An electronic packaging optimization method and system

Abstract

The present invention discloses an electronic packaging optimization method and system. By obtaining the detection data and device density of the packaged product, wherein the detection data includes the image to be detected. Introducing the device density can improve the quality of image acquisition, thereby improving the quality of the detection data, and further improving the accuracy of the quality judgment result. According to the detection data and the device density, obtain the packaging quality result of the packaged product; perform packaging optimization on the packaged product according to the packaging quality result. While improving the accuracy of electronic packaging quality evaluation, packaging optimization can also be carried out in a timely manner.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic packaging, and particularly to an electronic packaging optimization method and system. Background Art

[0002] An electronic device is a very complex system, and the defects and failures in its packaging process are also very complex. In the process of electronic packaging, the factors causing packaging defects and failures are diverse. For example, material composition and properties, packaging design, environmental conditions, and process parameters, etc. will all affect the quality of electronic packaging. Therefore, in order to improve the yield rate, qualification rate of electronic packaging products, and reduce production costs, during the electronic packaging process, it is necessary to detect and evaluate the packaging quality of the packaged products in order to timely discover packaging defects and process them.

[0003] For the existing microelectronic packaging process quality detection devices and methods, the optical image and thermal image of the electronic packaging are collected by an image acquisition device, and the similarity between the optical image and thermal image and the standard image is calculated respectively to conduct quality evaluation. However, in actual operation, it is very difficult to obtain a perfect standard image. Only using the standard image for quality evaluation is very likely to cause misjudgment of qualified products. Moreover, by collecting and comparing the optical image, thermal image with the standard image, generally only the apparent quality of the packaged product can be evaluated, and it is difficult to accurately evaluate or predict the internal defects of the packaged product.

[0004] Therefore, it is necessary to propose an electronic packaging optimization method to improve the accuracy of electronic packaging quality evaluation and conduct packaging optimization in a timely manner. Summary of the Invention

[0005] In view of this, the present invention provides an electronic packaging optimization method and system, which can improve the accuracy of electronic packaging quality evaluation and conduct packaging optimization in a timely manner.

[0006] The specific technical solutions adopted by the present invention are as follows:

[0007] According to an embodiment of the present invention, there is provided an electronic packaging optimization method, including: obtaining the detection data and device density of the packaged product, wherein the detection data includes the image to be detected; obtaining the packaging quality result of the packaged product according to the detection data and the device density; and performing packaging optimization on the packaged product according to the packaging quality result.

[0008] According to another embodiment of the present invention, an electronic packaging optimization system is provided, including: a data acquisition module for acquiring detection data and device density of a packaged product, wherein the detection data includes an image to be detected; a result determination module for obtaining a packaging quality result of the packaged product according to the detection data and the device density; and a packaging optimization module for optimizing the packaging of the packaged product according to the packaging quality result.

[0009] According to another embodiment of the present invention, an electronic packaging optimization device is further provided, including at least one memory and at least one processor; the at least one memory is used for storing computer instructions, and the at least one processor is used for executing at least part of the computer instructions to implement the above method.

[0010] According to another embodiment of the present invention, a computer-readable storage medium is further provided, and the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer runs the above method.

[0011] Beneficial effects:

[0012] (1) An electronic packaging optimization method, which acquires detection data and device density of a packaged product, wherein the detection data includes an image to be detected. Introducing device density can improve the quality of image acquisition, thereby improving the quality of detection data, and further improving the accuracy of the quality judgment result. According to the detection data and the device density, a packaging quality result of the packaged product is obtained; and the packaging of the packaged product is optimized according to the packaging quality result. While improving the accuracy of electronic packaging quality assessment, packaging optimization can also be carried out in a timely manner.

[0013] (2) The detection data further includes at least one of the following: a qualified reference image of the packaged product; an unqualified reference image of the packaged product; packaging process flow data; packaging environment data; packaging material data. It avoids the problem that it is difficult to obtain a perfect qualified reference image, resulting in a large error in the judgment of the quality result based on the qualified reference image. Considering both the qualified reference image and the unqualified reference image can improve the accuracy of the packaging quality result.

[0014] (3) Inputting the packaging process flow data, packaging environment data, packaging material data, and device density into a similarity judgment model and / or a quality judgment model to obtain a packaging quality result of the packaged product. Introducing various detection data in different links. By introducing packaging process flow data, packaging environment data, and packaging material data, the accuracy of the packaging quality result can be further enhanced. At the same time, it can be introduced in different links, and then different judgment results are obtained. The different judgment results are comprehensively judged to obtain the final packaging quality result, further improving the practicality and accuracy. Brief Description of the Drawings

[0015] Figure 1 is an exemplary block diagram of an electronic package optimization system according to some embodiments of the present invention;

[0016] Figure 2 is an exemplary flowchart of an electronic package optimization method according to some embodiments of the present invention;

[0017] Figure 3 is an exemplary diagram for determining a package quality result based on a judgment model according to some embodiments of the present invention;

[0018] Figure 4 is an exemplary diagram of a similarity judgment model according to some embodiments of the present invention;

[0019] Figure 5 is an exemplary diagram for determining a package quality result based on a judgment model according to some embodiments of the present invention. Detailed Description of the Embodiments

[0020] The present invention will be described in detail below with reference to the accompanying drawings and by way of examples.

[0021] To more clearly illustrate the technical solutions of the embodiments of this specification, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some examples or embodiments of this specification. For those of ordinary skill in the art, without creative efforts, this specification can also be applied to other similar scenarios based on these drawings. Unless obvious from the language context or otherwise stated, the same reference numerals in the figures represent the same structure or operation.

[0022] It should be understood that the "system", "device", "unit" and / or "module" used in the present invention is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the said words can be replaced by other expressions.

[0023] As shown in this specification and the claims, unless the context clearly indicates an exception, words such as "a", "an", "one" and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0024] Flowcharts are used in this specification to illustrate the operations performed by the systems according to the embodiments of this specification. It should be understood that the preceding or subsequent operations are not necessarily executed precisely in sequence. On the contrary, the steps can be processed in reverse order or simultaneously. Also, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0025] The present invention provides an electronic packaging optimization method, including the following steps:

[0026] Step 1: Obtain the detection data and device density of the packaged product, where the detection data includes the image to be detected;

[0027] In a specific embodiment, the detection data further includes at least one of the following: the qualified reference image of the packaged product; the unqualified reference image of the packaged product; the packaging process flow data; the packaging environment data; the packaging material data.

[0028] In a specific embodiment, obtaining the device density of the packaged product includes: obtaining the wiring diagram of the packaged product through the packaging control application, and obtaining the device density based on the wiring diagram.

