CN117092113A - Camera mould welding quality detection device and system thereof - Google Patents

Abstract

The application relates to the technical field of welding quality detection, in particular to a welding quality detection device and a welding quality detection system for a camera die. According to the application, the camera is fixedly connected with the third motor through the connecting plate, the third motor is in threaded connection with the first fixed plate through the connecting rod, so that the camera can be conveniently detached, installed and maintained, and the camera can be prevented from shaking during rotation by connecting the connecting plate with the second fixed plate through the second limiting rod.

Description

A camera mold welding quality inspection device and its system

Technical field

The invention relates to the technical field of welding quality detection, specifically a camera mold welding quality detection device and its system.

Background technique

With the rapid development of the smart device industry, the demand for camera modules CCM is increasing rapidly. At the same time, equipment specifications are trending toward ultra-thin and high-pixel dimensions, resulting in continued miniaturization of equipment camera modules. High-pixel camera modules are mainly composed of Lens (lens), VCM (voice coil motor), base bracket, SensorIC (image sensor) and PCB/FPC (circuit board) and other components. In the process of producing CCM, the welding between components is an important part. The number of VCM welding pins of CCM products ranges from 2 to 20, and they are connected to the PCB through laser solder ball spray welding and other methods. During the welding process, welding machine errors and other reasons may cause welding defects in the camera module, which in turn affects the reliability of the integrated circuits in the module, such as missing solders, missing solders, solder joint adhesion (bridging), etc. If such defective modules are applied to electronic products, it will cause quality problems in the product and even cause the entire product to be scrapped. Therefore, effective detection of CCM welding defects is crucial to improving the quality and reliability of electronic products.

Machine vision inspection technology refers to a non-contact inspection method that uses industrial cameras to collect images of products to be inspected during product quality inspection, and then uses image processing technology to determine the quality problems of the products to be inspected. It can be used very well in automated production lines of electronic components. Realize pre-weld and post-weld inspection. Due to the interference of local high light reflection, complex lighting environment, camera distortion, background interference and other factors on the CCM welding surface to be tested, visual inspection of CCM surface welding defects has become a challenging task. Camera module solder joints and ordinary PCB solder joints Differently, there are unmarked spot fillet welding and multi-side connected welding, which makes the existing detection technology unable to be well applied to actual CCM production lines. Existing quality inspection solutions have the disadvantages of high cost, high misjudgment rate, and low efficiency, and rely on secondary manual visual inspection to ultimately determine defects. However, the manual inspection process is easily affected by external factors and the proficiency of inspection workers. Moreover, it is difficult to ensure product quality consistency and cannot meet the detection of camera module welding defects that require high detection accuracy and recognition speed.

In response to the above problems, we provide a camera mold welding quality inspection device and its system.

Contents of the invention

The purpose of the present invention is to provide a camera mold welding quality detection device and its system to solve the problems raised in the above background technology.

In order to solve the above technical problems, the present invention adopts the following technical solutions:

A camera mold welding quality inspection device and its system, including a base (1), a work chamber (2) provided above the base (1), and a storage drawer (3) provided inside the work chamber (2) ), a handle (4) is provided above the storage drawer (3), a detection structure (5) is provided above the work compartment (2), and a bracket (5) is provided on the left side of the detection structure (5) 6). A support rod (7) is provided above the bracket (6). A buzzer (9) is provided inside the support rod (7). A machine is provided below the buzzer (9). Vision structure (10), a lighting lamp (8) is provided on the left side of the machine vision structure (10), and a control unit (11) is provided on the right side of the buzzer (9).

Preferably, the detection structure (5) includes a first motor (501), a first threaded barrel (502) is provided above the first motor (501), and a first threaded barrel (502) is provided inside. There is a first threaded rod (503), the first threaded barrel (502) is threadedly connected to the first threaded rod (503), and a placement plate (504) is provided above the first threaded rod (503). A suction cup (505) is provided above the placing plate (504), a second threaded barrel (506) is provided inside the bracket (6), and a second threaded rod (506) is provided inside the second threaded barrel (506). 507), the second threaded barrel (506) is threadedly connected to the second threaded rod (507), and a driving motor (508) is provided at the outer end of the second threaded barrel (506).

