate the real CCD so that it can establish the entire PCB reference template B15. The virtual CCD is currently produced by the B1 program (as shown in FIG. 4 ) in the present invention, and the B program and the D program are used by off-line operation software. The image generation of the reference template B15 established by the virtual CCD, regardless of the movement of the biaxial stage 10 on the X-axis or the Y-axis, the general processing steps are as follows (as shown in Figure 7):

(a) Move the CCD on the biaxial stage by a fixed distance 40 to generate the first image before the movement and the second image after the movement; the fixed distance 40 is about 1/3 of the length and width of the image (expected overlap area).

(b) The same CCD position only fixedly uses the same light source as the judgment processing image.

(c) The image overlapping area 43 is an area for numerical analysis.

(d) Obtain the overlapping area 43 between the first image 41 and the second image 42 through calculation, and cut it off from the second image 42 .

To remove overlapping images, the present invention utilizes the overlapping region 43 on the first image 41 as an identification template for pattern matching (Pattern Matching) method, and searches for a similar region on the second image 42 to obtain the pattern from the second image 42 upper resection.

(3), classification detection unit:

In the present invention, the classification detection method can also be called a defect classification algorithm, which is mainly divided into two parts: offline (or training) operation and online (or detection) operation.

When working offline, the characteristic value 50 of the standard component is first extracted from the standard component data of the A program, and the detection frame 51 is set to facilitate the online operation to perform the test 52, and compare or compare the standard component and the component to be tested. Relevant characteristic values, and the qualified component characteristic values are stored 5 (as shown in FIG. 10 ).

1. Capacitor missing and skew detection processing model;

When the grayscale image of the missing capacitor on the PCB is taken, there are two situations: (A) when the capacitor is missing on the PCB, the component position does not contain a circuit; when the capacitor is missing on the 2PCB, the vertical center of the component contains the circuit through . Capacitors exist as standard components on PCBs.

In practical testing, the components on the PCB are allowed to be slightly offset, and are not flaws. Therefore, the present invention intends to use the Pattern Matching method to obtain the correct component position in the first stage, and then propose an algorithm to determine whether there is a missing part (or wrong piece).

The first stage - pattern matching (Pattern Matching) method - to obtain the correct component position. The acceptance (Acceptance) threshold value setting of the graphic comparison method does not have a certain standard, so the present invention intends to adopt a lower acceptance (Acceptance) threshold value first, so that the result of the graphic comparison method includes (A) correct components, ( B) Misjudgment of missing parts and (C) Misjudgment of wrong parts, etc., and then classify and screen based on the characteristic differences between standard components and misjudged blocks. In the second stage, an algorithm is proposed to determine whether the capacitor is missing. For the convenience of explanation, it is referred to as the Black Percentage method hereinafter.

The second stage - black block ratio (Black Percentage) method - to determine whether the capacitor is missing. This method intends to use appropriate light source lighting to cause the difference between the characteristics of the capacitor itself and the image characteristics of missing or wrong parts on the printed circuit board (PCB). Take the similar components found by graphic comparison as an example. Observe the gray scale distribution diagram of the component block (as shown in Table 3). Table 3 contains a dotted line, which can be clearly distinguished. The gray scale distribution diagram of standard component 01 is in The left side of the dotted line does not contain any pixel (pixel), and the rest are the gray scale distribution diagrams of misjudgment components 02, 03, 04. The left side of the dotted line contains image pixels.

Table 3: Gray scale distribution chart of misjudged components and standard components

2. Bridge (short circuit) detection processing model;

Bridging defects only appear on components with IC pins. Figure 13 is an enlarged view of a quad flat package (QFP) IC pin with bridging defects.

In practice, due to the detection of the tin foot area, it may be caused by the offset when the detection area is manually selected during offline training, and the PCB is slightly offset within the tolerance range during online inspection operations, or the components are slightly offset within the tolerance range; If an incorrect action has occurred when the detection area is framed, the subsequent algorithm will not be able to correctly judge the detection result due to the offset of the detection starting point and the position of the judgment detection point.

The present invention intends to solve the problem of positioning by searching for the IC pin in the detection area, and then cooperate with the Image Projection method to perform detection. The method is described as follows:

(A) Expansion of the detection area (Inflate region): Due to the number of IC pins to be detected, it is set for offline training operations, and the location of the detection area needs to be informed manually. In order to avoid human-factor deviation, the user determines the detection When changing the area position, increase the size of the original detection area without changing the center position of the detection area.

