A system and method for inductance product package detection

Abstract

The application relates to the technical field of visual detection of semiconductor manufacturing, in particular to a packaging detection system and method for inductance products. The system comprises a packaging detection device with a first station and a second station and a visual detection device. The inductance products are loaded and positioned at the first station, and packaging defect detection and unloading are carried out at the second station. The system further comprises a control module. Contour feature data and texture feature data are obtained according to original images at the first station, and positioning control is carried out. Then, pixel statistics and comparison analysis are carried out on the packaging images based on the feature data, packaging defect detection results are obtained, and unloading indication control is carried out. Compared with the prior art, the application innovatively utilizes the process characteristics of the inductance products between the packaging double stations, realizes efficient and light-weight defect detection through a feature data cross-station reuse mechanism, greatly reduces the image processing data volume, and solves the problem that the packaging detection by using vision in the prior art has high requirements on hardware.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual inspection for semiconductor manufacturing, and in particular to a packaging detection system and method for inductance products. BACKGROUND

[0002] Packaging detection is a key link in the production of electronic components, and its core is to ensure that the product meets the design specifications during the packaging process through high-precision measurement and defect recognition. Traditional detection methods rely on manual visual inspection or contact gauges, but are limited by subjective judgment differences and measurement efficiency, making it difficult to meet the production needs of modern electronic components in large quantities and high consistency.

[0003] With the development of machine vision technology, automated detection schemes based on image processing have gradually become the mainstream - by industrial cameras to capture product images, combined with edge detection, feature matching and other algorithms to extract key parameters, and use pre-set thresholds to realize defect classification. Compared with traditional methods, machine vision assisted detection has the advantages of non-contact measurement and suitability for large-scale detection.

[0004] However, the existing machine vision detection technology still has obvious bottlenecks when applied to inductance packaging scenarios - its algorithm usually relies on complex deep learning models, requiring extremely high hardware computing power. For example, traditional solutions require high-performance GPUs to process high-resolution images (such as more than 5 million pixels) in real time, and multi-scale feature extraction, multi-target segmentation and other steps will further increase the computational load, resulting in single-frame processing time of tens of milliseconds, which is difficult to match the fast production line rhythm, and also increases the cost of equipment. Therefore, people need a more lightweight visual-based packaging detection scheme. SUMMARY

[0005] Therefore, the present application provides a packaging detection system and method for inductance products to solve the problem of high hardware requirements for packaging detection using vision in the prior art.

[0006] The present application provides a packaging detection system for inductance products, comprising a packaging detection device with a first station and a second station, and a visual detection device, the visual detection device is used to shoot the original image of the inductance product at the first station, and shoot the packaging image of the inductance product at the second station, the inductance product is loaded and positioned at the first station based on the original image, and after packaging by the packaging detection device, the packaging defect detection and unloading are carried out at the second station based on the packaging image, the system further comprises a control module, the control module comprises:

[0007] a contour analysis module for obtaining contour feature data of the inductance product according to the original image;

[0008] a positioning control module for obtaining positioning instruction data of the first station according to the contour feature data;

[0009] a texture analysis module configured to obtain texture feature data of the inductor product according to the original image;

[0010] a package detection module configured to perform pixel statistics and comparison analysis on the package image based on the contour feature data and the texture feature data, and obtain a package defect detection result;

[0011] a detection control module configured to obtain blanking instruction data according to the package defect detection result.

[0012] In a preferred implementation, the contour feature data of the inductor product is obtained according to the original image, including:

[0013] gradient calculation is performed on the original image to obtain a gradient amplitude map and a gradient direction map;

[0014] a direction filtering range is obtained according to the size and shape of the inductor product;

[0015] non-maximum suppression is performed in the direction filtering range based on the gradient direction map to obtain a thinned contour map;

[0016] the contour feature data is obtained according to the thinned contour map.

