Intelligent visual detection method on inductor hot-pressing production line and related equipment

Abstract

The invention provides an intelligent visual inspection method on an inductor hot-pressing production line and related equipment, and relates to the field of quality control of automatic production lines. The method comprises the steps that image information of various inductors of different models is obtained firstly, the image information comprises boundary dimensions and surface defect information, and after preprocessing, a quality judgment model is constructed by combining a deep learning algorithm and qualified inductor image information; and after receiving a detection starting instruction of the display end, shooting a to-be-detected inductor by using a plurality of image acquisition devices pre-deployed in a key area of the production line to obtain a to-be-analyzed image, and inputting the to-be-analyzed image into the model to judge the quality of the inductor. And if the detection result is not matched with the quality threshold value of the qualified inductor of the corresponding model, an unqualified prompt message is sent to the display end, so that efficient and accurate detection of the inductor quality is realized.

Description

Technical Field

[0001] The present application relates to the field of quality control of automated production lines, and in particular to an intelligent visual inspection method and related equipment on an inductive hot pressing production line. Background Art

[0002] In the field of modern electronic industry production, inductors, as key electronic components, play an indispensable role in many electronic devices. Quality control of inductor hot pressing production lines is extremely important to ensure the performance and reliability of inductors.

[0003] At present, inductor detection mainly adopts manual visual detection and some simple automatic detection equipment. Among them, the simple automatic detection equipment measures and compares certain key dimensions of the inductor based on simple optical principles or dimension measurement tools.

[0004] However, existing detection methods face problems such as low recognition accuracy and poor adaptability in complex environments, and are difficult to meet the high-quality detection needs on high-speed production lines, especially in terms of inductor size consistency and surface defect identification. There is a problem of low efficiency. Summary of the invention

[0005] The present application provides an intelligent visual inspection method and related equipment on an inductor hot pressing production line, which are used to efficiently and accurately detect the quality of the inductor, overcoming the problems of low efficiency, large errors and difficulty in adapting to changes in complex production environments existing in traditional inspection methods.

[0006] In the first aspect, the present application provides an intelligent visual inspection method on an inductor hot pressing production line, which is applied to a detection device, and the method includes: obtaining inductor image information of multiple existing inductors of different models, and the inductor image information includes external dimension information and surface defect information; after preprocessing the inductor image information, constructing an inductor quality judgment model based on the inductor image information and existing qualified inductor image information in combination with a deep learning algorithm; after receiving a start-up detection function sent by a display end, photographing the inductor to be tested through multiple image acquisition devices to obtain an inductor image to be analyzed, and the image acquisition device is pre-deployed in a preset key area of ​​the inductor hot pressing production line; inputting the inductor image to be analyzed into the inductor quality judgment model to judge the inductor quality detection result of the inductor to be tested; if the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, sending an unqualified prompt information to the display end.

[0007] By adopting the above technical solutions, we can take advantage of the existing image information and deep learning algorithms to accurately build models and greatly reduce the workload and errors of manual inspection. In the inductor hot pressing production line, we can quickly and accurately judge the quality of inductors, find unqualified products in time, improve the inspection efficiency and quality, and ensure the consistency of products.

[0008] In combination with some embodiments of the first aspect, in some embodiments, after receiving the start detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices, and before the step of obtaining the inductor image to be analyzed, it also includes: communicating with the main controller of the hot pressing production line through a high-speed synchronous trigger module to obtain the inductor movement speed and the estimated residence time of the inductor to be tested in the detection area; dynamically adjusting the exposure time and image storage time of the image acquisition device according to the inductor movement speed and the estimated residence time.

[0009] By adopting the above technical solution, the image acquisition device can be adaptively adjusted according to the actual movement of the inductor on the production line. In the fast production process, it is ensured that each inductor to be tested can be photographed at a suitable exposure and storage time to avoid image blur or information loss, thereby improving the quality of the collected inductor image to be analyzed, making the subsequent quality inspection results more accurate and reliable.

[0010] In combination with some embodiments of the first aspect, in some embodiments, if the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, then after the step of suspending the detection work and sending an unqualified prompt message to the display terminal, it also includes: periodically obtaining the current inductor quality detection result and production line operation data; judging whether there is a preset deviation between the inductor quality detection result and the historical record under the production line operation data, and the production line operation data at least includes the materials used for inductor production and the production environment data; if the preset deviation exists, generating a parameter adjustment instruction according to the deviation analysis algorithm; and adjusting the parameters in the inductor quality judgment model according to the parameter adjustment instruction.