(B) Find Stripe: After the IC feet are binarized, it presents black and white stripe features, and uses the Find Stripe method to locate them. Since the bottom board circuit will be hidden between the IC pins, the circuit will be affected by the height of the IC pins on both sides, resulting in a decrease in the brightness of the received light source, so it can be eliminated by using a binarization method; but in some IC pin rows, For example, the side of the first IC pin may also contain circuit lines on the bottom board. In order to maintain the distinctive features of the IC pin, it is not intended to completely eliminate it by binarization, because the brightness of the light source received by the first IC pin is higher than that of the IC. The backplane circuit between the feet is high.

Considering practical identification, the Find Stripe method will use "black-white-black" stripe marks to estimate the position where the second IC pin appears in the expanded detection area (Inflate region). Stripe) compares the starting point of the area to find out the correct position of the second IC pin, and the positioning position of the first IC pin can be deduced from the known width of the IC pin, and the correct detection area can be obtained.

(C) Image Projection (Image Projection) method; its algorithm is as follows:

(a) Frame the detection area;

(b) image binarization processing;

(c) Image forward projection processing to obtain grayscale accumulation;

(d) Numerical analysis: Set the detection starting point, the spacing of the IC pins, the width of the IC pins, and the number of IC pins, and then the correct position of the IC pins can be calculated. If the cumulative value of the binary gray scale between the IC pins is too high, Then it is determined that a bridging defect occurs at this place (as shown in FIG. 11 ).

3. Polarity reverse detection processing model;

In the PCB components (including SOP and QFP) discussed in the present invention, the polarity representation is divided into strip polarity and hole polarity. The polarity reversal phenomenon does not have appearance defects on the components. The main reason is that the component is placed in the reverse position and the function of the component is lost. Therefore, the polarity direction will be marked on the component, and this mark can be used to find out the position of the polarity . This section describes the detection model of the strip polarity and the detection model of the hole polarity respectively.

(A) Strip polarity detection model;

The body of the component to be tested mainly contains two kinds of information, the serial number and the polarity of the component. First, the binarization process is used to separate this information from the background. Since the component serial number and the polarity have the same gray scale, it is necessary to further separate these two types of information; the position of all strip polarity marks is at the end of the component body, so according to this position-related characteristic, the first step is to detect The frame sets the position of the component body, and the image in the detection frame is processed by the orthographic projection method to obtain the result in Figure 12. The orthographic projection processing can convert the two-dimensional image data into a one-dimensional array of numerical data, and further use this numerical data to obtain The position of its maximum value can be used to know the position of the polarity strip on the component.

(B) Hole polarity detection model;

Most QFP components have a concave circular hole to indicate the polarity position, and the degree of depression of the polar circular hole and the size of the hole will vary with different types of QFP components.

When the polar hole is irradiated by side light, a white ring-shaped aperture will appear around the hole. Therefore, the present invention intends to use the reflection characteristics formed by the side light, and use the morphology processing method (Morphology) in image processing to strengthen the required The halo information can further determine the position of the polar hole. The method is described as follows:

(a) Binarization processing: For polarity detection, the information that needs to be obtained from the image is the presence or absence of polar holes, so a grayscale image can first use binarization to convert unnecessary gray Step value removed. After binarization, the annular aperture can be slightly separated from the background, but there is still some noise mixed in it, which needs to be further eliminated by morphological processing operations.

(b) Morphological processing: In the application of morphological processing, we often design a suitable matrix and apply specific algorithms to the graphics to be processed to eliminate or strengthen certain signals. The present invention intends to use image erosion (Erosion) algorithm to eliminate noise, and image expansion (Dilation) algorithm to enhance signal. This successfully preserves most of the ring aperture signal and eliminates major noise.

(c) Particle processing: due to the smooth surface of the QFP component body, only the concave polar hole position will reflect light when taking an image with side light. The present invention intends to use particle processing (Blob process) to calculate the white point particles in the image The number of pixels (pixels) occupied, that is, the particle area (Blob Area), is used as the basis for detecting whether there are polar holes in the detection area.