[0029] Step 2: Obtain the packaging quality result of the packaged product according to the detection data and the device density;

[0030] In a specific embodiment, obtaining the packaging quality result of the packaged product according to the detection data and the device density includes: inputting the detection data and the device density into a judgment model to obtain the packaging quality result of the packaged product.

[0031] In a specific embodiment, the judgment model is a machine learning model, where the machine learning model at least includes: a similarity judgment model and a quality judgment model.

[0032] In a specific embodiment, inputting the detection data and the device density into the judgment model to obtain the packaging quality result of the packaged product includes: inputting the image to be detected, the qualified reference image of the packaged product, and the unqualified reference image of the packaged product into the similarity judgment model to obtain the qualified quality similarity and the unqualified quality similarity, and obtaining the qualified product similarity set and the unqualified product similarity set according to the qualified quality similarity and the unqualified quality similarity; inputting the qualified product similarity set and the unqualified product similarity set into the quality judgment model to obtain the packaging quality result.

[0033] In a specific embodiment, in the process of inputting the detection data and the device density into the judgment model to obtain the packaging quality result of the packaged product, it further includes: inputting the packaging process flow data, the packaging environment data, the packaging material data, and the device density into the similarity judgment model and / or the quality judgment model to obtain the packaging quality result of the packaged product.

[0034] Step 3. Optimize the encapsulated product according to the encapsulation quality result.

[0035] An embodiment of the present invention further provides an electronic packaging optimization system, including: a data acquisition module for acquiring detection data and device density of an encapsulated product, where the detection data includes an image to be detected; a result determination module for obtaining an encapsulation quality result of the encapsulated product according to the detection data and the device density; and an encapsulation optimization module for optimizing the encapsulated product according to the encapsulation quality result.

[0036] To enable those skilled in the art to better understand the technical solution of the present invention, the following will be described in combination with a specific real-time scenario.

[0037] Figure 1 It is an exemplary module diagram of an electronic packaging optimization system shown according to some embodiments of the present invention. In some embodiments, the control system 100 for the mobile phone to communicate with the audio device may include a first acquisition module 110, a second acquisition module 120, a first determination module 130, and a second determination module 140. Among them, the first acquisition module 110 and the second acquisition module 120 may correspond to the data acquisition module in the above embodiments, the first determination module 130 may correspond to the result determination module in the above embodiments, and the second determination module 140 may correspond to the encapsulation optimization module in the above embodiments. In the actual implementation process, the division, naming, and function division of the modules are not specifically limited as long as the technical solution of the present invention can be implemented.

[0038] The first acquisition module 110 may be used to acquire detection data of the encapsulated product, and the detection data may include an image to be detected collected based on an image acquisition device; among them, the image to be detected may be an image of a finished or semi-finished product in the electronic packaging process. In some embodiments, the acquisition parameters of the image acquisition device may be determined based on the type of the encapsulated product; the type of the encapsulated product may be determined based on the user's input. For more content about detection data, image acquisition device, image to be detected, acquisition parameters, and type of encapsulated product, please refer to Figure 2 And its related descriptions.

[0039] The second acquisition module 120 may be used to obtain a wiring diagram input by the user through the encapsulation control APP, and determine the device density used to characterize the density of components on the substrate based on the wiring diagram. For more content about the wiring diagram and device density, please refer to Figure 2 And its related descriptions.

[0040] The first determination module 130 can be used to determine the packaging quality result by detecting data and device density. In some embodiments, the first determination module 130 can be further configured to process the detected data and device density based on a judgment model to determine the packaging quality result; the judgment model is a machine learning model. For more information about the judgment model and the packaging quality result, please refer to Figure 2 and its related descriptions.

[0041] The second determination module 140 can be used to determine the packaging optimization method of the packaged product based on the packaging quality result. The packaging optimization method includes: the optimized position of the target component on the substrate. In some embodiments, the packaging optimization method further includes reinstallation, supplementary installation, and adjustment of the packaging scheme; different packaging quality results correspond to different packaging optimization methods. For more information about the packaging optimization method, please refer to Figure 2 and its related descriptions.

[0042] It should be noted that the above descriptions of the electronic packaging optimization system and its modules are only for convenience of description and do not limit this specification to the scope of the examples given. It can be understood that for those skilled in the art, after understanding the principle of the system, they may, without departing from this principle, make any combination of the various modules, or form a subsystem and connect it to other modules. In some embodiments, Figure 1 the first acquisition module 110, the second acquisition module 120, the first determination module 130, and the second determination module 140 disclosed in

[0043] Figure 2 can be different modules in a system, or a module can implement the functions of two or more of the above modules. For example, the various modules can share a storage module, or each module can have its own storage module respectively. Such variations are all within the protection scope of this specification.

[0044]

[0045] is an exemplary flowchart of an electronic packaging optimization method according to some embodiments of the present invention. In some embodiments, process 200 can be executed by a processor.

[0044] Among them, the processor can be used to process data and / or information related to electronic packaging optimization. For example, the processor can process the to-be-detected image, qualified reference image, and unqualified reference image of the packaged product collected by the image acquisition device to obtain the to-be-detected image feature, qualified reference image feature, and unqualified reference image feature. For another example, the processor can determine the packaging quality result based on the detected data and device density; determine the packaging optimization method based on the packaging quality result.

[0045] In some embodiments, the processor may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor may be local or remote. For example, the processor may access information and / or data from a memory, an image acquisition device, and an environmental sensor through a network. For another example, the processor may be directly connected to a memory, an image acquisition device, and an environmental sensor to access information and / or data. In some embodiments, the processor may be implemented on a cloud platform. For example, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, a multi-cloud, etc., or any combination thereof. In some embodiments, the processor may be part of an electronic packaging optimization system. In some embodiments, the processor may also be embedded in a user terminal together with a memory and a packaging control APP. For specific content about the packaging control APP, reference may be made to the relevant description below.

[0046] As Figure 2 shown, process 200 includes the following steps:

[0047] Step 210, obtaining detection data of a packaged product, where the detection data includes a to-be-detected image collected by an image acquisition device; wherein, the to-be-detected image is an image of a packaged finished product or semi-finished product during the electronic packaging process.

[0048] Electronic packaging may refer to the process of assembling an integrated circuit into a final chip product. Simply put, it is to place the integrated circuit die produced by Foundry (wafer foundry) on a substrate that serves as a carrier, lead out the pins, and then fix and package them into a whole. Packaging plays an important role in protecting the chip, enhancing electrothermal performance, and facilitating the assembly of the whole machine.