[0011] By adopting the above technical solution, during the production process, the model can be continuously optimized according to the actual production situation to adapt to the fluctuation of inductor quality caused by material or environmental changes. Even if the production conditions change, the accuracy of quality inspection can be guaranteed, the misjudgment caused by the fixed model can be reduced, and the adaptability of the entire inspection system to different production environments can be improved.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after receiving the startup detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices to obtain the inductor image to be analyzed, and it also includes: using an adaptive threshold segmentation algorithm to pre-process the inductor image to be analyzed to extract the inductor target to be analyzed; and using an image registration algorithm to register the inductor target to be analyzed to obtain the inductor image under different viewing angles.

[0013] By adopting the above technical solutions, the adaptive threshold segmentation algorithm can accurately separate the inductor target from the complex background, avoiding irrelevant information from interfering with subsequent analysis. The image registration algorithm can integrate image information from different perspectives to more comprehensively present the appearance and details of the inductor. The combination of the two makes the analysis of the inductor more comprehensive and accurate, and the detection of both surface defects and external dimensions can be more accurate, thereby improving the accuracy of inductor quality detection.

[0014] In combination with some embodiments of the first aspect, in some embodiments, after receiving the startup detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices, and before the step of obtaining the inductor image to be analyzed, it also includes: after calibrating the image acquisition device, determining the lighting conditions of multiple preset key areas corresponding to the image acquisition devices through environmental sensors; adjusting the polarization filter on the image acquisition device according to the lighting conditions to reduce the impact of light on the inductor image to be analyzed.

[0015] By adopting the above technical solutions, the calibration equipment ensures the initial accuracy of image acquisition, and the environmental sensor's monitoring of lighting conditions provides a basis for the adjustment of the polarization filter. Under different lighting conditions, the polarization filter can effectively reduce the impact of reflected light and scattered light on image quality, ensuring that the collected inductance image to be analyzed is clear and stable, and avoiding image distortion caused by problems such as uneven or excessive lighting.

[0016] In combination with some embodiments of the first aspect, in some embodiments, if the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, then after the step of sending unqualified prompt information to the display end, it also includes: obtaining unqualified image samples of unqualified inductors through the image acquisition device; and inputting the unqualified image samples as new training samples into the inductor quality judgment model for updating.

[0017] By adopting the above technical solutions, the model can continuously learn new unqualified situations and enrich the model's sample library. As the samples are increased and updated, the model's ability to identify the characteristics of various types of unqualified inductors will be enhanced, and it will be able to more accurately judge the quality of inductors, reduce missed detections caused by new types of unqualified products, and improve the accuracy and stability of the model in the long-term production process.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after preprocessing the inductor image information, after the step of constructing an inductor quality judgment model based on the inductor image information and the existing qualified inductor image information in combination with a deep learning algorithm, it also includes: during the inductor hot pressing forming process, obtaining the inductor forming dynamic image frame by frame by the image acquisition device; judging whether the inductor forming dynamic image matches the preset inductor forming dynamic image standard; if not, judging that the inductor forming process is abnormal, sending an abnormal alarm and the corresponding inductor forming dynamic image to the display end.

[0019] By adopting the above technical solution, dynamic images can fully reflect the changes in the molding process. By comparing with the preset standards, problems can be quickly located, allowing operators to take timely measures to make adjustments, avoid the production of a large number of unqualified products, and ensure the stability of the production process and product quality.

[0020] In a second aspect, the present application provides a detection device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the detection device to perform the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, the present application provides a computer-readable storage medium comprising instructions, which, when executed on a detection device, causes the detection device to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, the present application provides a computer program product, which, when executed on a detection device, enables the detection device to execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The inductor quality judgment model is constructed based on a deep learning algorithm, and the image information of inductors of various models is obtained for model training, and the image of the inductor to be tested is collected for detection after receiving the start-up detection function. The problem of large workload and high error in manual inductor detection in the existing technology is effectively solved, thereby realizing the technical effect of quickly and accurately detecting the quality of inductors on the inductor hot pressing production line, timely discovering unqualified products, improving detection efficiency and quality, and ensuring product consistency.

[0024] 2. The high-speed synchronous trigger module is used to communicate with the main controller of the hot pressing production line to obtain the moving speed and dwell time of the inductor, so as to dynamically adjust the exposure time and image storage time of the image acquisition device. This effectively solves the problem of image blur or information loss caused by the rapid movement of the inductor on the production line in the prior art, thereby achieving the technical effect of improving the image quality of the inductor to be analyzed and making the subsequent quality inspection results more accurate and reliable.