[0017] In a preferred implementation, the direction filtering range is obtained according to the size and shape of the inductor product, including:

[0018] a first range component is obtained according to the aspect ratio of the inductor product;

[0019] a second range component is obtained according to the size ratio of the inductor product and background objects in the original image;

[0020] the first range component and the second range component are integrated to obtain the direction filtering range.

[0021] In a preferred implementation, the texture feature data of the inductor product is obtained according to the original image, including:

[0022] a ROI region of the inductor product is obtained in the original image according to the contour feature data;

[0023] statistical analysis is performed on the gray value in the ROI region to obtain the texture feature data.

[0024] In a preferred implementation, the package defect detection result includes a glue amount uniformity detection result; the pixel statistics and comparison analysis on the package image based on the contour feature data and the texture feature data to obtain the package defect detection result, including:

[0025] the package image is obtained, and the package image has a preset positioning center corresponding thereto;

[0026] obtaining a first detection region in the packaging image according to the contour feature data, wherein a range of the first detection region is obtained based on an expansion of the contour represented by the contour feature data;

[0027] statistically counting the gray value of the first detection region to obtain a gray histogram of the first detection region;

[0028] calculating a kurtosis of the gray histogram, and obtaining a glue amount uniformity detection result according to a comparison between the kurtosis and a preset kurtosis threshold.

[0029] In a preferred implementation manner: the packaging defect detection result includes a bubble defect detection result; the packaging defect detection result is obtained by performing pixel counting and comparison analysis on the packaging image based on the contour feature data and the texture feature data, and includes:

[0030] obtaining a packaging image, the packaging image corresponding to a preset positioning center;

[0031] obtaining a second detection region in the packaging image according to the contour feature data based on the preset positioning center, wherein a range of the second detection region is obtained based on an expansion of the contour represented by the contour feature data;

[0032] dividing a uniform grid in the second detection region, and randomly selecting a plurality of windows in each grid to obtain a plurality of detection windows;

[0033] statistically counting the gray value in each detection window to obtain a texture detection feature of each detection window;

[0034] comparing the texture detection feature with the texture feature data to obtain an abnormal window number;

[0035] obtaining the bubble defect detection result according to a comparison between the abnormal window number and a preset number threshold.

[0036] In a preferred implementation manner: the packaging detection device further includes a third station for dispensing and packaging the inductor product; the packaging defect detection result includes an overflow glue detection result; the packaging defect detection result is obtained by performing pixel counting and comparison analysis on the packaging image based on the contour feature data and the texture feature data, and includes:

[0037] obtaining a packaging image, the packaging image corresponding to a preset positioning center;

[0038] obtaining a third detection region in the packaging image according to the contour feature data based on the preset positioning center, wherein a range of the third detection region is obtained based on an expansion of the contour represented by the contour feature data;

[0039] Obtain the dispensing pressure and glue nozzle moving speed of the third station, and obtain the glue area threshold according to the dispensing pressure and glue nozzle moving speed;

[0040] Based on the gray value of the third detection area, the number of pixels capable of representing the glue area in the third detection area is counted;

[0041] According to the comparison of the number of pixels and the glue area threshold, the glue overflow detection result is obtained.

[0042] The application also provides a packaging detection method of an inductor product, which is applied to a packaging detection system of the inductor product.

[0043] According to the original image, the contour feature data of the inductor product is obtained;

[0044] According to the contour feature data, the positioning indication data of the first station is obtained;

[0045] According to the original image, the texture feature data of the inductor product is obtained;

[0046] Based on the contour feature data and the texture feature data, pixel statistics and comparison analysis are performed on the packaging image to obtain the packaging defect detection result;

[0047] According to the packaging defect detection result, the unloading indication data is obtained.

[0048] The application also provides an electronic device, which comprises:

[0049] A memory and a processor;

[0050] The memory is used for storing a program, and the processor is used for executing the steps in the packaging detection method of the inductor product when the program is executed.

[0051] The application also provides a computer readable storage medium, which is used for storing a computer readable program or instruction.