[0049] The packaged product may refer to the final product formed by electronic packaging. For example, the packaged product may include one or more integrated circuit chips packaged in a suitable packaging form, and the solder pads of the chips are connected to the external pins of the package by wire bonding (WB), tape automated bonding (TAB), and flip chip bonding (FCB) to form a functional electronic component or assembly (single chip module (SCM) and multi-chip module (MCM)), which can also be called a microelectronic packaging product; the corresponding packaging process can be called primary packaging (chip-level packaging). For another example, the packaged product may include installing the microelectronic packaging product together with passive components on a printed circuit board or other substrate to form a component or a whole machine; the corresponding packaging process can be called secondary packaging (board-level packaging). For yet another example, the packaged product may include connecting the product of secondary packaging to a motherboard through layer selection, an interconnect socket, a wire harness cable, or a flexible circuit board to form a three-dimensional package, constituting a complete whole machine system; the corresponding packaging process can be called tertiary packaging (system-level packaging).

[0050] The detection data may refer to the data used for detecting the quality of the packaged product. For example, the detection data may include the image to be detected, the qualified reference image, the unqualified reference image, etc. In some embodiments, the detection data may further include process flow data, environmental data, and packaging material data. For more information about the process flow data, environmental data, and packaging material data, reference can be made to Figure 3 and its related descriptions.

[0051] The image to be detected may refer to the image of the packaged finished product / semi-finished product during the electronic packaging process; the qualified reference image may refer to the image of the qualified packaged finished product / semi-finished product that has been pre-collected and stored for reference; the unqualified reference image may refer to the image of the qualified packaged finished product / semi-finished product that has been pre-collected and stored for reference. In some embodiments, the image to be detected, the qualified reference image, and the unqualified reference image may include optical images and thermal images.

[0052] The packaged finished product may refer to a product that has completed all the required packaging processes and is available for use or sale. The packaged semi-finished product may refer to an intermediate product that has undergone a certain packaging process but has not completed all the packaging and still requires further processing.

[0053] In some embodiments, the processor may acquire the image of the packaged finished product / semi-finished product through an image acquisition device to obtain the detection data. For example, the image to be detected is obtained by taking a picture of the current packaged finished product / semi-finished product through the image acquisition device; the qualified reference image is obtained by acquiring the image of the qualified packaged finished product / semi-finished product of the historical packaging through the image acquisition device; the unqualified reference image is obtained by acquiring the image of the unqualified packaged finished product / semi-finished product of the historical packaging through the image acquisition device.

[0054] Among them, the image acquisition device may refer to a device that can be used to acquire the image of the packaged finished product / semi-finished product. For example, the image acquisition device may include an optical camera, an infrared thermal imager, etc.

[0055] In some embodiments, the acquisition parameters of the image acquisition device may be determined based on the type of the packaged product; the type of the packaged product may be determined based on the user's input.

[0056] Among them, the acquisition parameters may refer to various parameters when the image acquisition device acquires the image of the packaged finished product / semi-finished product. For example, the acquisition parameters may include resolution, focal length, sensitivity, focusing parameters, fill light parameters, shooting angle, shooting distance, etc.

[0057] The type of the packaged product may refer to the category of the packaged product. In some embodiments, the processor may determine the type of the packaged product based on the model (number), purpose, etc. of the packaged product. For example, the processor may classify the packaged products of the same model (number) into the same type. For another example, the processor may determine the packaged products of the same purpose as a major category; further, the packaged products may be divided into more specific minor categories based on the model (number) of the packaged products in each major category.

[0058] In some embodiments, the processor may determine the acquisition parameters of the image acquisition device based on the type of the packaged product. For example, for different types of packaged products, the device density is different. For a packaged product with a large device density, in order to obtain a clearer image, a closer shooting distance is required. For another example, for different types of packaged products, the distribution positions and quantities of their corresponding components are different, and the shooting angles when acquiring images will also be different. In some embodiments, for different types of packaged products, the processor may preset corresponding acquisition parameters in advance.

[0059] In some embodiments, the type of the packaged product may be determined based on user input. For example, the user may input the type of the packaged product currently being packaged through the packaging control APP.

[0060] In some embodiments, the data acquisition device may further include an environmental sensor, and the environmental sensor may be used to acquire environmental data; the acquisition parameters of the image acquisition device are also related to the environmental data; the environmental data may include light intensity, temperature, etc.

[0061] The environmental sensor may refer to an acquisition device for acquiring environmental data. For example, the environmental sensor may include a light sensor, a temperature sensor, a gas detector, etc.

[0062] The environmental data may refer to the environmental state information during the electronic packaging process and when taking images of the packaged finished products / semi-finished products. For example, the environmental data may include packaging environmental data and detection environmental data. Among them, the packaging environmental data refers to the environmental data during the electronic packaging process, which may include temperature, atmospheric environment (oxidation), chemical corrosion conditions, etc.; the detection environmental data is the environmental data when acquiring detection data, which may include light intensity, temperature, etc. In some embodiments, the processor may acquire environmental data based on the environmental acquisition device. For example, the processor may use a light sensor to collect light intensity; use a temperature sensor to collect temperature; use a gas detector to acquire atmospheric environmental data. Specifically, the oxygen concentration in the atmosphere, the concentration of other corrosive molecules, etc. may be acquired.

[0063] In some embodiments, the processor may determine acquisition parameters based on detected environmental data. In some embodiments, the processor may determine the acquisition parameters of an optical camera based on detected environmental data. For example, the fill light parameters and shooting parameters (sensitivity, EV value, etc.) of the optical camera are negatively correlated with the light intensity. In some embodiments, the shooting parameters of the thermal imager are related to the temperature. For example, if the temperature change is not obvious, the acquisition parameters of the thermal imager can be set to be more sensitive.

[0064] In some embodiments of this specification, by collecting environmental data based on environmental sensors, real-time environmental data during acquisition of detection data can be obtained, and the acquisition parameters of the image acquisition device can be adjusted based on the real-time environmental data, which can improve the accuracy and quality of the obtained detection data and provide strong data support for subsequent determination of the encapsulation quality result based on the detection data.

[0065] Step 220: Obtain the wiring diagram input by the user through the encapsulation control APP, and determine the device density used to characterize the density of components on the substrate based on the wiring diagram.

[0066] The encapsulation control APP may refer to an APP used to control the process flow, encapsulation quality, etc. of electronic encapsulation. The encapsulation control APP may be an APP installed in the user terminal used by the user (such as a mobile phone, tablet computer, wired computer, encapsulation control workbench, etc.). Through the encapsulation control APP, the user can obtain data and / or information related to electronic encapsulation. For example, obtain the encapsulation progress, encapsulation quality result, etc. of the encapsulated product. On the other hand, the user can also input data and / or information related to electronic encapsulation through the encapsulation control APP. For example, input the type of the electronic encapsulation product, the corresponding encapsulation material data, process flow data, etc.