[0025] 3. After the inductor quality test result fails, the current result and production line operation data are periodically obtained, the deviation of historical records is compared and the parameters of the inductor quality judgment model are adjusted accordingly. Therefore, the problem that the model is difficult to adapt to the fluctuation of inductor quality due to changes in production conditions in the existing technology is effectively solved, thereby achieving the technical effect of improving the adaptability of the entire detection system to different production environments, ensuring the accuracy of quality detection and reducing misjudgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of an intelligent visual inspection method on an inductive hot pressing production line in an embodiment of the present application; Figure 2 It is another flow chart of the intelligent visual inspection method on the inductive hot pressing production line in the embodiment of the present application; Figure 3 It is a schematic diagram of the structure of a physical device of the detection device in the embodiment of the present application. DETAILED DESCRIPTION

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more listed items.

[0028] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0029] For ease of understanding, the following is a description of the process of the method provided by this implementation. Figure 1, which is a flow chart of the intelligent visual inspection method on the inductive hot pressing production line in an embodiment of the present application.

[0030] S101, obtaining inductor image information of a plurality of existing inductors of different models, the inductor image information including shape size information and surface defect information; The detection equipment first needs to connect to a database or image warehouse that stores a large amount of inductor product image information. The database collects pictures of inductor products of various models. The detection equipment can read the pixel matrix information of each inductor product picture in the database and use image processing methods to parse and extract the key external dimension parameters of the inductor. For example, the detection equipment can mark the length and width coordinate points of the inductor picture, and calculate the length and width of the inductor according to the coordinate point positions; the detection equipment can also extract the contour lines of the inductor image, and analyze and extract the external dimensions of the inductor product according to the geometric shape formed by the pixel distribution of the contour lines; in addition, the detection equipment can mark the contour lines of specific parts on the inductor image, such as the inner and outer contours of the inductor coil, and calculate the thickness parameters of the inductor coil according to the contour area. By parsing the pixel information of the inductor image, the detection equipment can efficiently and automatically obtain the external dimension parameters of the inductor.

[0031] In addition to the dimensional information, the testing equipment also needs to analyze whether there are defects on the surface of the inductor. After obtaining the image of the inductor product, the testing equipment can use image enhancement, filtering and other algorithms to improve the image quality, and then use edge detection and other algorithms to mark the edge contours of various cracks, scratches and other edge contours on the surface of the inductor, and judge whether there are surface defects based on the edge features. For some lighter scratches, the testing equipment can also use image segmentation algorithms to extract areas of different colors on the surface of the inductor to determine the location of the defect. By automatically parsing the pixel information of the inductor image, the testing equipment can obtain the inductor's dimensional and surface defect information, obtain image information containing different models of inductors, and provide sufficiently rich sample data for the subsequent construction of the judgment model.

[0032] S102, after preprocessing the inductor image information, construct an inductor quality judgment model according to the inductor image information and existing qualified inductor image information in combination with a deep learning algorithm; After obtaining the inductor image information including the dimensions and surface defects, the inspection equipment needs to use image processing algorithms to pre-process the image to improve the quality. This can include using the median filter method to reduce image noise points; using histogram equalization to improve image contrast, making the inductor edges and defect areas clearer and easier to detect; using grayscale processing to eliminate color interference and highlight the outline of the inductor. In addition, the inspection equipment can also crop and rotate the inductor image to adjust it to a uniform shape and direction to facilitate subsequent feature extraction and judgment.

[0033] After preprocessing the image, the detection equipment will use deep learning models such as convolutional neural networks to build a judgment system. The detection equipment can input preprocessed images of inductors of different models and corresponding qualified inductor images into the convolutional neural network for training. Through training, the network can gradually adjust the convolution kernel parameters, learn to extract the key features of the inductor, and integrate the qualified standards of different types of inductors into a unified judgment system. After a large number of training samples, the convolutional neural network can form an efficient inductor quality judgment model that can quickly judge the qualification of the input inductor image. By building a judgment model, the detection equipment can automatically and efficiently perform quality inspection on the inductor.