Preferably, a mounting plate (509) is provided on the right side of the second threaded rod (507), and a second motor (510) is provided on the right side of the mounting plate (509). The second motor (510) A clamping plate (511) is provided on the right side of the clamping plate (511), and an anti-slip pad (512) is provided on the right side of the clamping plate (511). The anti-slip pad (512) is adhesively connected to the clamping plate (511), so A first limiting rod (513) is provided below the second motor (510).

Preferably, the machine vision structure (10) includes a first fixed plate (1001), a connecting rod (1002) is provided below the first fixed plate (1001), and a connecting rod (1002) is provided below the connecting rod (1002). The second fixing plate (1003), the connecting rod (1002) is connected to the first fixing plate (1001) through threads.

Preferably, a second limiting rod (1004) is provided below the second fixed plate (1003), and a third motor (1005) is provided on the left side of the second limiting rod (1004). The third motor (1005) is threadedly connected to the second fixing plate (1003), a connecting plate (1006) is provided below the third motor (1005), and a camera (1007) is provided below the connecting plate (1006).

A camera mold welding quality inspection system, including:

The image acquisition unit acquires the image sequence of the camera module to be detected;

The image preprocessing unit performs welding area positioning and sequence image fusion processing on the acquired image sequence to be detected to obtain high-quality welding area images;

The welding quality inspection unit performs preliminary quality inspection of the welding quality of the camera module on a module basis and neural network secondary defect determination of the welding quality of the camera module on a solder point basis;

The display unit displays the quality inspection results of the camera module welding and marks the location and type of welding defects.

Preferably, the image sequence is collected by an optical imaging unit, wherein the optical imaging unit includes:

The image acquisition unit collects the camera module image sequence, and the image sequence is the camera module image collected under different exposure times; the lighting unit has a preset fixed angle between the camera module welding area and the image acquisition equipment, and is used to acquire the solder and solder Imaging of welding areas with high disc contrast;

The image preprocessing unit includes:

The welding area positioning unit is used to automatically and accurately position the welding area of the camera module and obtain the welding area position information of each image in the collected camera module image sequence;

The image sequence fusion unit is used to fuse the acquired welding area image sequences to obtain a welding area image with uniform exposure and rich detailed information;

The welding area positioning unit includes:

The template area selection component is used to use the principle that the welding area is fixed relative to the camera module's spatial position, and uses the template matching method to select the template image for area matching and positioning;

The similarity comparison component is used to compare and calculate the similarity between the template image and the search sub-image in the original image. The position of the maximum similarity sub-image is the target position;

The search acceleration component is used to accelerate the speed of matching and positioning. It adopts a pyramid hierarchical search strategy and presets the starting search coordinate point. It first performs rough matching and then precise matching;

The welding area cropping component is used to obtain the matching positioning coordinates, and by presetting the length and width of the welding area, crop the welding area of each image in the image sequence and store it as a welding area image with a smaller size and less background area. ;

The image sequence fusion unit includes:

The image alignment component is used to align the corresponding pixel positions of the acquired welding area image sequence;

The image fusion component is used to calculate the weight value of each image pixel based on the two indicators of exposure weight and local detail retention weight to obtain a comprehensive weight map, thereby obtaining a high-quality fused welding area image;

The quality inspection unit includes:

The preliminary quality inspection unit is used to quickly determine whether the welding quality of the camera module is qualified and output the results;

The secondary defect detection unit is used to further conduct secondary quality inspections on modules that are judged to be unqualified by the preliminary quality inspection unit, eliminate qualified modules that are mistakenly judged to be defective, and check the types and locations of defects in unqualified modules. Identify and label.

Preferably, the preliminary quality detection unit includes:

Image enhancement component, used to improve the interference of flux impurities in the welding area, enhance edge contour information, and improve the accuracy of subsequent feature extraction;

The feature extraction component is used to effectively characterize the characteristics of various types of welding defects. The welding area processed by the image enhancement component uses an edge detection algorithm to identify the edge contour of the solder joint, and obtains the solder joint contour through hole filling and morphological operations, and then calculates the solder joint. Point outline features;

The preliminary classification component is used to initially classify qualified camera modules and unqualified camera modules after extracting the welding defect characteristics of the camera module. It uses a machine learning method with adjustable output categories to make quality judgments so that defective camera modules can be identified. All groups are detected.