[0052] The beneficial effects of the above scheme are:

[0053] The application provides a packaging detection system of an inductor product, which comprises a packaging detection device with a first station and a second station and a visual detection device. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The system architecture diagram of the packaging detection system of the inductor product provided by the application is shown in the figure.

[0055] Figure 2 The method flow chart of the packaging detection method of the inductor product provided by the application is shown in the figure.

[0056] Figure 3 The specific step chart of step S201 in the method is shown in the figure. Figure 2 The specific step chart of step S201 in the method is shown in the figure. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0058] In combination withFigures 1-2 As shown in the figure, one specific embodiment of the present application discloses a packaging detection system of inductance products, which comprises a packaging detection device 110 with a first station 111 and a second station 112, and a visual detection device 120, the visual detection device is used to shoot the original image of the inductance products at the first station, and shoot the packaging image of the inductance products at the second station, the inductance products are fed and positioned at the first station based on the original image, and after being packaged by the packaging detection device, the inductance products are detected for packaging defects and discharged at the second station based on the packaging image, the system further comprises a control module 130, the control module comprises:

[0059] a contour analysis module 131, used to execute step S201, and obtain the contour feature data of the inductance products according to the original image;

[0060] a positioning control module 132, used to execute step S202, and obtain the positioning indication data of the first station according to the contour feature data;

[0061] a texture analysis module 133, used to execute step S203, and obtain the texture feature data of the inductance products according to the original image;

[0062] a packaging detection module 134, used to execute step S204, and perform pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data, to obtain the packaging defect detection result;

[0063] a detection control module 135, used to execute step S205, and obtain the discharge indication data according to the packaging defect detection result.

[0064] In this embodiment, the control module mainly refers to a computer program product, and the packaging detection device and the visual detection device can be realized in any form. For example, in the working platform of a certain packaging detection integrated automatic device, three turntable mechanisms are simultaneously arranged, which are A disc, B disc and C disc. The A disc is provided with a corresponding feeding mechanism, a camera and a four-jaw chuck and other clamping mechanisms, and the feeding and positioning are controlled through the image recognition of the camera. The inductance products positioned well are transferred to the B disc through the A disc, the B disc is provided with corresponding glue dispensing and baking mechanisms, which are used to package the inductance products. The packaged products are transferred to the C disc through the B disc, the C disc is provided with a camera and a suction disc and other structures, which are used to analyze the defects of the packaging through the image recognition of the camera, and to decide to control the qualified products to be discharged and taken by the next process, while the unqualified products are NG rejected. The A disc is the first station in the present application, the C disc is the second station in the present application, the camera is the visual detection device, the image shot at the A disc is the original image, and the image shot at the C disc is the packaging image.

[0065] Compared with the prior art, the application innovatively utilizes the process characteristics of the inductor product between the packaging double stations, realizes efficient and lightweight defect detection through a feature data cross-station multiplexing mechanism. Specifically, the contour feature data extracted at the first station is not only used for positioning, but also directly transmitted to the second station for rapid target identification (for example, the contour is expanded by 3mm, and the area in the packaging image that can represent the workpiece is quickly obtained), that is, the repeated target identification calculation process in the defect detection process is avoided, and the image size that needs to be processed is greatly reduced. At the same time, the process of extracting texture features at the first station can utilize the inductor product during the packaging process (such as the dispensing time of the B disc in the above embodiment), fully utilizing the processing time gap between the first station and the second station, forming a perfect staggered pipeline, and greatly shortening the operation time. Most importantly, based on the appearance characteristics of the inductor product packaging (unchanged shape, enlarged size, different packaging glue color, the color of the inductor product may change before and after packaging, and the background texture of the turntable is basically the same), the texture feature data can be directly used at the second station, and the packaging detection can be quickly performed through pixel statistics and comparison analysis (i.e. a non-artificial intelligence recognition method), without running complex image recognition algorithms based on deep learning, that is, the algorithm complexity is reduced, the hardware cost is reduced, the efficiency and accuracy requirements are perfectly balanced, and the problem of high hardware requirement for packaging detection by vision in the prior art is solved.