[0034] In some embodiments, when the inductor enters the hot pressing forming process, the detection device instructs the image acquisition device (high-speed industrial camera) to start dynamic shooting of the inductor forming process. The image acquisition device shoots the inductor forming process frame by frame at multiple frames per second to obtain a dynamic video of inductor forming containing multiple frames of images. After the detection device obtains the dynamic video of inductor forming, it uses an image processing algorithm to extract each frame of the image to obtain a multi-phase image of inductor forming. The detection device compares the inductor multi-phase forming image with the preset standard inductor forming dynamic image to determine whether it matches. The standard inductor forming dynamic image contains normal images of the inductor at each stage of the forming process in chronological order, which can be used as a standard template for the inductor forming process. If there are unmatched frames on the time axis of the extracted inductor forming multi-phase image and the standard dynamic image, it can be judged that the inductor forming process is abnormal. When the abnormality is detected, the detection device immediately sends an abnormal alarm to the display end, and at the same time sends the corresponding inductor forming dynamic image frame to display the abnormal details. After receiving the abnormal alarm, the display end can immediately notify the operator to check and adjust the inductor forming process to avoid the generation of a large number of defective products.

[0035] S103, after receiving the start detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices to obtain the inductor image to be analyzed, and the image acquisition devices are pre-deployed in the preset key area of ​​the inductor hot pressing production line; The detection equipment is connected to a human-computer interaction interface of the display end. When the operator chooses to start the automatic detection function on the display end, the detection equipment can receive a start signal. After receiving the signal, the detection equipment will instruct multiple image acquisition devices configured on the production line to take pictures of the inductor to be tested conveyed by the conveyor belt. These image acquisition devices, such as high-speed industrial cameras, will be arranged in key positions of the production line in advance, such as the inspection section after the inductor forming process, to ensure that the complete inductor product can be photographed. The detection equipment can synchronously control the configured multiple cameras to capture images of the inductor to be tested from different angles and transmit the images to the detection equipment end. After receiving the inductor pictures taken from multiple perspectives, the detection equipment can use image processing algorithms to crop and correct the pictures, retain the image area of ​​the inductor product, remove irrelevant content such as the background, and obtain a clear and complete image of the inductor to be tested as a sample image for subsequent quality inspection.

[0036] In some embodiments, after obtaining the inductor image to be analyzed, the detection device can also traverse each small area of ​​the inductor image to be analyzed, calculate the average value and standard deviation of the pixel values ​​of the small area, and use the average value minus a weight coefficient multiplied by the standard deviation as the threshold of the area, and use the adaptive threshold to binarize the small area, convert pixels above the threshold to white, and convert pixels below the threshold to black, repeat the above steps, traverse each small area of ​​the image, and obtain the binarization result of the entire image. In the binary image, the main part of the inductor product is converted to white, while the background and other debris are converted to black. In this way, the target area of ​​the inductor product can be separated, the image area of ​​the inductor product marked as white can be extracted, the black background and debris parts can be filtered out, and a clear target image of the inductor to be analyzed can be obtained.

[0037] It is also possible to obtain images of inductor products taken by multiple image acquisition devices at different positions and different viewing angles in advance, extract the inductor target area in each image using the above method, obtain a clear image containing the inductor product, and then input the inductor images at different viewing angles into the image registration algorithm for processing. The algorithm can calculate the common feature points in different images and determine the geometric transformation relationship between different views of the same inductor. According to the geometric relationship, the inductor images of different viewing angles are coordinate transformed and aligned, so that different views of the same inductor are mapped to a common coordinate system, and finally the inductor images that have been registered and aligned at multiple viewing angles are obtained. The positions and directions of the parts corresponding to the same inductor in these images are unified and coincident. These registered and aligned multi-view inductor images can provide data support for subsequent modeling and analysis.

[0038] In some embodiments, before this step, before deploying the image acquisition device to the production line, the detection device needs to be calibrated to ensure the accuracy of image acquisition. Calibration may include the following steps: using a calibration plate containing characteristic patterns such as black and white chessboards, placing it within the shooting range of the image acquisition device, ensuring that the calibration plate completely enters the framing range, and the detection device controls the image acquisition device to shoot the calibration plate picture. The detection device then uses edge detection, corner point detection and other algorithms to automatically parse the characteristic points of the calibration plate in the picture, such as the four corners of the chessboard, intersections, etc. The extracted characteristic point coordinates are the calibration points. The detection device combines the calibration point coordinates, uses algorithms such as Zhang Zhengyou calibration method, calculates the internal parameters of the image acquisition device, including focal length, optical center, distortion coefficient, etc., obtains the stereoscopic projection model of the image acquisition device, uses the verification calibration plate to shoot the sample picture, extracts the characteristic point coordinates of the detection device, substitutes them into the stereoscopic projection model to calculate their position in three-dimensional space, and compares them with the actual position to verify the calibration accuracy. Through the above-mentioned method of calibrating checkerboard feature points, the internal parameters of the image acquisition device can be accurately obtained, and a mapping model from two-dimensional images to three-dimensional actual scenes can be established, providing a basis for subsequent inductive image acquisition and analysis.