Preferably, the secondary defect detection unit includes:

The data set construction component is used to build a data set of the solder joints of the camera module to be detected. It uses the preset pad positions in the template area to crop the welding area, obtains the target image of each solder joint, and classifies the solder joint defects according to the type. And divided into sample training set and sample test set according to a certain proportion;

The model training component is used to train a neural network model for identifying solder joint defect types, and uses the sample training set to perform parameter iterative optimization until the training model inputs the sample test set and reaches the expected prediction accuracy to stop training and obtain the solder joint defect classification. Model;

The defect detection component is used to determine whether there are defects in the input solder joint image and what type of defects there are.

Preferably, the display unit includes:

The first display screen is used to display the current inspection camera module image and the welding area image, mark whether it is qualified, and check the location and type of defective solder joints;

The second display screen is used to display parameter settings for setting adjustable parameters involved in the machine vision-based camera module welding defect detection device and method.

It can be seen from the above description that the technical problems to be solved by the present application can certainly be solved through the above technical solutions of the present application.

At the same time, through the above technical solutions, the present invention at least has the following beneficial effects:

The invention obtains the image sequence of the camera module to be detected through the optical imaging unit; performs image preprocessing on the image sequence to obtain high-quality, uniformly exposed welding area images; performs quality judgment on the camera module to be detected through the quality detection unit, and identifies The location and type of solder joints with welding defects, the detection method can be better applied to actual production lines where the number of defective samples is small and the number of qualified modules is much larger than the defective modules. It meets the actual production requirements of high precision and efficiency and avoids manual inspection. Insufficient inspection.

Description of the drawings

Figure 1 is a schematic front view of the structure of the present invention;

Figure 2 is a front cross-sectional structural schematic diagram of the present invention;

Figure 3 is a schematic diagram of the detection structure of the present invention;

Figure 4 is a schematic structural diagram of position A in Figure 3 of the present invention;

Figure 5 is a schematic diagram of the machine vision structure of the present invention;

Figure 6 is a schematic diagram of the overall process in the embodiment of the present invention;

Figure 7 is a flow chart of the welding area positioning unit in the embodiment of the present invention;

Figure 8 is a flow chart of the image sequence fusion unit in the embodiment of the present invention;

Figure 9 is a flow chart of the preliminary quality detection unit in the embodiment of the present invention;

Figure 10 is a schematic structural diagram of the neural network of the secondary defect detection unit in the embodiment of the present invention.

In the picture: 1. Base; 2. Working compartment; 3. Storage drawer; 4. Handle; 5. Detection structure; 501. First motor; 502. First threaded barrel; 503. First threaded rod; 504. Placement plate; 505, suction cup; 506, second threaded barrel; 507, second threaded rod; 508, drive motor; 509, mounting plate; 510, second motor; 511, clamping plate; 512, anti-slip pad; 513, No. 1. Limiting rod; 6. Bracket; 7. Support rod; 8. Light; 9. Buzzer; 10. Machine vision structure; 1001. First fixed plate; 1002. Connecting rod; 1003. Second fixed plate; 1004 , second limiting rod; 1005, third motor; 1006, connecting plate; 1007, camera; 11, control unit.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.

Implementation case one

As shown in Figures 1 and 2, the present invention provides a technical solution: a camera mold welding quality detection device and its system, including a base 1, a working chamber 2 arranged above the base 1, and an internal setting of the working chamber 2 There is a storage drawer 3. A handle 4 is provided above the storage drawer 3. A detection structure 5 is provided above the working compartment 2. A bracket 6 is provided on the left side of the detection structure 5. A support rod 7 is provided above the support 6. A buzzer 9 is provided inside the rod 7 , a machine vision structure 10 is provided below the buzzer 9 , a lighting lamp 8 is provided on the left side of the machine vision structure 10 , and a control unit 11 is provided on the right side of the buzzer 9 .

Embodiment 2

The solution in Embodiment 1 will be further introduced below in conjunction with the specific working methods. For details, see the description below:

As shown in Figures 3 and 4, as a preferred embodiment, based on the above method, further, the detection structure 5 includes a first motor 501, and a first threaded barrel 502 is provided above the first motor 501. A first threaded rod 503 is provided inside the threaded barrel 502. A placing plate 504 is provided above the first threaded rod 503. A suction cup 505 is provided above the placing plate 504. A second threaded barrel 506 is provided inside the bracket 6. A second threaded rod 507 is provided inside the threaded barrel 506, a driving motor 508 is provided at the outer end of the second threaded barrel 506, a mounting plate 509 is provided on the right side of the second threaded rod 507, and a third mounting plate 509 is provided on the right side of the mounting plate 509. Two motors 510. A clamping plate 511 is provided on the right side of the second motor 510. An anti-slip pad 512 is provided on the right side of the clamping plate 511. A first limiting rod 513 is provided below the second motor 510. The first threaded barrel 502 It is threadedly connected with the first threaded rod 503, the second threaded barrel 506 is threadedly connected with the second threaded rod 507, the anti-slip pad 512 is adhesively connected with the clamping plate 511, the object is placed on the placing plate 504 and contacts with the suction cup 505, and the placed object falls out. After falling, the first motor 501 starts to work and drives the object to the clamping plate 511 through the threaded connection between the first threaded barrel 502 and the first threaded rod 503. The driving motor 508 starts to work. Through the second threaded barrel 506 and the first The threaded connection between the two threaded rods 507 drives the clamping plate 511 to clamp the object, and through the work of the second motor 510, the object is driven to rotate, enabling all-round detection.

As shown in Figure 5, as a preferred embodiment, on the basis of the above method, the machine vision structure 10 further includes a first fixed plate 1001, and a connecting rod 1002 is provided below the first fixed plate 1001. The connecting rod 1002 A second fixed plate 1003 is provided below, a second limiting rod 1004 is provided below the second fixed plate 1003, a third motor 1005 is provided on the left side of the second limiting rod 1004, and a connecting plate 1006 is provided below the third motor 1005 , a camera 1007 is provided below the connecting plate 1006, the connecting rod 1002 is threadedly connected to the first fixed plate 1001, the third motor 1005 is threadedly connected to the second fixed plate 1003, and the camera 1007 is fixedly connected to the third motor 1005 through the connecting plate 1006 , the third motor 1005 is threadedly connected to the first fixed plate 1001 through the connecting rod 1002, which can be easily disassembled, installed and repaired. The connecting plate 1006 and the second fixed plate 1003 are connected through the second limiting rod 1004 to prevent the camera. 1007 shakes when turning.

Example 1:

As shown in Figures 6, 7, 8, 9 and 10, the camera module welding defect detection system provided by the embodiment of the present invention includes:

Optical imaging unit, used for imaging and image sequence collection of the welding area of the camera module;

The image preprocessing unit is used to perform welding area positioning and sequence image fusion processing on the acquired image sequence to be detected, so as to obtain high-quality, easy-to-process welding area images;

The quality inspection unit is used for welding quality inspection of the welding area to be inspected. It conducts preliminary quality inspection of the welding quality of the camera module on a module basis and performs neural network secondary defect determination on the welding quality of the camera module on a solder point basis;

The display unit is used to display the processed camera module and welding area images, and display the location and type of marked defects.

In this example, the image preprocessing unit is connected to the optical imaging unit and the quality detection unit respectively, and the display unit is connected to the quality detection unit;

In this example, in order to avoid the uneven exposure that often occurs in one imaging, resulting in greater reflection in the welding area or unclear edges, and poor imaging quality, the imaging device collects image sequences with different exposure times. Specifically, the exposure time is set to low , medium and high, the same camera module collects three images;

In this example, the welding area positioning method uses a feature matching algorithm optimized by the pyramid search strategy. The positioning is only performed on high-exposure images. After the positioning coordinates are obtained, all image sequence data are shared;

In this example, in order to further improve the speed of the algorithm, image sequence fusion only fuses the welding area images cropped after positioning the welding area;

In this example, the solder joint features of the welding area in the preliminary quality inspection are extracted as the contour geometric features of the solder joint. Specifically, the feature quantities are area, perimeter, regional convexity, tightness, angle, and aspect ratio; through adjustable decision-making Use the above machine learning method to determine whether the welding quality of the camera module is qualified;

In this example, the neural network model of the secondary defect determination method is obtained through training on the solder joint data set. The input image is a high-quality solder joint image after image fusion, and the output is whether the solder joint is qualified and the type and location of the defect;

Beneficial effects of the above technical solution: In order to change the shortcomings of manual visual inspection such as fatigue and slow speed in the past, an inspection method based on machine vision is set up. The image of the item to be inspected is collected through the acquisition unit, and is processed by the image preprocessing unit. The quality inspection unit determines the quality according to the qualification standards, and transmits the inspection results to the display unit for display, making it easier for staff to view.