[0067] The wiring diagram may refer to the circuit design diagram, component distribution diagram, etc. of the electronic encapsulation product. For example, the wiring diagram can reflect the distribution of components on the substrate and the connection relationship of each pin of the components. For each type of encapsulated product, its corresponding wiring diagram can be determined.

[0068] The device density may refer to the density of components on the substrate. For example, the distance between the pins of the components can be used to measure the density of components on the substrate, that is, the device density.

[0069] In some embodiments, the processor may obtain the wiring diagram input by the user through the encapsulation control APP and determine the device density based on the wiring diagram. For example, the processor may determine the distance between each pin of the component based on the wiring diagram and calculate the average value, and determine the average pin distance as the device density.

[0070] Step 230: Determine the encapsulation quality result based on the detection data and the device density.

[0071] The encapsulation quality result may refer to the data used to evaluate the encapsulation quality of the encapsulated product / semi-finished product. For example, the encapsulation quality result may be a specific quality score, or it may be qualified / unqualified, or it may also be a quality grade (such as first-class product, second-class product, unqualified product, etc.).

[0072] In some embodiments, the processor may obtain the detection data and the device density, and determine the encapsulation quality result. In some embodiments, the processor may determine the encapsulation quality result based on the detection data of the current encapsulated product and the detection data of other encapsulated products with the same device density and the same type as the current encapsulated product. For example, the processor may obtain the detection images of other historical encapsulated products with the same type and the same device density as the current encapsulated product, compare them with the image to be detected of the current encapsulated product, and obtain the encapsulation quality result of the other historical encapsulated product with the highest image similarity as the encapsulation quality result of the current encapsulated product.

[0073] In some embodiments, the processor may process the detection data and the device density based on a judgment model to determine the encapsulation quality result; wherein, the judgment model is a machine learning model. For more content on determining the encapsulation quality result based on the judgment model, reference can be made to Figure 3 and its related descriptions.

[0074] Step 240, determining the encapsulation optimization method of the encapsulated product based on the encapsulation quality result, wherein the encapsulation optimization method includes: the optimized position of the target component on the substrate.

[0075] The encapsulation optimization method may refer to the method of how to optimize the encapsulation of the encapsulated product. For example, the encapsulation optimization method may include: the optimized position of the target component on the substrate. Among them, the target component may refer to the component that needs to be encapsulated and optimized, and its optimized position on the substrate may include the specific pin positions that need to be encapsulated and optimized. In some embodiments, the optimized position may be determined by identifying the image to be detected collected based on image recognition technology.

[0076] In some embodiments, the processor may determine the encapsulation optimization method of the encapsulated product based on the encapsulation quality result. For example, the processor may determine the corresponding encapsulation optimization method for different encapsulation quality results in advance. In some embodiments, the encapsulation optimization method is also related to the type of the encapsulated product. For different types of encapsulated products, even if the encapsulation quality results are the same, their corresponding encapsulation optimization methods are also different. Therefore, the processor may also determine the corresponding encapsulation optimization methods for different types and different encapsulation quality results of the encapsulated products in advance.

[0077] In some embodiments, the encapsulation optimization method may further include reinstalling, supplementing, and adjusting the encapsulation scheme; different encapsulation quality results may correspond to different encapsulation optimization methods.

[0078] Among them, reinstalling may refer to reinstalling the encapsulated finished products / semi-finished products whose encapsulation quality results meet the preset requirements. For example, the encapsulation of an encapsulated product with the number of encapsulation defects exceeding the first preset requirement (such as 5 apparent defects) is removed and reinstalled. Supplementing may refer to encapsulating the positions where components are missed during the encapsulation process (such as a certain pin of a certain component not being connected).

[0079] Adjusting the encapsulation scheme may refer to adjusting the original encapsulation design scheme. For example, optimizing the encapsulation process flow, replacing encapsulation materials, etc.

[0080] In some embodiments, different encapsulation quality results may correspond to different encapsulation optimization methods. In some embodiments, the processor may determine the specific defect conditions of the encapsulation based on the encapsulation quality results, and determine the encapsulation optimization method based on the specific defect conditions. For example, the encapsulation quality results can be divided into different quality grades (such as first-class products, second-class products, non-conforming products, etc.); for non-conforming products, the corresponding defect condition is that the number of encapsulation defects exceeds 5, and the corresponding encapsulation optimization method can be determined as reinstalling; for second-class products, the corresponding defect condition is that the number of encapsulation defects is greater than 0 and less than 5, and the corresponding encapsulation optimization method is supplementing. Another example is that the processor may determine whether to adjust the encapsulation scheme based on the proportion of non-conforming products among the multiple encapsulation quality results corresponding to multiple encapsulation products. Exemplarily, it is assumed that when the proportion of non-conforming products is greater than 2%, the processor may adjust the encapsulation scheme. It should be noted that the aforementioned number of defects, proportion of non-conforming products, etc. can be adjusted based on the actual production situation, and the examples here are only for illustration and are not intended to limit the scope of this specification.

[0081] In some embodiments, the encapsulation quality result may be determined based on the first encapsulation quality result and the second encapsulation quality result. Among them, the first encapsulation quality result and the second encapsulation quality result are similar to the encapsulation quality result, and may refer to data used to evaluate the encapsulation quality of encapsulated products / semi-finished products, which can be represented by quality scores, quality grades, etc. Among them, the first encapsulation quality result is the encapsulation quality result determined by processing the first set of qualified product similarity, the first set of non-conforming product similarity, device density, process flow data, environmental data, and encapsulation material data based on a quality judgment model; the second encapsulation quality result is the encapsulation quality result determined by processing the second set of qualified product similarity and the second set of non-conforming product similarity based on a quality judgment model. Specifically, reference can be made to Figure 3 、 Figure 4 、 Figure 5 and their related descriptions.

[0082] In some embodiments, the processor may determine a first package quality result and a second package quality result based on a judgment model. For details, refer to Figure 4 、 Figure 5 and its related descriptions.

[0083] In some embodiments, the processor may use various methods to determine the package quality result based on the first package quality result and the second package quality result. For example, the processor may respectively determine the accuracy of the first package quality result and the accuracy of the second package quality result (such as the confidence level output by the judgment model), and determine the higher accuracy one as the final package quality result. For another example, the processor may perform weighted summation, averaging, etc. on the first package quality result and the second package quality result to determine the final package quality result.

[0084] In some embodiments of this specification, by comprehensively determining the final package quality result based on the first package quality result and the second package quality result, the accuracy of the determined package quality result can be improved, providing reliable data support for subsequent determination of package optimization methods.

[0085] In some embodiments, the optimization position may be determined based on a judgment model.