[0039] After calibrating the image acquisition device, the detection device needs to monitor the lighting conditions of multiple key areas of the production line through environmental sensors. The environmental sensors here can be illuminometers that can detect the light intensity in real time and convert it into digital signal output. The detection device will arrange the environmental sensor at a position corresponding to the image acquisition device to monitor the changes in the lighting conditions of the detection area where the image acquisition device needs to shoot the inductance. After obtaining the lighting condition data, the detection device can determine whether there is excessive light or reflection in the area based on the light intensity value. If the light exceeds a certain threshold, it indicates that there is strong light interference in the area. The detection device will adjust the polarization filter angle or polarization direction of the front end of the image acquisition device corresponding to the area accordingly. The polarization filter can effectively filter out the reflected light and scattered light entering the lens. By adjusting the direction of the polarization filter to be perpendicular to the polarization direction of the strong light, the influence of the reflected light can be greatly reduced. The detection device can adjust the polarization filter to the optimal direction to reduce the light interference. This can effectively reduce the impact of changes in lighting conditions on the captured inductor images, ensure that the collected inductor images are clear and have high contrast, facilitate subsequent image analysis and quality inspection, and avoid a decrease in detection accuracy due to uneven lighting.

[0040] S104, inputting the inductor image to be analyzed into the inductor quality judgment model to judge the inductor quality detection result of the inductor to be tested; After obtaining the image sample of the inductor to be tested, the testing equipment needs to input the sample image into the pre-built inductor quality judgment model for testing. The judgment model is a model obtained by the testing equipment through deep learning algorithm training, which contains multiple convolution layers, pooling layers and other hierarchical structures. Specifically, the testing equipment will send the image of the inductor to be tested as the input layer data into the first convolution layer of the model. The first convolution layer is set with multiple convolution kernels with different parameters. These convolution kernels will slide on the inductor image and perform convolution operations with the image pixels to extract the underlying features of the image, such as image edges, lines, corners and other low-level features. After the first convolution layer extracts the underlying features, the feature map will be input into the second convolution layer. The second convolution layer can continue to extract higher-level features based on the first layer features, such as the shape, size, contour and other structured features of the inductor coil area. With the stacking of convolution layers, the model can continuously extract highly abstract feature representations, gradually represent the key parts of the inductor, and judge its shape, size, surface defects and other information. Finally, the high-level features will enter the fully connected layer, which can integrate all features and map them into output categories representing different quality levels. In this way, the testing equipment can obtain the quality test results of the inductor to be tested from the result output of the judgment model.

[0041] S105: If the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, a non-qualified prompt message is sent to the display terminal.

[0042] After obtaining the test results of the inductor to be tested, the testing equipment needs to find the qualified standard threshold of the quality of the inductor model according to the specific model of the inductor. For example, the qualified standard of a certain type of toroidal inductor is that the quality test result needs to reach the "good" level. If the testing equipment determines that the result of the inductor to be tested only reaches the "general" level, it means that its quality assessment result does not match the qualified threshold. At this time, the testing equipment can immediately send a prompt message of unqualified test results to the human-computer interaction interface of the display end, and can also feed back the image and model information of the unqualified inductor to the display end for the operator to view and analyze. After receiving the feedback of the unqualified prompt, the display end can let the operator pause the production line and stop the outflow of unqualified inductors. In this way, the testing equipment detects quality problems through the judgment model and promptly feeds back to the display end, which can prevent defective products from going offline, thereby improving production quality and output.

[0043] In some embodiments, when the inductor quality detection result shows that the inductor to be tested is unqualified, the detection device will determine the corresponding inductor image as an image sample of unqualified inductor, and the detection device will extract the key feature information of the unqualified inductor image sample, including the shape size parameters and the surface defect area. These feature information can be obtained by comparing with the qualified inductor image in the database. The extracted key feature information of the unqualified inductor is digitized and vectorized to generate a feature vector describing the unqualified inductor, and the feature vector and the unqualified inductor image are used as new training samples and input into the constructed inductor quality judgment model. The inductor quality judgment model updates the model parameters through the back propagation process of the new sample, so that the convolution kernel is more adapted to the unqualified features of this type. With the increase in the number of unqualified samples, the accuracy and robustness of the inductor quality judgment model in the detection of various types of unqualified inductors are improved, thereby reducing the missed detection rate and improving the detection performance. In summary, by continuously acquiring unqualified samples and using them for model incremental training, the inductor quality judgment model can better adapt to various unqualified situations that may occur in actual production, thereby achieving the technical effect of improving the detection accuracy.