(d) Numerical analysis: The particle area can be used as the detection parameter value of the standard part. However, in addition to the processed ring aperture area, the particle area still includes noise that has not been completely removed. Therefore, in the design of the detection parameter value of the circular hole type polarity detection, the calculated total particle area should be multiplied by a weight to filter out the particle area occupied by noise, and the setting of the weight value requires further experiments. Decide. When the area of the particles obtained after the test is processed is smaller than the detection parameter value, it can be determined that it is a defect with reversed polarity.

4. The detection model of solder volume;

The QFP component on the PCB may produce secondary defects such as too much tin or too little tin during the SMT process. Take the grayscale image of the tin solder joint with normal QFP tin content on the PCB. The detection method is briefly described as follows:

(A) Set the pitch of tin pins D and the width of tin pins W, and use the method of image cutting to make a QFP component containing N tin pins, starting from S, every (D+W)×i, (0≤i< N, i∈integer) is to cut out the solder joint image.

(B) Calculate the following parameter values by using the solder joint image cut out in item (A) above.

Let U _i be the average value of the gray scale of solder joints under the upper light source environment, (0≤i<N, i∈integer)

L _i is the average value of the gray scale of solder joints in the lower light source environment, (0≤i<N, i∈integer)

${\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; :</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}$ ${\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; :</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}$

${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; :</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="u\; i"\; filenames="C0215382900211.png"\; state="\{\{state\}\}">u</figure-callout></span><figure-callout\; id="2"\; label="u\; i"\; filenames="C0215382900211.png"\; state="\{\{state\}\}">\; </figure-callout>}}_{\mathrm{<span\; class="notranslate">\; </span>}}^{\mathrm{<span\; class="notranslate">i</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$ ${\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; :</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="L\; i"\; filenames="C0215382900211.png"\; state="\{\{state\}\}">L</figure-callout></span><figure-callout\; id="2"\; label="L\; i"\; filenames="C0215382900211.png"\; state="\{\{state\}\}">\; </figure-callout>}}_{\mathrm{<span\; class="notranslate">\; </span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

(C) Use the upper and lower light sources to calculate the average gray scale of solder joints, and then use the classification method (Classfication) in visual processing to separate out the normal tin content, excessive tin content and too little tin solder joints .

Now set forth the application example of the present invention as follows:

The detection system developed by the present invention considers the needs on the actual production line. The program design includes a three-layer structure (A, B and C programs), a virtual CCD (B1 program) and a test result report output (D program). This example will be based on This framework and the aforementioned detection method fully illustrate the operation process of this system with a PCB example. In this example, the components to be tested include chip resistors, chip capacitors and quad flat package integrated circuits (QFP), a total of 78 components.

(A) Establish the A program (A-Prog.) of the standard component: firstly establish the standard component database in the A program, the steps are to move the CCD to the standard component to be detected, frame the standard component image, and set each component Detection items and their detection algorithms.

(B) B1 program (B1-Prog.) for establishing PCB virtual CCD data: set the PCB length 225mm, width 230mm and the moving distance required by the CCD to capture images (as shown in the B1-Prog operation screen in Figure 34), During execution, the CCD will automatically capture the sub-images of the PCB in sequence, and combine all the sub-images into a complete standard PCB information to create a reference template (Reference Template) for the offline operation program to simulate the real CCD capture image action. Combine the completed full PCB image.

(C) Establish the B program (B-Prog.) for detecting PCB data: use the image combined by B1-Prog to browse the PCB offline, select the position of the component on the PCB, and increase the detection range slightly for easy search when selecting the frame Component location, select the corresponding standard component from the standard component database completed by A-Prog, and select the item to be tested.

(D) C program (C-Prog.) of the online inspection program: use the standard inspection version data files generated by B-Prog to inspect the entire batch of PCBs to be tested, and output the inspection data and defects of each PCB information.

(E) D program (D-Prog.) for viewing PCB defect data: use the defect information generated by C-Prog to indicate the location and defect type of defective components. .

The automatic optical inspection system (AOI) provided by the present invention has the following advantages:

1. Consistency of quality: The machine does not allow products with poor quality to pass the factory because of inconsistent quality standards caused by human factors, such as mental state, laziness, negligence, and fatigue.

2. Improve judgment ability: Some defects, such as SMT empty soldering, tin bridge, tin beads, etc., cannot be found out with the naked eye. The detection time of the AOI system is not only short, but also has a high judgment for such defects, and there will be no missed.