[0086] In some embodiments, the processor may use the judgment model to process the detection data and device density to determine the package quality result and the optimization position. For example, the judgment model may output the position with the lowest similarity between the image to be detected and the qualified reference image and the highest similarity between the image to be detected and the unqualified reference image as the optimization position. In some embodiments, in order to enable the judgment model to output the optimization position while determining the package quality result, a training label: sample optimization position may be added during the training of the judgment model. The sample optimization position refers to the position in the sample detection data (image to be detected) that needs package optimization, which can be determined by manual annotation. For details of the training of the specific model, refer to Figure 3 and its related descriptions.

[0087] In some embodiments of this specification, by automatically determining the optimization position of the target component on the substrate through the judgment model, the accuracy and efficiency of determining the optimization position can be improved.

[0088] In some embodiments of this specification, by obtaining the detection data and device density, determining the package quality result, and determining the package optimization method of the package product based on the package quality result, the efficiency and accuracy of package quality evaluation in the electronic packaging process can be improved, so as to timely process and optimize the unqualified packaging situation, and improve the automation level and efficiency of electronic packaging production.

[0089] Figure 3It is an exemplary schematic diagram for determining the encapsulation quality result based on a judgment model as shown in some embodiments of the present invention.

[0090] In some embodiments, the processor may process the detection data and the device density based on the judgment model to determine the encapsulation quality result; wherein, the judgment model is a machine learning model.

[0091] In some embodiments, the judgment model may include a similarity judgment model and a quality judgment model. The similarity judgment model may determine a first qualified quality similarity and a first unqualified quality similarity based on the image to be detected, the qualified reference image, and the unqualified reference image. The quality judgment model may determine a first encapsulation quality result based on the first set of qualified product similarities, the first set of unqualified product similarities, and the device density; wherein, multiple first qualified quality similarities form the first set of qualified product similarities, and multiple first unqualified quality similarities form the first set of unqualified product similarities.

[0092] In some embodiments, the judgment model may include one or more of Convolutional Neural Networks (CNN), Deep Neural Networks (DNN), Recurrent Neural Network (RNN), or other custom networks.

[0093] In some embodiments, as Figure 3 shown, the judgment model may include a similarity judgment model 313 and a quality judgment model 322. Among them, the network structures of the similarity judgment model 313 and the quality judgment model 322 may include one or more of CNN, DNN, RNN, or other custom networks.

[0094] In some embodiments, the input of the similarity judgment model 313 may be the image to be detected 310, multiple qualified reference images 311, and multiple unqualified reference images 312, and the output may be multiple first qualified quality similarities 314 and multiple first unqualified quality similarities 315. It should be noted that the image to be detected 310, the qualified reference images 311, and the unqualified reference images 312 input into the similarity judgment model 313 are collected from encapsulation products of the same type and at the same encapsulation stage (process).

[0095] In some embodiments, the processor may input the image to be detected 310 and a qualified reference image 311 into the similarity judgment model 313 each time, and obtain a corresponding first qualified quality similarity 314; through multiple inputs, multiple first qualified quality similarities 314 are obtained. Similarly, the processor may also input the image to be detected 310 and multiple unqualified reference images 312 into the similarity judgment model 313 in sequence, and respectively obtain corresponding multiple first unqualified quality similarities 315. Among them, for the content of the image to be detected 310, the qualified reference image 311, and the unqualified reference image 312, reference can be made to Figure 2 and its related description.

[0096] In some embodiments, multiple first qualified quality similarities 314 may form a first qualified product similarity set 316, and multiple first unqualified quality similarities 315 may form a first unqualified product similarity set 317.

[0097] In some embodiments, the input of the quality judgment model 322 may be the first qualified product similarity set 316, the first unqualified product similarity set 317, and the device density 318, and the output may be the first package quality result 323.

[0098] Among them, the first qualified quality similarity 314 may refer to the similarity between the image to be detected 310 and the qualified reference image 311; the first unqualified quality similarity 315 may refer to the similarity between the image to be detected 310 and the unqualified reference image 312; the first qualified product similarity set 316 may refer to a set composed of multiple first qualified quality similarities 314; the first unqualified product similarity set 317 may refer to a set composed of multiple first unqualified quality similarities 315. Among them, for the content of the device density 318, reference can be made to Figure 2 and its related description.

[0099] In some embodiments, as Figure 3 shown, the input of the quality judgment model 322 may further include process flow data 319, environmental data 320, and package material data 321. For the content of the environmental data 320, reference can be made to Figure 2 and its related description.

[0100] The process flow data 319 may refer to information related to the process flow of electronic packaging. For example, the process flow data 319 may include the sequence of the process flow, the specific operation information of each process flow, etc. In some embodiments, the processor may obtain the process flow data input by the user through the packaging control APP. Specifically, the process flow data of each type of packaging product may be stored in advance, and the user may determine the process flow data corresponding to each type of packaging product through the packaging control APP, and then select to input.

[0101] The encapsulation material data 321 may refer to the material information of the materials used in the electronic encapsulation. For example, the encapsulation material data 321 may include the material type, quality parameters, etc. Among them, the encapsulation material type may include metals, glass, ceramics, optoelectronic materials, epoxy resin materials, etc.; the quality parameters may include the service life of the encapsulation material, oxidation resistance, corrosion resistance, etc. In some embodiments, the processor may obtain the encapsulation material data 321 input by the user through the encapsulation control APP.

[0102] In some embodiments of this specification, when predicting the encapsulation quality result, adding the influence of the process flow data, encapsulation material data, and environmental data on the encapsulation quality can enable the model to pay attention to the influence of the process flow data, encapsulation material data, and environmental data on the encapsulation quality (especially internal defects of the encapsulated product) during prediction, making the result more in line with the actual situation and improving the accuracy of determining the encapsulation quality result.

[0103] In some embodiments, the similarity judgment model 313 and the quality judgment model 322 can be obtained through separate training; among them, the first training sample for training the similarity judgment model 313 includes multiple groups of sample encapsulated finished / semi-finished product images of different types. The sample encapsulated finished / semi-finished product images may include sample images to be detected, sample qualified reference images, and sample unqualified reference images; the first label is the similarity.

[0104] It can be understood that, in some embodiments, the first training sample may also be images of multiple historical encapsulated products; the first label is the similarity situation between any two of the images of multiple historical encapsulated products. Specifically, the first label can be obtained through manual annotation. For example, if any two images are images of qualified encapsulated products, they are labeled as similar; if one is an image of a qualified encapsulated product and one is an image of an unqualified encapsulated product, they are labeled as dissimilar. Since the similarity judgment model 313 outputs the similarity of two images (the image to be detected and the qualified reference image / unqualified reference image) each time it is applied, its essence is also to judge the similarity between two images. Therefore, here the images of historical encapsulated products can be directly used as training samples for training, making it easier to obtain training samples.