[0044] In the embodiment of the present application, since the inductor image information including size and defects is obtained, the inductor quality judgment system is constructed using a deep learning model, and the real-time inductance on the production line can be detected and judged, automatic and efficient inductor quality control can be achieved, effectively solving the problem of low efficiency of traditional manual quality inspection, thereby improving the quality qualification rate and output of inductor products.

[0045] After combining the above embodiments, the method provided by this embodiment is further described in more detail below. Figure 2 , is another flow chart of the intelligent visual inspection method on the inductive hot pressing production line in an embodiment of the present application.

[0046] S201, communicating with the main controller of the hot pressing production line through a high-speed synchronous trigger module to obtain the moving speed of the inductor and the estimated residence time of the inductor to be tested in the detection area; The detection equipment integrates a high-speed synchronous trigger module, which is a communication interface specifically used for data interaction with industrial equipment. The detection equipment will use the synchronous trigger module to establish a connection through standard industrial communication methods to form a data transmission channel with the main controller of the hot pressing production line. The detection equipment needs to regularly send data reading instructions to the communication interface of the main controller, and then the main controller will feedback the operating parameter data of the current production line in real time. It mainly contains two key information: one is the actual operating speed of the production line conveyor belt, that is, the moving speed of the inductor product when it flows through the conveyor belt; the other is the expected residence time of each inductor type product from entering the image detection area to completely moving out of the detection area. More specifically, the main controller can monitor the operating speed of the entire conveyor belt in real time, and calculate the actual conveying speed of the conveyor belt to the workpiece based on the speed of the conveyor belt motor, belt shrinkage and other factors. This speed data will be regularly fed back to the detection equipment. Then, the speed parameters of the conveyor belt, the size of different types of inductor products and the specific position interval of the detection area on the conveyor belt are obtained at the same time. Based on the above data, the time required for each inductor model from entering to moving out of the detection area can be roughly calculated in advance. And this expected time parameter can also be optimized and updated according to actual detection.

[0047] S202, dynamically adjusting the exposure time and image storage time of the image acquisition device according to the inductor moving speed and the estimated dwell time; After receiving the inductor movement speed data v sent by the processor through the image acquisition device, the detection device calculates the exposure time. The detection device first queries the effective pixel width w of the image sensor, which is usually the factory parameter of the sensor. Then the inductor movement speed data v is converted into the pixel displacement per second d = v / w. Then the minimum exposure e required for inductor image shooting is determined according to the empirical formula, and the ideal exposure time t1 = e / d for clear shooting is calculated. After the image acquisition device captures the inductor image, it will immediately send the image to the detection device for storage. When the detection device receives the image data, it will query the estimated residence time t2 of the current inductor, and then the detection device will reduce the actual storage time by the safety factor k, and the storage time = k * t2. For example, if the residence time is 0.5 seconds, the actual storage time will be set to 0.5 * 0.8 = 0.4 seconds. The detection device will use this time as a limit to ensure that when the inductor completely passes through the detection area, its image has been fully saved and there will be no data loss.

[0048] S203, periodically obtaining current inductor quality test results and hot pressing production line operation data; In this step, the detection equipment will automatically obtain the quality inspection results of the latest batch of inductor products output by the image acquisition device within a preset time period (for example, every 1 hour). The results exist in the form of statistical data of the pass rate or defect rate. The detection equipment also calls the sensor network interface to obtain the environmental parameter data collected in real time by the temperature sensors, humidity sensors and other sensor devices pre-set at the key parts of the hot pressing production line, as well as the production operation data such as the use of raw materials and equipment debugging parameters related to each inductor batch. The acquired inductor quality data and production operation data will be automatically identified and formatted by the detection equipment, and then formed into a data pair with the historical statistical records in the database, ready for the next step of difference analysis and processing.