Example 2:

Based on Embodiment 1, the camera module welding defect detection system and optical imaging unit provided by this embodiment of the present invention include:

The image acquisition unit is used to collect camera module images, and uses the preset image acquisition exposure time and the number of images contained in the image sequence to collect a preset number of images with different exposure times for the same camera module to be detected;

The lighting unit uses a preset fixed angle between the welding area of the camera module and the image acquisition device to obtain images of the welding area with a high contrast between the solder and the pad;

In this example, the image acquisition unit can use low-cost CMOS industrial cameras and ordinary industrial lenses. The conditions that must be met are that high-definition solder joint images can be obtained and parameters such as exposure time can be freely set through programming;

In this example, in order to obtain an image with a high contrast between the solder joints and the solder pad, the specific lighting method is a low-angle ring light source dark field illumination, and the solder pad and the light source need to be at a certain angle so that the light reflected by the solder pad does not enter the camera. It is a dark area in imaging;

The beneficial effects of the above technical solution: collecting multiple images with different exposure times can use image processing algorithms to further eliminate solder joint reflections and enhance image details; the cost of the lighting unit used is low, the welding area has clear images and obvious defect characteristics, which can further Convenient for subsequent image processing steps.

Example 3:

Based on Embodiment 1, a machine vision-based camera module welding defect detection device and method, image preprocessing unit, includes:

The welding area positioning unit uses the feature matching method of the pyramid hierarchical search strategy to automatically and accurately locate the welding area of the camera module, and obtains the welding area position information of each image in the collected camera module image sequence;

The image sequence fusion unit uses a multi-exposure image fusion method in which the exposure function and local detail retention jointly determine pixel weights to fuse the acquired welding area image sequences to obtain a welding area image with uniform exposure and rich detail information.

In this example, after the pyramid layering strategy is introduced into the search process, the number of pyramid layers is first determined. Then the optimization search process is divided into two parts: coarse positioning of low-resolution images at the top of the pyramid and fine positioning of high-resolution images in the upper layer. Specifically, the coarse positioning part is performed on the top-level low-resolution image, and the search sub-image generated at each candidate position is compared exhaustively with the template image. The top-level image generates fewer search sub-images due to its smaller size. Therefore, firstly, coarse positioning of the low-resolution top image excludes irrelevant areas and then narrows the fine positioning search range, which can greatly improve the search speed. At the same time, during the top-level coarse positioning search, the judgment threshold is appropriately lowered to reduce the risk of losing image details at low resolution. Impact, the candidate point areas that meet the requirements are passed into the precise positioning process. The precise positioning search at each level of the image pyramid only selects the position near the positioning area of the previous layer to search, and calculates the similarity. In this way, the original image positioned at the bottom of the pyramid is searched cyclically, and the image positioning accuracy is highest in the original image.

In this example, feature matching uses geometric invariant moment features. After the first layer of the pyramid template is accurately positioned, the position is selected according to the monocular module and binocular module templates, and the welding area is obtained after being offset in the x and y directions respectively. . Define the position, width and height of each group of template pads based on the pads in the welding area for subsequent solder joint image defect determination.

In this example, multi-exposure image fusion uses certain image fusion rules to present both darker area information and reflective area information in one image, ultimately obtaining a welding area image with low reflection and uniform exposure.

The beneficial effects of the above technical solution: taking the positioned welding area as the image processing object, improving the speed of subsequent image processing and reducing interference in non-detection areas; the detail-enhanced multi-exposure image fusion method can weld the collected camera modules with different exposure times Regional image sequences are fused to obtain high-quality welding area imaging, which can further improve the accuracy of quality inspection.

Example 4:

On the basis of Embodiment 3, the camera module welding defect detection system and image sequence fusion unit provided by this embodiment of the present invention include:

The image alignment component is used to align the corresponding pixel positions of the acquired welding area image sequence;

The image fusion component is used to calculate the weight value of each image pixel based on the two indicators of exposure weight and local detail retention weight to obtain a comprehensive weight map, thereby obtaining a high-quality fused image.

In this example, before image fusion, in order to avoid slight offsets between images with different exposure times, resulting in blurred artifacts and other phenomena in the fused image, the images to be fused first need to be registered and corrected to align their pixels. The steps are as follows : First, grayscale the image, determine the median value of the image pixels, use the median value as the threshold to obtain the median threshold bitmap, and then construct the image pyramid to calculate the offset to obtain the offset (Δx, Δy). The quasi-image is moved by Δx in the horizontal direction and Δy in the vertical direction to obtain the final registration image sequence.