[0105] In some embodiments, the first label may include the similarity between the sample image to be detected and the sample qualified reference image, and the similarity between the sample image to be detected and the sample unqualified reference image in each group of the first training samples. In some embodiments, the processor may collect the first training samples through an image acquisition device. For example, the processor may use the image acquisition device to collect historical images of multiple historical packaged products as the first training samples. Among them, some historical images may be randomly selected as the sample images to be detected, and the qualified historical images among the remaining historical images may be determined as the sample qualified reference images, and the unqualified historical images may be determined as the unqualified reference images.

[0106] In some embodiments, the second training samples for training the quality judgment model 322 may be a set of sample qualified product similarities composed of the similarities between multiple groups of sample images to be detected and sample qualified reference images, a set of sample unqualified product similarities composed of the similarities between multiple groups of sample images to be detected and sample unqualified reference images, and the sample device density. In some embodiments, the second training samples may be determined by the trained similarity judgment model 313, or may be determined based on other image similarity algorithms. In some embodiments, the second training samples may further include sample process flow data, sample environment data, and sample packaging material data. Among them, the sample process flow data and the sample packaging material data may be obtained through user input of the packaging control APP; the sample environment data may be determined by the environmental data historically collected by the environmental sensor. For the content of the process flow data, environmental data, and packaging material data, reference can be made to Figure 2 、 Figure 3 and its related descriptions.

[0107] In some embodiments, the second label of the training quality judgment model 322 may be the packaging quality result of the sample packaged product corresponding to the sample image to be detected, which can be obtained through manual annotation.

[0108] In some embodiments, the processor may input the first training samples into the initial similarity judgment model to obtain the initial similarity between the sample image to be detected and the sample qualified reference image, and the initial similarity between the sample image to be detected and the sample unqualified reference image; construct a loss function based on the foregoing two initial similarities and the first training label, and use the loss function to update the parameters of the initial similarity judgment model; through parameter update, obtain the trained similarity judgment model 313.

[0109] In some embodiments, the processor may input the second training samples into the initial quality judgment model to obtain the initial packaging quality result; construct a loss function based on the initial packaging quality result and the second training label; use the loss function to update the parameters of the initial quality judgment model; through parameter update, obtain the trained quality judgment model 322.

[0110] In some embodiments of this specification, by training the similarity judgment model and the quality judgment model separately, it is possible to adapt to encapsulated finished or semi-finished products of different types and at different encapsulation stages. Because when training the models, images of encapsulated finished or semi-finished products of different types and at different encapsulation stages can be used as training data, and the obtained models can adapt to encapsulated products of various types and encapsulation stages. At the same time, by using the method of separate training, sufficient training data can be obtained, and training labels are also easier to obtain.

[0111] In some embodiments of this specification, by setting the judgment model as a combination of a similarity judgment model and a quality judgment model, different models can be used to process different data, improving the efficiency of data processing. At the same time, by using different models to process data specifically, deep-level data features can be extracted, improving the accuracy of model prediction.

[0112] In some embodiments of this specification, by using the judgment model to determine the encapsulation quality result, the self-learning ability of the machine learning model can be utilized to find patterns from a large amount of data, improving the accuracy of determining the encapsulation quality result. At the same time, in some embodiments of this specification, for each type of encapsulated product, a number of qualified reference images and unqualified reference images are taken as references. The model determines the encapsulation quality result based on the image to be detected, a number of qualified reference images, and unqualified reference images. The focus inside the model is to judge the consistency of different images of the same type of encapsulated product in terms of being qualified or not, and to synthesize multiple groups of consistencies to evaluate the encapsulation quality result, which can avoid the situation of training a model for each type of encapsulated product, improving the generality of the model. Therefore, in the embodiments of this specification, when training the model, pictures of different types of encapsulated products can be used for unified training to obtain a general judgment model.

[0113] Figure 4 It is an exemplary schematic diagram of the similarity judgment model shown according to some embodiments of the present invention.

[0114] In some embodiments, the similarity judgment model may include an image feature extraction layer and a similarity judgment layer. The image feature extraction layer is used to process the image to be detected, the qualified reference image, and the unqualified reference image to determine the image feature to be detected, the qualified reference image feature, and the unqualified reference image feature. The similarity judgment layer is used to process the image feature to be detected, the qualified reference image feature, and the unqualified reference image feature to determine the first quality similarity. The first quality similarity includes the first qualified quality similarity and the first unqualified quality similarity.

[0115] In some embodiments, such as Figure 4As shown, the similarity judgment model 313 may include an image feature extraction layer 324 and a similarity judgment layer 328. In some embodiments, the network structure of the image feature extraction layer 324 may be a Graph Neural Network (GNN), and the network structure of the similarity judgment layer 328 may be an NN.

[0116] As Figure 4 shown, the input of the image feature extraction layer 324 may be the image to be detected 310, multiple qualified reference images 311, and multiple unqualified reference images 312, and the output may be the image feature to be detected 325, multiple qualified reference image features 326, and multiple unqualified reference image features 327. The input of the similarity judgment layer 328 may be the image feature to be detected 325, multiple qualified reference image features 326, and multiple unqualified reference image features 327, and the output may be the first quality similarity 329. The first quality similarity 329 may include multiple first qualified quality similarities 314 and multiple first unqualified quality similarities 315.

[0117] In some embodiments, the image feature extraction layer 324 may process one image in sequence to obtain the corresponding image feature. For example, the image feature extraction layer 324 may first process the image to be detected 310 to obtain the image feature to be detected 325; then process each qualified reference image 311 respectively to obtain the corresponding qualified reference image feature 326; and finally process each unqualified reference image 312 respectively to obtain the corresponding unqualified reference image feature 327. Similarly, when the similarity judgment layer 328 processes the image features, it may process the image feature to be detected 325 and a qualified reference image feature 326 / unqualified reference image feature 327 respectively each time to obtain the corresponding quality similarity.

[0118] It should be noted that the order in which the image feature extraction layer 324 processes the images and the order in which the similarity judgment layer 328 processes the image features may be random, and the foregoing examples are not intended to limit the scope of this specification.

[0119] Among them, for the content of the image to be detected 310, the qualified reference images 311, and the unqualified reference images 312, reference may be made to Figure 2 and its related descriptions.

[0120] The image feature to be detected 325 may refer to data that can reflect the feature information of the image to be detected 310; the qualified reference image feature 326 may refer to data that can reflect the feature information of the qualified reference image 311; the unqualified reference image feature 327 may refer to data that can reflect the feature information of the unqualified reference image 312.