[0049] S204, determining whether there is a preset deviation between the inductor quality test result and the historical record under the operation data of the hot pressing production line, where the operation data of the hot pressing production line at least includes the materials used in the production of the inductor and the production environment data; The testing equipment will continuously count and record the daily inductor quality test results in the recent period (for example, the past week), including the total number of inductors tested daily, the number of qualified inductors, the number of unqualified inductors, and the classification statistics of the reasons for unqualified inductors. These test results will serve as the basic data for judging whether there is quality deviation. In the same time period corresponding to the inductor quality test results, the testing equipment needs to collect various operating data of the hot pressing production line, including: (1) information on raw materials used in inductor production, such as the material, specification, source batch number, etc. of the inductor wire; (2) parameter information of the inductor production process, such as molding temperature, pressure, time and other parameters; (3) inductor production environment information, such as workshop temperature, humidity, vibration and other data. These operating data are derived from various sensors on the hot pressing production line. The testing equipment obtains these operating data through the communication interface with the main controller of the hot pressing line. The testing equipment compares and analyzes the collected inductor quality test results in the recent period with the hot pressing line operating data in the same period to determine whether the operating data has also changed accordingly during the period when the quality test results fluctuate greatly or change abnormally. For example, if the qualified rate of inductors drops significantly during a certain day, the testing equipment will analyze whether there are any abnormalities in the production materials, process parameters, and environmental data during that period, such as material replacement, temperature fluctuations, humidity exceeding the standard, etc. If there are highly correlated abnormal changes, it can be determined that there is a deviation in the quality test results.

[0050] If the quality results fluctuate but the operating data is stable, it can be determined that the quality deviation may be caused by the testing equipment itself. Through this comparative analysis, the testing equipment can determine the specific cause of the quality result deviation. In order to accurately determine whether there is an abnormal deviation, the testing equipment also needs to preset the allowable fluctuation range of the quality results and operating data. Only when the fluctuation of the quality results exceeds the preset range will it be determined that there is a deviation. The preset range can be determined by historical statistical data.

[0051] If there is a preset deviation, execute step S205; If there is no preset deviation, step S207 is executed.

[0052] S205, generating a parameter adjustment instruction according to the deviation analysis algorithm; When a deviation is detected between the inductor quality test results and the historical records of the hot pressing production line, the testing equipment needs to call the preset deviation analysis algorithm to generate instructions for adjusting the model parameters. First, the testing equipment will collect and statistically analyze the distribution of inductor quality data during the deviation period, such as what types of defects appear most often, the distribution of the number of defects, etc. At the same time, the corresponding production line operation data, such as temperature, pressure, environmental parameters, etc., are collected.

[0053] Then, the detection equipment inputs the inductor quality data and operating data into the deviation analysis algorithm to find the main correlation between the two and determine the operating parameters that have a key impact on the quality results, such as excessive temperature. Next, the algorithm will automatically determine which features of the inductor the quality problem corresponds to based on the inductor image samples that produce deviations, such as dimensional deformation, coil cracks, etc. The algorithm can give the degree of correlation between each feature item and the quality result. Finally, the algorithm will combine the above analysis and give the model parameter items that should be adjusted, such as improving the detection sensitivity of the coil crack feature, and give the direction and amplitude of the adjustment. Finally, detailed instructions for adjusting the model parameters are formed.

[0054] S206, adjusting the parameters in the inductance quality judgment model according to the parameter adjustment instruction; After receiving the parameter adjustment instruction, the detection device will adjust the part of the model related to the instruction. First, locate the model component specified in the instruction, such as the convolution layer related to coil crack discrimination. Then, read the convolution kernel parameters in the convolution layer. According to the adjustment range given by the instruction, modify the specific weights in the convolution kernel, such as increasing the weight of the corresponding coil crack feature to enhance the detection strength of the feature. At the same time, the detection device will adjust the node weights in the fully connected layer related to the specific feature to change the influence of the feature on the model output results.

[0055] By adjusting the combination of convolution kernel and node weight, the corresponding parameters in the model are corrected, thereby improving the effect of the model in distinguishing the quality problem.

[0056] S207: No model adjustment operation is performed.

[0057] If the comprehensive assessment of the detection equipment shows that the deviation is accidental and will not have a long-term impact on the system, the current model parameters will remain unchanged and no adjustments will be made.

[0058] In the embodiments of the present application, due to the use of methods such as dynamically adjusting the parameters of the image acquisition equipment, establishing an inductor quality judgment model, and analyzing the deviation between the quality results and the production data, the quality of the inductor product can be monitored in real time and the causes of quality problems can be quickly located, effectively solving the problems of low efficiency of traditional manual quality inspection and difficulty in quickly tracing the source of the fault, thereby realizing automatic and efficient quality control on the inductor hot pressing production line.

[0059] The following describes the detection device in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of a physical device structure of a detection device in an embodiment of the present application.