In this example, the exposure weight function is constructed as formula (I):

In the formula, α and β are the balance coefficients in the weight function, which can balance the global brightness and local details of the image;

In this example, the homomorphic filtering algorithm is used to preserve local details to enhance the amount of local detail information to obtain the detail weight map. Homomorphic filtering is processed in the frequency domain, which can enhance the contrast while compressing the grayscale range of the image, reducing the low-frequency information in the image and increasing the high-frequency information, thereby reducing brightness changes and sharpening image edge details;

Perform a homomorphic filtering operation on each image in the image sequence to obtain k images T _k (i, j), and the difference between them and the corresponding original image can obtain the area with the richest details in the image. The weight function is as shown in formula (II):

In this example, a comprehensive weight calculation is finally performed, taking the above two factors into consideration to calculate the weight value of the image. First, the two weight indicators are multiplied and normalized to obtain the weight of each pixel in the image. The comprehensive weight calculation is as follows (III):

Among them, σ is a very small positive number that avoids the denominator being zero. This article takes σ = 10 ^-7 , μ ₁ and μ ₂ are the weight values of the weight parameters. In the embodiment, their values are all 1;

In this example, in order to obtain better image quality, optimize weight map fusion, and reduce artifacts in the fused image, a multi-scale pyramid method is used for image fusion.

The beneficial effects of the above technical solution: the fused welding area has clear layers around it, better highlight suppression, and retains more detailed information than other methods, which is beneficial to later feature extraction and final defect detection.

Example 5;

Based on Embodiment 1, the camera module welding defect detection system and quality inspection unit provided by this embodiment of the present invention include:

The preliminary quality inspection unit is used to quickly determine whether the welding quality of the camera module is qualified and output the results;

The secondary defect detection unit is used to further conduct secondary quality inspections on modules that are judged to be unqualified by the preliminary quality inspection unit, eliminate qualified modules that are mistakenly judged to be defective, and check the types and locations of defects in unqualified modules. Identify and label.

In this example, the input of the preliminary quality inspection unit is the high-exposure welding area image in the image sequence. The welding area image is enhanced through a nonlinear image enhancement method to remove the interference of flux impurities and enhance the edge information. Then the edge contour of the solder joint is extracted through an edge detection algorithm. Obtain the geometric contour characteristics of the solder joints and input the minimum risk Bayesian decision-making algorithm of the adjustable risk function to determine whether the module is qualified;

Nonlinear image enhancement method. First, taking into account the real-time requirements of solder joint detection, a fast median filtering algorithm is used to filter out interference around the solder joint; secondly, based on the adaptive Gamma algorithm, the image quality between the welding area and the non-welding area is improved. Gray value difference.

In this example, a risk factor function is introduced to measure the risk when an error occurs, so that defective modules can be detected as much as possible. In this process, a small number of qualified camera modules are misjudged as defects.

Example 6:

On the basis of Embodiment 5, the camera module welding defect detection system and secondary defect detection unit provided by this embodiment of the present invention include:

The data set construction component is used to build a data set of the solder joints of the camera module to be detected. It uses the preset pad positions in the template area to crop the welding area, obtains the target image of each solder joint, and classifies the solder joint defects according to the type. And divided into sample training set and sample test set according to a certain proportion;

The model training component is used to train the neural network model for identifying solder joint defect types, and uses the sample training set to iteratively optimize the parameters until the training model input sample test set reaches the expected prediction accuracy to stop training and obtain the solder joint defect classification model;

The defect detection component is used to determine whether there are defects in the input solder joint image and what type of defects there are.

In this example, in the data set in the model training component, high-quality solder spot images after multi-exposure image fusion are selected to create a data set, and the center point of each solder pad is predefined in the extracted welding area, and the size of the solder pad is determined according to the size. The solder spot images are intercepted to ensure that adjacent solder spots have overlapping areas, and finally several solder spot images are obtained through manual screening. In addition, due to the small sample size, this article uses data enhancement to expand the sample size of the solder joint data set, and uses mirror flipping and adding noise perturbation to expand the sample size based on the actual situation of camera module welding. After expansion, the final solder joint data set is obtained. Randomly select 70% of the data set as the training set, and 30% as the verification set, including four welding types: qualified, bridge, missing welding and less welding.