[0066] In combination Figure 3 In a more preferred embodiment, the step S201 of obtaining the contour feature data of the inductor product according to the original image specifically comprises:

[0067] S301, gradient calculation is performed on the original image to obtain a gradient amplitude graph and a gradient direction graph;

[0068] S302, a direction filtering range is obtained according to the specification shape of the inductor product;

[0069] S303, based on the gradient direction graph, non-maximum suppression is performed within the direction filtering range to obtain a thinned contour graph;

[0070] S304, the contour feature data is obtained according to the thinned contour graph.

[0071] The above process is an improvement on the existing Canny edge detection algorithm, wherein the gradient calculation in step S301 can be performed by a Sobel operator or the like, and the profile representing the inductor product can be quickly located in step S304 based on the positional relationship between the camera and the turntable itself, and the length, width, and center coordinates of the minimum circumscribed rectangle can be calculated to obtain the profile feature data, all of which are existing technologies that can be understood by those skilled in the art, and therefore will not be described in detail herein. It should be noted that steps S302 and S303 of the present embodiment are an improvement on the non-maximum suppression process in the existing Canny algorithm. In non-maximum suppression, the usual approach is to check the gradient direction of the current pixel, and then compare the gradient amplitudes of the adjacent pixels in the positive and negative directions of the direction to retain the maximum value for screening. In the present embodiment, however, the direction screening range is obtained using the size and shape of the inductor product, and screening is performed within the range rather than in a single direction, which has the following beneficial effects:

[0072] Improving the retention rate of inductor edges: Since inductor edges have continuity in multiple directions, while background edges are often isolated, multi-directional non-maximum suppression within a range can better retain inductor edges.

[0073] Effectively suppressing background edges: Background edges usually do not satisfy the local maximum in multiple directions (e.g., the isolated turntable can be considered a straight line compared to an inductor), and are therefore suppressed.

[0074] Adaptive adjustment: The direction range is dynamically adjusted according to the size and shape of the specific inductor product, so that high detection rates can be maintained in complex backgrounds, and positioning accuracy can be improved in simple backgrounds.

[0075] Specifically, in a preferred embodiment, the above step S302 obtains the direction screening range based on the size and shape of the inductor product, specifically including:

[0076] Obtaining a first range component based on the aspect ratio of the inductor product;

[0077] Obtaining a second range component based on the size ratio of the inductor product and the background objects in the original image;

[0078] Combining the first range component and the second range component to obtain the direction screening range.

[0079] It can be understood that, on the one hand, in practice, the inductor usually has a long axis and a short axis, and the edge direction is mainly along the long axis direction (0 degrees) and the short axis direction (90 degrees), and the 45-degree and 135-degree directions (because the inductor is usually rectangular or oval, there are four main edge directions). Therefore, in the non-maximum suppression of the present embodiment, by considering the non-maximum suppression in multiple directions (such as the gradient direction θ, considering the direction in the range of θ+45°~θ-45°), it can be ensured that the current edge point is the local maximum in multiple directions, so as to be more likely to be the inductor edge (because the inductor edge usually has continuity and regularity in different directions, while the background texture often does not have such characteristics, and is more random). The present embodiment requires that the gradient amplitude of the current pixel is the maximum in this range (the first range component), and the edge point is retained.

[0080] On the other hand, the size of the background object (such as a turntable) is usually larger than that of the inductor, and the change of the edge direction is smaller than that of the inductor product, so the present embodiment also fine adjusts by setting a smaller angle range (i.e. the second range component, such as 5 degrees) to exclude the background edge.

[0081] One formula of the above process is as follows:

[0082] ;

[0083] Where Δθ is the direction screening range, e is the natural constant, k is the preset shape sensitive coefficient, such as 0.8 obtained according to experiments, and R is the aspect ratio of the current inductor product, is the minimum area of the inductor, is the background area. In the above formula, the first term is the first range component, which realizes the aspect ratio deviation penalty through the natural exponential function, and the second term is the second range component, which judges the background interference degree through the size ratio of the inductor and the background, so as to correct the screening angle range to obtain more accurate screening angle range, improve the inductor edge retention rate in the non-maximum suppression process, and reduce the background false detection rate.