3. Real-time response: AOI cooperates with the function of statistical process control (SPC), which can quickly feed back the relevant information of the collected defective products, discover process problems in real time and adjust machine parameters to maintain process stability and reduce defects caused by defective products. loss.

4. Reduce inadvertent damage: AOI system is a non-contact detection system, which can reduce or eliminate the chance of hand touching the product, so as to avoid damage to the product such as static electricity and handprint.

In summary, the development of the automatic optical inspection system (AOI) of defective components on the printed circuit board of the present invention can not only reduce the production cost, improve the detection speed and reduce the misjudgment rate, but also can achieve the level of full inspection, and its efficiency The consistency of performance, performance and quality is far superior to traditional manual inspection. In the current situation, customers gradually regard AOI as the basic requirement of product quality. Therefore, if the domestic industry wants to increase product competitiveness, it is believed that it is necessary to develop AOI and quickly introduce it. inevitable trend.

Claims (10)

1. An automatic optical inspection method for defective components on a printed circuit board, comprising:

(1) procedure a to build a database of standard components: after moving the image vision device CCD to the standard component to be detected, framing and capturing the standard component image, and setting the detection items and detection algorithms of all the components;

(2) b1 procedure for creating virtual image viewer CCD data of printed circuit board PCB: setting the information of the PCB, including the length, the width and the moving distance required by the image vision device CCD to shoot the image, automatically shooting the subimages of the PCB in sequence by the image vision device CCD during execution, combining all the subimages into complete standard PCB information to establish a reference template, and combining the complete PCB image;

(3) establishing a B procedure for detecting PCB data of the printed circuit board: browsing the PCB off-line by using the combined image, selecting the position of the component on the PCB, selecting the corresponding standard component from the standard component database, and selecting the item to be detected;

(4) procedure C for on-line detection: using the generated standard detection board data file to detect the whole batch of PCB to be detected and generating the detection data and flaw information of each PCB;

(5) d, inspecting the PCB flaw data of the printed circuit board: utilizing the generated defect information to indicate the position and the defect type of the defective component;

the method adopts an automatic optical detection system comprising a system architecture unit, a practical identification unit and a classification detection unit, wherein the system architecture unit is provided with a hardware architecture and a software architecture, so that a user can perform practical identification and classification detection during off-line operation and on-line operation, the off-line operation is to establish a standard detection value and relevant environmental parameters of a component to be detected on the printed circuit board, and the on-line operation is to perform detection on the defect state of the component to be detected on the printed circuit board.

2. The method of claim 1, wherein the standard cell database of the A program is used for user to set the defect type of the detectable standard cell.

3. The method of claim 1 wherein the standard cell image data is obtained by setting the defect type of the standard cell and selecting an appropriate inspection algorithm.

4. The method of claim 1, wherein the B1 program sets distance data for moving the biaxial stage to a fixed position.

5. The method of claim 1, wherein the B1 program is used for offline operations in performing real estate identification to create a reference template, the B1 program sets the data for selecting image combinations by:

(a) moving an image vision device on a double-shaft carrying platform for a fixed distance to generate a first image before moving and a second image after moving;

(b) the same image vision device only uses the same LED annular light source as the basis for judging the image;

(c) the image overlapping area is a numerical analysis area;

(d) the overlapping region of the first image and the second image is obtained by calculation, and is cut out from the second image.

6. The method of claim 6, wherein the fixed distance is 1/3 image length.

7. The method of claim 6, wherein the overlap area of the first image is used as a recognition template for pattern matching when the overlap image is cut, and a similar area is searched for in the second image to cut the second image.

8. The method of claim 1, wherein the B program uses a pattern matching method in the real estate identifying unit to first frame and record the positioning characteristics of the PCB and record the relative positions of the characteristics of the components on the PCB when selecting to read the standard component data to be tested, thereby finding the X-axis or Y-axis offset of the dual-axis stage, or the X-axis and Y-axis offsets, so that the offset can be corrected when moving the dual-axis stage to calibrate and establish the automatic positioning data of the PCB.

9. The method of claim 10, wherein the relative positions of features of the component on the printed circuit board include the position of the image relative to the biaxial stage and the position of the locating feature relative to the image.

10. The method of claim 1, wherein the B program reads the dummy CCD data of the B1 program.

Images