In this example, since there are not many types of defects that need to be detected, and the sample size is small and the input image is small, consider using the Bayesian method to optimize the shallow and simple LeNet-5 convolutional neural network structure, and add the batch normalization design network structure.

Beneficial effects of the above technical solution: In view of the real-time requirements in actual industrial production and the fact that defective module samples are less prone to over-fitting problems, a preliminary inspection based on modules and two inspections based on solder joints are designed. . Since the probability of defective modules in industrial practice is small, in order to ensure real-time requirements, a Bayesian decision-making algorithm is used to extract the geometric features of solder joints after image enhancement processing and input the minimum risk, and adjust the risk function to ensure that there are welding defects. All modules are checked out. Due to the misjudgment of some qualified modules in the first step and the need to determine the type and location of welding defects, the second step of detection designed a convolutional neural network structure improved by the Bayesian method. The accuracy of the algorithm on the camera module welding data set is as high as Good results have been achieved in terms of prediction speed. This camera module defect detection solution solves the problem of industrial real-time detection and reduces misjudgment of camera module welding defects, improving detection efficiency and accuracy.

Example 7:

Based on Embodiment 1, the camera module welding defect detection system and display unit provided by the embodiment of the present invention include:

The first display screen is used to display the current inspection camera module image and the welding area image, mark whether it is qualified, and check the location and type of defective solder joints.

The second display screen is used to display parameter settings for setting adjustable parameters involved in the machine vision-based camera module welding defect detection device and method.

In this example, all parameters involved in the present invention can be adjusted on the second display screen, and the results are displayed on the first display screen.

Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art will understand that various changes, modifications, and substitutions can be made to these embodiments without departing from the principles and spirit of the invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (10)