[0084] The actual angle range for maximum value comparison based on the gradient direction θ obtained finally is:

[0085] ;

[0086] Further, in a preferred embodiment, the above step S203, obtaining the texture feature data of the inductor product according to the original image, specifically includes:

[0087] Obtaining the ROI region of the inductor product in the original image according to the contour feature data;

[0088] Statistically analyzing the gray value in the ROI region to obtain the texture feature data.

[0089] In the above process, the ROI region is a range region in the original image representing the inductance according to the contour feature, and the texture feature data is obtained by means of gray value statistical analysis (for example, calculating the gray value mean, gray variance, and contrast of the gray co-occurrence matrix). In this embodiment, the texture feature is obtained by means of gray statistical analysis, realizing position-independent texture comparison. In subsequent package detection, only the texture feature data is used as a reference, and the local region texture in the package image is quickly compared with the reference, so that the detection result can be obtained, without repositioning or global scanning, greatly reducing the hardware cost and significantly improving the efficiency.

[0090] Further, in a preferred embodiment, the package defect detection result includes a glue amount uniformity detection result, and on this basis, the above step S204, based on the contour feature data and the texture feature data, performs pixel statistics and comparison analysis on the package image to obtain the package defect detection result, specifically including:

[0091] Obtaining a package image, the package image corresponding to a preset positioning center;

[0092] Based on the preset positioning center, a first detection region in the package image is obtained according to the contour feature data, wherein the range of the first detection region is obtained by expanding the contour represented by the contour feature data;

[0093] Statistically analyzing the gray values of the first detection region to obtain a gray histogram of the first detection region;

[0094] Calculating the kurtosis of the gray histogram, and obtaining a glue amount uniformity detection result according to the comparison between the kurtosis and a preset kurtosis threshold.

[0095] In the above process, the preset positioning center can be preset by a person, for example, the center of the suction cup is defined as the preset positioning center, and the region centered on the preset positioning center is the detection region, which is used to represent the approximate position of the inductance product in the package image after packaging. The detection region is obtained by expanding the contour represented by the contour feature data (the typical expansion range is 2-3 mm), and the degree of expansion can be different for different defect detection targets. In this embodiment, the first detection region is dynamically generated based on the preset positioning center and the contour feature data, realizing accurate positioning of the detection region, and reducing most of the invalid scanning area compared with the fixed region detection method.

[0096] The kurtosis is a concept used to measure the steepness of a probability distribution, and its calculation method is a prior art that can be understood by those skilled in the art. In this embodiment, the gray histogram kurtosis analysis is used as a criterion for glue amount uniformity, and the preset kurtosis threshold can be used to simultaneously identify the glue amount deficiency (distribution is sharp, kurtosis > 3.5) and the glue amount oversaturation (distribution is flat, kurtosis < 2.5) defects, with high detection sensitivity.

[0097] Further, in a preferred embodiment, the packaging defect detection result includes a bubble defect detection result. On this basis, the step S204 described above, based on the contour feature data and the texture feature data, performs pixel statistics and comparison analysis on the packaging image to obtain the packaging defect detection result, including:

[0098] Obtaining a packaging image, the packaging image corresponding to a preset positioning center;

[0099] Based on the preset positioning center, obtaining a second detection area in the packaging image according to the contour feature data, wherein the range of the second detection area is obtained based on the contour feature data.

[0100] Dividing a uniform grid in the second detection area, and randomly selecting a plurality of windows in each grid to obtain a plurality of detection windows;

[0101] Statistically analyzing the gray value in each detection window to obtain the texture detection feature of each detection window;

[0102] Comparing the texture detection feature with the texture feature data to obtain the number of abnormal windows;

[0103] According to the comparison between the number of abnormal windows and the preset number threshold, the bubble defect detection result is obtained.