[0060] It should be noted that Figure 3 The structure of the detection device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0061] like Figure 3 As shown, the detection device includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method described in the above embodiment. In RAM 303, various programs and data required for system operation are also stored. CPU 301, ROM 302 and RAM 303 are connected to each other through bus 304. Input / output (I / O) interface 305 is also connected to bus 304.

[0062] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0063] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section 309, and / or installed from a removable medium 311. When the computer program is executed by a central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0064] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0065] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0066] Specifically, the detection device of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the intelligent visual detection method on the inductive hot pressing production line provided in the above embodiment is implemented.

[0067] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the detection device described in the above embodiment; or may exist independently without being assembled into the detection device. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the detection device, the detection device implements the intelligent visual detection method on the inductive hot pressing production line provided in the above embodiment.

[0068] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0069] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0070] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. An intelligent visual inspection method for an inductive hot pressing production line, applied to a detection device, characterized in that: The method comprises: Acquire inductor image information of a plurality of existing inductors of different models, the inductor image information including shape dimension information and surface defect information; After preprocessing the inductor image information, building an inductor quality judgment model based on the inductor image information and existing qualified inductor image information in combination with a deep learning algorithm; After receiving the start detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices to obtain the inductor image to be analyzed. The image acquisition devices are pre-deployed in the preset key area of ​​the inductor hot pressing production line; Inputting the inductor image to be analyzed into the inductor quality judgment model to judge the inductor quality detection result of the inductor to be tested; If the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, unqualified prompt information is sent to the display end.

2. The method according to claim 1, characterized in that If the inductor quality detection result does not match the qualified inductor quality threshold of the corresponding model, after the step of suspending the detection work and sending unqualified prompt information to the display terminal, the method further includes: Periodically obtain the current inductor quality test results and hot pressing production line operation data; Determine whether there is a preset deviation between the inductor quality test result and the historical record under the operation data of the hot pressing production line, wherein the operation data of the hot pressing production line at least includes the materials used in the production of the inductor and the production environment data; If the preset deviation exists, generating a parameter adjustment instruction according to the deviation analysis algorithm; The parameters in the inductance quality judgment model are adjusted according to the parameter adjustment instruction.

3. The method according to claim 2, characterized in that Before the step of periodically obtaining the current inductor quality test results and hot pressing production line operation data, it also includes: The high-speed synchronous trigger module is connected to the main controller of the hot pressing production line to obtain the moving speed of the inductor and the estimated residence time of the inductor to be tested in the detection area; The exposure time and image storage time of the image acquisition device are dynamically adjusted according to the inductor moving speed and the estimated dwelling time.

4. The method according to claim 1, characterized in that: After receiving the start detection function sent by the display end, the inductor to be tested is photographed by multiple image acquisition devices to obtain the inductor image to be analyzed, and the following steps are also included: Preprocessing the inductance image to be analyzed by using an adaptive threshold segmentation algorithm to extract the inductance target to be analyzed; The inductance target to be analyzed is registered using an image registration algorithm to obtain inductance images at different viewing angles.

5. The method according to claim 1, characterized in that After receiving the start detection function from the display end, the inductor to be tested is photographed by multiple image acquisition devices to obtain the inductor image to be analyzed, and before the step, the method also includes: After calibrating the image acquisition device, determining the lighting conditions of the preset key areas corresponding to the plurality of image acquisition devices through an environmental sensor; The polarization filter on the image acquisition device is adjusted according to the lighting condition to reduce the influence of the lighting on the inductance image to be analyzed.

6. The method according to claim 1, characterized in that If the inductance quality detection result does not match the qualified inductance quality threshold of the corresponding model, after the step of sending unqualified prompt information to the display terminal, the method further includes: Acquire unqualified image samples of unqualified inductors by using the image acquisition device; The unqualified image samples are input as new training samples into the inductance quality judgment model for updating.

7. The method according to claim 1, characterized in that After preprocessing the inductance image information, and building an inductance quality judgment model according to the inductance image information and existing qualified inductance image information in combination with a deep learning algorithm, the method further includes: During the inductor hot pressing forming process, the image acquisition device acquires frame-by-frame dynamic images of the inductor forming; Determining whether the inductor forming dynamic image matches a preset inductor forming dynamic image standard; If they do not match, when it is determined that the inductor forming process is abnormal, an abnormal alarm and the corresponding inductor forming dynamic image are sent to the display end.

8. A detection device, characterized in that: The detection device comprises: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and the one or more processors call the computer instructions so that the detection device executes the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the detection device, the detection device is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a detection device, the detection device is caused to perform the method according to any one of claims 1 to 7.

Images