1. A camera mold welding quality detection device, characterized in that it includes a base (1), a working chamber (2) is provided above the base (1), and a storage chamber is provided inside the working chamber (2) Drawer (3), a handle (4) is provided above the storage drawer (3), a detection structure (5) is provided above the working compartment (2), and a detection structure (5) is provided on the left side of the There is a bracket (6), a pole (7) is provided above the bracket (6), a buzzer (9) is arranged inside the pole (7), and below the buzzer (9) A machine vision structure (10) is provided, an illumination lamp (8) is provided on the left side of the machine vision structure (10), and a control unit (11) is provided on the right side of the buzzer (9). 2. A camera mold welding quality detection device according to claim 1, characterized in that the detection structure (5) includes a first motor (501), and a third motor (501) is provided above the first motor (501). A threaded barrel (502). A first threaded rod (503) is provided inside the first threaded barrel (502). The first threaded barrel (502) is threadedly connected to the first threaded rod (503). A placement plate (504) is provided above the first threaded rod (503), a suction cup (505) is provided above the placement plate (504), and a second threaded barrel (506) is provided inside the bracket (6). , a second threaded rod (507) is provided inside the second threaded barrel (506), the second threaded barrel (506) is threadedly connected to the second threaded rod (507), the second threaded barrel (506) ) is provided with a driving motor (508) at its outer end. 3. A camera mold welding quality inspection device according to claim 2, characterized in that a mounting plate (509) is provided on the right side of the second threaded rod (507), and a mounting plate (509) is provided on the right side of the second threaded rod (507). A second motor (510) is provided on the right side, a clamping plate (511) is provided on the right side of the second motor (510), and an anti-slip pad (512) is provided on the right side of the clamping plate (511). The anti-slip pad (512) is adhesively connected to the clamping plate (511), and a first limiting rod (513) is provided below the second motor (510). 4. A camera mold welding quality inspection device according to claim 2, characterized in that the machine vision structure (10) includes a first fixed plate (1001), and the bottom of the first fixed plate (1001) A connecting rod (1002) is provided, and a second fixing plate (1003) is provided below the connecting rod (1002). The connecting rod (1002) is threadedly connected to the first fixing plate (1001). 5. A camera mold welding quality inspection device according to claim 2, characterized in that a second limiting rod (1004) is provided below the second fixed plate (1003), and the second limiting rod (1004) A third motor (1005) is provided on the left side of the motor (1004). The third motor (1005) is threadedly connected to the second fixing plate (1003). A connecting plate (1006) is provided below the third motor (1005). , a camera (1007) is provided below the connecting plate (1006). 6. A camera mold welding quality inspection system, characterized by including: The image acquisition unit acquires the image sequence of the camera module to be detected; The image preprocessing unit performs welding area positioning and sequence image fusion processing on the acquired image sequence to be detected to obtain high-quality welding area images; The welding quality inspection unit performs preliminary quality inspection of the welding quality of the camera module on a module basis and neural network secondary defect determination of the welding quality of the camera module on a solder point basis; The display unit displays the quality inspection results of the camera module welding and marks the location and type of welding defects. 7. A camera module welding defect detection system according to claim 1, characterized in that: The image sequence is collected by an optical imaging unit, wherein the optical imaging unit includes: The image acquisition unit collects the camera module image sequence, and the image sequence is the camera module image collected under different exposure times; the lighting unit has a preset fixed angle between the camera module welding area and the image acquisition equipment, and is used to acquire the solder and solder Imaging of welding areas with high disc contrast; The image preprocessing unit includes: The welding area positioning unit is used to automatically and accurately position the welding area of the camera module and obtain the welding area position information of each image in the collected camera module image sequence; The image sequence fusion unit is used to fuse the acquired welding area image sequences to obtain a welding area image with uniform exposure and rich detailed information; The welding area positioning unit includes: The template area selection component is used to use the principle that the welding area is fixed relative to the camera module's spatial position, and uses the template matching method to select the template image for area matching and positioning; The similarity comparison component is used to compare and calculate the similarity between the template image and the search sub-image in the original image. The position of the maximum similarity sub-image is the target position; The search acceleration component is used to accelerate the speed of matching and positioning. It adopts a pyramid hierarchical search strategy and presets the starting search coordinate point. It first performs rough matching and then precise matching; The welding area cropping component is used to obtain the matching positioning coordinates, and by presetting the length and width of the welding area, crop the welding area of each image in the image sequence and store it as a welding area image with a smaller size and less background area. ; The image sequence fusion unit includes: The image alignment component is used to align the corresponding pixel positions of the acquired welding area image sequence; The image fusion component is used to calculate the weight value of each image pixel based on the two indicators of exposure weight and local detail retention weight to obtain a comprehensive weight map, thereby obtaining a high-quality fused welding area image; The quality inspection unit includes: The preliminary quality inspection unit is used to quickly determine whether the welding quality of the camera module is qualified and output the results; The secondary defect detection unit is used to further conduct secondary quality inspections on modules that are judged to be unqualified by the preliminary quality inspection unit, eliminate qualified modules that are mistakenly judged to be defective, and check the types and locations of defects in unqualified modules. Identify and label. 8. A camera module welding defect detection system according to claim 7, characterized in that the preliminary quality detection unit includes: Image enhancement component, used to improve the interference of flux impurities in the welding area, enhance edge contour information, and improve the accuracy of subsequent feature extraction; The feature extraction component is used to effectively characterize the characteristics of various types of welding defects. The welding area processed by the image enhancement component uses an edge detection algorithm to identify the edge contour of the solder joint, and obtains the solder joint contour through hole filling and morphological operations, and then calculates the solder joint. Point outline features; The preliminary classification component is used to initially classify qualified camera modules and unqualified camera modules after extracting the welding defect characteristics of the camera module. It uses a machine learning method with adjustable output categories to make quality judgments so that defective camera modules can be identified. All groups are detected. 9. A camera module welding defect detection system according to claim 7, characterized in that the secondary defect detection unit includes: The data set construction component is used to build a data set of the solder joints of the camera module to be detected. It uses the preset pad positions in the template area to crop the welding area, obtains the target image of each solder joint, and classifies the solder joint defects according to the type. And divided into sample training set and sample test set according to a certain proportion; The model training component is used to train a neural network model for identifying solder joint defect types, and uses the sample training set to perform parameter iterative optimization until the training model inputs the sample test set and reaches the expected prediction accuracy to stop training and obtain the solder joint defect classification. Model; The defect detection component is used to determine whether there are defects in the input solder joint image and what type of defects there are. 10. A camera module welding defect detection system according to claim 6, characterized in that the display unit includes: The first display screen is used to display the current inspection camera module image and the welding area image, mark whether it is qualified, and check the location and type of defective solder joints; The second display screen is used to display parameter settings for setting adjustable parameters involved in the machine vision-based camera module welding defect detection device and method.