[0104] It can be understood that in this embodiment, the way of extracting the texture feature in the window is the same as the way in the original image (statistical features of the gray value such as gray mean value, variance, etc.) to facilitate comparison. The difference between the texture detection feature and the texture feature data can be evaluated by calculating the cosine similarity. The specific threshold for defect judgment can also be flexibly set according to actual needs, for example, the detection window with a difference greater than 0.15 is marked as an abnormal window, and when the number of abnormal windows exceeds three, it can be determined that there is a bubble defect.

[0105] This embodiment constructs a multi-scale detection system through uniform grid division (such as 3x3 grid) and random window sampling (5 5x5 pixel windows in each grid), without the need for accurate identification of bubble position, and compared with traditional full-area scanning, it greatly reduces the amount of calculation and greatly reduces the required hardware level.

[0106] Further, in a preferred embodiment, the packaging detection device further includes a third station for dispensing packaging of inductance products; and the packaging defect detection result includes an overflow detection result. On this basis, the step S204 described above, based on the contour feature data and the texture feature data, performs pixel statistics and comparison analysis on the packaging image to obtain the packaging defect detection result, including:

[0107] An encapsulation image is acquired, and the encapsulation image has a preset positioning center corresponding thereto;

[0108] Based on the preset positioning center, a third detection region in the encapsulation image is obtained according to the contour feature data, wherein a range of the third detection region is obtained based on an outward expansion of a contour represented by the contour feature data;

[0109] A dispensing pressure and a glue nozzle moving speed of the third station are acquired, and a glue area threshold is obtained according to the dispensing pressure and the glue nozzle moving speed;

[0110] Based on a gray value of the third detection region, a number of pixels in the third detection region capable of representing a glue area is counted;

[0111] According to a comparison between the number of pixels and the glue area threshold, an overflow glue detection result is obtained.

[0112] In the above process, the third station is a dispensing station located between the first station and the second station, for example, the B disc in the foregoing embodiment. The third detection region is outwardly expanded to a range greater than the first detection region and the second detection region.

[0113] One specific embodiment for obtaining the glue area threshold is as follows:

[0114]

[0115] wherein, is the glue area threshold, w is a preset glue amount diffusion coefficient, for example, 0.18 obtained according to experience, P is the dispensing pressure of the third station, V is the glue nozzle moving speed of the third station, and b is a preset adjustment coefficient, for example, 20 obtained according to experience.

[0116] It can be understood that how to judge the pixels representing the glue area in the above process can be designed according to specific dispensing types. For example, if the encapsulation glue is black and the background turntable is silver, the pixels with a large gray value (for example, a gray value greater than a mean value + 20) in the third detection region are the pixels representing the glue area.

[0117] The embodiment creatively couples real-time dispensing parameters of the third station with the depth of visual detection, generates an adaptive area threshold through a dynamic glue amount model, and greatly improves the adaptability. Most importantly, the embodiment is assisted by process parameters, effectively overcomes the instability problem of relying solely on image visual detection, improves the detection accuracy, ensures the high-speed processing performance, and perfectly solves the detection stability problem caused by environmental fluctuations in mass production.

[0118] It is worth mentioning that the above three defect detection methods do not need to call complex artificial intelligence visual algorithms, and only need to perform simple statistical analysis and comparison on pixel gray values to realize defect judgment, greatly reducing the complexity and hardware requirements.​

[0119] Please combine again Figure 2 The application further provides a packaging detection method of an inductor product, applied to a packaging detection system of the inductor product. The system comprises a packaging detection device with a first station and a second station and a visual detection device. The visual detection device is used to shoot an original image of the inductor product at the first station and a packaging image of the inductor product at the second station. The inductor product is fed and positioned at the first station based on the original image, and is detected for packaging defects and discharged at the second station after being packaged by the packaging detection device. The system further comprises a control module used to run the packaging detection method of the inductor product. The method comprises the following steps:

[0120] S201. Obtaining contour feature data of the inductor product according to the original image.