[0121] The first qualified quality similarity 314 may refer to the similarity between the image to be detected 310 and the qualified reference image 311; the first unqualified quality similarity 315 may refer to the similarity between the image to be detected 310 and the unqualified reference image 312.

[0122] In some embodiments, the similarity judgment model 313 may be obtained through the joint training of the image feature extraction layer 324 and the similarity judgment layer 328. In some embodiments, the training samples for training the similarity judgment model 313 may be the first training samples. For the specific content of the first training samples, reference may be made to Figure 3 and its related descriptions. In some embodiments, the labels for training the similarity judgment model 313 may be the first labels. For the content of the first labels, reference may be made to Figure 3 and its related descriptions.

[0123] In some embodiments, the processor may input the first training samples into the initial image feature extraction layer to obtain the initial image features to be detected, the initial qualified reference image features, and the initial unqualified reference image features; input the initial image features to be detected, the initial qualified reference image features, and the initial unqualified reference image features into the initial similarity judgment layer to obtain the initial first qualified quality similarity and the initial first unqualified quality similarity. A loss function is constructed based on the initial first qualified quality similarity, the initial first unqualified quality similarity, and the first labels; the parameters of the initial image feature extraction layer and the initial similarity judgment layer are updated synchronously using the loss function; through parameter update, the trained image feature extraction layer 324 and similarity judgment layer 328 are obtained.

[0124] In some embodiments of this specification, by setting the similarity judgment model as a network structure composed of an image feature extraction layer and a similarity judgment layer, the features of the image to be detected, the qualified reference image, and the unqualified reference image can be extracted first using the image feature extraction layer to obtain more accurate image features, and then the similarity can be determined based on the similarity judgment layer, which can improve the accuracy of model prediction.

[0125] In some embodiments, the similarity judgment model may include an image feature extraction layer and a similarity judgment layer; the image feature extraction layer is used to process the image to be detected, the qualified reference image, and the unqualified reference image to determine the image feature to be detected, the qualified reference image feature, and the unqualified reference image feature; the similarity judgment layer is used to process the image feature to be detected, the qualified reference image feature, the unqualified reference image feature, the process flow data, the environmental data, the encapsulation material data, and the device density to determine the second quality similarity; the quality judgment model is used to process the second qualified product similarity set and the second unqualified product similarity set to determine the second encapsulation quality result; wherein, the second quality similarity may include a second qualified quality similarity and a second unqualified quality similarity, and a plurality of second qualified quality similarities may determine the second qualified product similarity set, and a plurality of second unqualified quality similarities may determine the second unqualified product similarity set.

[0126] In some embodiments, as Figure 5 shown, the judgment model may be composed of a similarity judgment model 313 and a quality judgment model 322. Among them, the similarity judgment model 313 may be divided into an image feature extraction layer 324 and a similarity judgment layer 328. In some embodiments, the network structure of the image feature extraction 324 may be a GNN, the network structure of the similarity judgment layer 328 may be an NN, and the network structure of the quality judgment model 322 may be an NN.

[0127] As Figure 5 shown, the input of the image feature extraction layer 324 may be the image 310 to be detected, multiple qualified reference images 311, and multiple unqualified reference images 312, and the output is the image feature 325 to be detected, multiple qualified reference image features 326, and multiple unqualified reference image features 327. For the content of the image 310 to be detected, the qualified reference image 311, and the unqualified reference image 312, please refer to Figure 2 , Figure 3 and its related description; for the content of the image feature to be detected, the qualified reference image feature, and the unqualified reference image feature, please refer to Figure 4 and its related description.

[0128] In some embodiments, the image feature extraction layer 324 may process one image (image 310 to be detected / qualified reference image 311 / unqualified reference image 312) each time to obtain the corresponding image feature, and the processing order is not limited.

[0129] The input of the similarity judgment layer 328 can be the image features to be detected 325, multiple qualified reference image features 326, multiple unqualified reference image features 327, as well as the process flow data 319, environmental data 320, packaging material data 321, and device density 318 of the packaged product corresponding to the image features to be detected 325. The output can be the second quality similarity 330, where the second quality similarity 330 can include the second qualified quality similarity 331 and the second unqualified quality similarity 332. In some embodiments, the similarity judgment layer 328 can process the image features to be detected 325 and one qualified reference image feature 326 / unqualified reference image feature 327 each time, as well as a set of process flow data 319, environmental data 320, packaging material data 321, and device density 318 of the packaged product corresponding to the image features to be detected 325, to obtain the corresponding second quality similarity 330, and the processing order is not restricted.

[0130] For the content of the process flow data 319, environmental data 320, and packaging material data 321, reference can be made to Figure 3 and its related descriptions; for the content of the device density 318, reference can be made to Figure 2 and its related descriptions. The second quality similarity 330 can refer to the similarity between the image to be detected 310, multiple qualified reference images 311, and multiple unqualified reference images 312; the second qualified quality similarity 331 can refer to the similarity between the image to be detected 310 and the qualified reference image 311 when factors such as the process flow data 319, environmental data 320, packaging material data 321, and device density 318 are added to the input of the similarity judgment layer 328; the second unqualified quality similarity 332 can refer to the similarity between the image to be detected 310 and the unqualified reference image 312 when factors such as the process flow data 319, environmental data 320, packaging material data 321, and device density 318 are added to the input of the similarity judgment layer 328. By adding the process flow data 319, environmental data 320, packaging material data 321, and device density 318 to the input of the similarity judgment layer 328, the accuracy of the second qualified quality similarity 331 output by the similarity judgment layer 328 can be further improved on the basis of the first qualified quality similarity 314; the accuracy of the second unqualified quality similarity 332 can be further improved on the basis of the first unqualified quality similarity 315.

[0131] In some embodiments, multiple second qualified quality similarities 331 can form a second qualified product similarity set 333; multiple second unqualified quality similarities 332 can form a second unqualified product similarity set 334.

[0132] The input of the quality judgment model 322 can be the second set of qualified product similarities 333 and the second set of unqualified product similarities 334, and the output can be the second packaging quality result 335.

[0133] In some embodiments, the similarity judgment model 313 can be obtained through the joint training of the image feature extraction layer 324 and the similarity judgment layer 328. In some embodiments, the third training samples for training the similarity judgment model 313 can include the sample image to be detected, the sample qualified reference image, the sample unqualified reference image, and the sample process flow data, sample environment data, sample packaging material data, and sample device density corresponding to the sample packaged product corresponding to the sample image to be detected. In some embodiments, the sample image to be detected, the sample qualified reference image, and the sample unqualified reference image can be obtained by collecting historical packaged products through an image acquisition device. In some embodiments, the sample process flow data and the sample packaging material data can be determined by obtaining the user's input through the packaging control APP. In some embodiments, the sample environment data can be determined based on the historical environment data collected by the environmental sensor; the sample device density can be determined based on the wiring diagram corresponding to the packaged product input by the user in the packaging control APP.