[0121] S202. Obtaining positioning instruction data of the first station according to the contour feature data.

[0122] S203. Obtaining texture feature data of the inductor product according to the original image.

[0123] S204. Performing pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data to obtain a packaging defect detection result.

[0124] S205. Obtaining discharge instruction data according to the packaging defect detection result.

[0125] The application further provides an electronic device, which comprises:

[0126] a memory and a processor.

[0127] The memory is used to store a program, and the processor is used to execute the steps in the packaging detection method of the inductor product when the program is executed.

[0128] The application further provides a computer readable storage medium used to store a computer readable program or instruction. The program or instruction can realize the steps in the packaging detection method of the inductor product when executed by a processor.

[0129] The application provides a packaging detection system of an inductance product, which comprises a packaging detection device with a first station and a second station and a visual detection device, the visual detection device is used for shooting an original image of the inductance product at the first station and shooting a packaging image of the inductance product at the second station, the inductance product is fed and positioned at the first station based on the original image, and packaging defect detection and discharging are carried out at the second station based on the packaging image after the inductance product is packaged by the packaging detection device, the system further comprises a control module, the control module comprises a contour analysis module used for obtaining contour feature data of the inductance product according to the original image, a positioning control module used for obtaining positioning indication data of the first station according to the contour feature data, a texture analysis module used for obtaining texture feature data of the inductance product according to the original image, a packaging detection module used for carrying out pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data to obtain a packaging defect detection result, and a detection control module used for obtaining discharging indication data according to the packaging defect detection result. Compared with the prior art, the inductance product is innovatively used for the process characteristics between the packaging double stations, the feature data cross-station reuse mechanism is used for realizing efficient and lightweight defect detection, and specifically, the contour feature data and the texture feature data extracted at the first station are directly transmitted to the second station, so that the feature extraction does not need to be repeatedly performed when the second station detects, and only the preset key detection area needs to be rapidly compared and analyzed, the image processing data amount is greatly reduced compared with a traditional full-image scanning scheme, the redundant feature extraction calculation at the second station is as much as possible omitted, the detection speed is improved, the algorithm complexity is reduced, the hardware cost is reduced, the efficiency and accuracy requirements are perfectly balanced, and the problem that the hardware requirement is high in the packaging detection by using the vision in the prior art is solved.

[0130] It should be noted that each of the embodiments in the present specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the embodiments can be referred to each other.

[0131] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An inductance product package detection system characterized by, The system comprises a packaging detection device and a vision detection device, the vision detection device is used for shooting an original image of an inductor product at a first station and shooting a packaging image of the inductor product at a second station, the inductor product is fed and positioned at the first station based on the original image, and after being packaged by the packaging detection device, the inductor product is detected for packaging defects and discharged at the second station based on the packaging image, the system further comprises a control module, and the control module comprises: a contour analysis module, configured to obtain contour feature data of the inductor product according to the original image; a positioning control module, configured to obtain positioning instruction data of the first station according to the contour feature data; a texture analysis module, configured to obtain texture feature data of the inductor product according to the original image; a packaging detection module, configured to perform pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data, and obtain a packaging defect detection result; a detection control module, configured to obtain discharge instruction data according to the packaging defect detection result.

2. The inductance product package detection system of claim 1, wherein, The contour feature data of the inductor product is obtained according to the original image, comprising: gradient calculation is performed on the original image to obtain a gradient amplitude graph and a gradient direction graph; a direction filtering range is obtained according to the size and shape of the inductor product; non-maximum suppression is performed on the gradient direction graph within the direction filtering range to obtain a thinned contour graph; the contour feature data is obtained according to the thinned contour graph.

3. The inductance product package detection system of claim 2, wherein, The direction filtering range is obtained according to the size and shape of the inductor product, comprising: a first range component is obtained according to the aspect ratio of the inductor product; a second range component is obtained according to the size ratio of the inductor product and background objects in the original image; the first range component and the second range component are integrated to obtain the direction filtering range.