[0134] In some embodiments, the third label for training the similarity judgment model 313 can be the similarities between multiple sets of images to be detected, sample qualified reference images, and sample unqualified reference images, including the sample qualified quality similarity and the sample unqualified quality similarity, which can be obtained through manual annotation.

[0135] In some embodiments, the processor can input the sample image to be detected, the sample qualified reference image, and the sample unqualified reference image into the initial image feature extraction layer to obtain the initial image feature to be detected, the initial qualified reference image feature, and the initial unqualified reference image feature; the processor can input the initial image feature to be detected, the initial qualified reference image feature, the initial unqualified reference image feature, the sample process flow data, the sample environment data, the sample packaging material data, and the sample device density into the initial similarity judgment layer to obtain the initial second quality similarity. The processor can construct a loss function based on the initial second quality similarity and the third training label, and use the loss function to synchronously update the parameters of the initial image feature extraction layer and the initial similarity judgment layer; through parameter update, the trained similarity judgment model 313 can be obtained.

[0136] In some embodiments, the fourth training samples for training the quality judgment model 322 can be the sample image to be detected, the sample qualified reference image, the sample second qualified product similarity set corresponding to the sample unqualified reference image, and the sample second unqualified product similarity set, which can be obtained through the trained similarity judgment model 313. In some embodiments, the third label of the quality judgment model 322 can be the actual packaging quality result corresponding to the third training sample, which can be obtained through manual annotation. For example, based on manual annotation of the sample image to be detected, the corresponding packaging quality result can be determined as the third label.

[0137] In some embodiments, the processor can input the fourth training sample into the initial quality judgment model to obtain an initial second packaging quality result; construct a loss function based on the initial second packaging quality result and the fourth label; use the loss function to update the parameters of the initial quality judgment model; and obtain the trained quality judgment model 322 through parameter update.

[0138] In some embodiments of this specification, by setting different layers for the judgment model and designing input data different from the foregoing embodiments for different layers, the input data of the model can be diverse and flexible; by adding process flow data, environmental data, packaging material data, and device density as the input of the similarity judgment layer, the accuracy of determining similarity can be improved.

[0139] One or more embodiments of this specification further provide an electronic packaging optimization device, including at least one memory and at least one processor; the at least one memory is used to store computer instructions, and the at least one processor is used to execute at least part of the computer instructions to implement the electronic packaging optimization method described in any one of the embodiments of this specification.

[0140] One or more embodiments of this specification further provide a computer-readable storage medium, and the storage medium stores computer instructions. When the computer reads the computer instructions in the storage medium, the computer runs the electronic packaging optimization method described in any one of the embodiments of this specification.

[0141] The beneficial effects that may be brought about by the embodiments of this specification include, but are not limited to: (1) By obtaining detection data and device density, determining the packaging quality result, and determining the packaging optimization method for the packaged product based on the packaging quality result, the efficiency and accuracy of packaging quality assessment in the electronic packaging process can be improved, so as to timely process and optimize the situation of unqualified packaging, and improve the automation level and efficiency of the entire electronic packaging production; (2) By using a judgment model to determine the packaging quality result, the self-learning ability of the machine learning model can be utilized to find patterns from a large amount of data, improving the accuracy of determining the packaging quality result; (3) For each type of packaged product, a number of qualified reference images and unqualified reference images are taken as references. The model determines the packaging quality result based on the image to be detected, a number of qualified reference images, and unqualified reference images. The focus inside the model is to judge the consistency of different images of the same type of packaged product in terms of being qualified or not, and to synthesize multiple groups of consistencies to evaluate the packaging quality result, which can avoid training a model for each type of packaged product, improving the versatility of the model; (4) By setting the judgment model as a combination of a similarity judgment model and a quality judgment model, different models can be used to process different data, improving the efficiency of data processing; at the same time, using different models to process data specifically can extract deep-level data features, improving the accuracy of model prediction.

[0142] The above specific embodiments only describe the design principles of the present invention. The shapes and names of the components in this description can be different and are not limited. Therefore, those skilled in the art of the present invention can modify or equivalently replace the technical solutions recorded in the foregoing embodiments; and these modifications and replacements do not depart from the purpose and technical solutions of the present invention, and shall all fall within the protection scope of the present invention.

Claims

1. An electronic packaging optimization method, characterized in that, Including: Obtain the wiring diagram of the packaged product through the packaging control application, and obtain the device density of the packaged product based on the wiring diagram; Input the detection data of the packaged product and the device density into the judgment model to obtain the packaging quality result of the packaged product, where the detection data includes the image to be detected; Perform packaging optimization on the packaged product according to the packaging quality result; The judgment model is a machine learning model, where the machine learning model at least includes: a similarity judgment model and a quality judgment model; Inputting the detection data and the device density into the judgment model to obtain the packaging quality result of the packaged product includes: Input the image to be detected, the qualified reference image of the packaged product, and the unqualified reference image of the packaged product into the similarity judgment model to obtain the qualified quality similarity and the unqualified quality similarity, and obtain the qualified product similarity set and the unqualified product similarity set according to the qualified quality similarity and the unqualified quality similarity; Input the qualified product similarity set and the unqualified product similarity set into the quality judgment model to obtain the packaging quality result.

2. The method according to claim 1, characterized in that, Wherein, The detection data further includes at least one of the following: The qualified reference image of the packaged product; the unqualified reference image of the packaged product; packaging process flow data; packaging environment data; packaging material data.

3. The method according to claim 1, characterized in that, Wherein, In the process of inputting the detection data and the device density into the judgment model to obtain the packaging quality result of the packaged product, it further includes: Input the packaging process flow data, packaging environment data, packaging material data, and the device density into the similarity judgment model and / or the quality judgment model to obtain the packaging quality result of the packaged product.

4. An electronic packaging optimization system based on the method according to claim 1, characterized in that, Including: A data acquisition module for obtaining the wiring diagram of the packaged product through the packaging control application and obtaining the device density of the packaged product based on the wiring diagram; A result determination module for inputting the detection data of the packaged product and the device density into the judgment model to obtain the packaging quality result of the packaged product, where the detection data includes the image to be detected; A packaging optimization module for performing packaging optimization on the packaged product according to the packaging quality result.

5. An electronic packaging optimization device, characterized in that, Including at least one memory and at least one processor; the at least one memory is used to store computer instructions, and the at least one processor is used to execute at least part of the computer instructions to implement the method according to any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The storage medium stores computer instructions, and when the computer reads the computer instructions in the storage medium, the computer runs the method according to any one of claims 1-3.

Images