4. The inductance product package detection system of claim 1, wherein, The texture feature data of the inductor product is obtained according to the original image, comprising: a ROI region of the inductor product in the original image is obtained according to the contour feature data; statistical analysis is performed on the gray value in the ROI region to obtain the texture feature data.

5. The inductance product package detection system of claim 1, wherein, The packaging defect detection result comprises a glue amount uniformity detection result; the pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data to obtain the packaging defect detection result, comprising: the packaging image is obtained, and the packaging image corresponds to a preset positioning center; a first detection region in the packaging image is obtained based on the preset positioning center and the contour feature data, wherein the range of the first detection region is obtained by expanding the contour represented by the contour feature data; the gray value of the first detection region is counted to obtain a gray histogram of the first detection region; the kurtosis of the gray histogram is calculated, and the glue amount uniformity detection result is obtained according to the comparison between the kurtosis and a preset kurtosis threshold.

6. The inductance product package detection system of claim 1, wherein, The packaging defect detection result comprises a bubble defect detection result; the pixel statistics and comparison analysis on the packaging image based on the contour feature data and the texture feature data to obtain the packaging defect detection result, comprising: the packaging image is obtained, and the packaging image corresponds to a preset positioning center; a second detection region in the packaging image is obtained based on the preset positioning center and the contour feature data, wherein the range of the second detection region is obtained by expanding the contour represented by the contour feature data. A uniform grid is divided in the second detection area, and a plurality of windows are randomly selected in each grid to obtain a plurality of detection windows; Statistical analysis is performed on the gray value in each detection window to obtain the texture detection feature of each detection window; The texture detection feature and the texture feature data are compared to obtain the number of abnormal windows; According to the comparison between the number of abnormal windows and the preset number threshold, a bubble defect detection result is obtained.

7. The inductance product package detection system of claim 1, wherein, The packaging detection device further comprises a third station for dispensing and packaging the inductor product; and the packaging defect detection result comprises an overflow detection result; Based on the contour feature data and the texture feature data, pixel statistics and comparison analysis are performed on the packaging image to obtain the packaging defect detection result, including: An packaging image is obtained, and the packaging image corresponds to a preset positioning center; Based on the preset positioning center, a third detection area in the packaging image is obtained according to the contour feature data, wherein the range of the third detection area is obtained based on the contour feature data. The dispensing pressure and the glue nozzle moving speed of the third station are obtained, and a glue area threshold is obtained according to the dispensing pressure and the glue nozzle moving speed; Based on the gray value of the third detection area, the number of pixels in the third detection area that can represent the glue area is counted; According to the comparison between the number of pixels and the glue area threshold, an overflow detection result is obtained.

8. A method of detecting the packaging of an inductive product, characterized in that, The packaging detection system applied to the inductor product includes a packaging detection device with a first station and a second station, and a visual detection device, the visual detection device is used to shoot an original image of the inductor product at the first station, and shoot a packaging image of the inductor product at the second station, the inductor product is based on the original image for feeding and positioning at the first station, and after being packaged by the packaging detection device, the inductor product is based on the packaging image for packaging defect detection and discharging at the second station, the system further includes a control module, which is used to run a packaging detection method of the inductor product, the method includes: According to the original image, contour feature data of the inductor product is obtained; According to the contour feature data, positioning indication data of the first station is obtained; According to the original image, texture feature data of the inductor product is obtained; Based on the contour feature data and the texture feature data, pixel statistics and comparison analysis are performed on the packaging image to obtain a packaging defect detection result; According to the packaging defect detection result, discharging indication data is obtained.

9. An electronic device, comprising: It includes: A memory and a processor; The memory is used to store a program, and the processor is used to execute the steps of the packaging detection method of the inductor product in claim 8 when executing the program.

10. A computer-readable storage medium, characterized in that, A computer readable program or instruction is used to store, which can realize the steps of the packaging detection method of the inductor product in claim 8 when the processor executes the program.

Images