Electronic component quality on-line monitoring and evaluation method, system and equipment

Abstract

The invention discloses an electronic component quality online monitoring and evaluation method, system and device. Establishing a historical monitoring image set, and obtaining specific values of resolution data, edge curve data, gray data and quality grade data; dividing the historical monitoring image set into a training set and a verification set, inputting the training set and the verification set into an initial model for training, optimizing parameters in the initial model, and constructing a target evaluation model; obtaining a to-be-evaluated monitoring image, and marking the to-be-evaluated monitoring image with a high-resolution label, a medium-resolution label and a low-resolution label according to the resolution data of the to-be-evaluated monitoring image; calculating quality grade data of the to-be-evaluated monitoring image marked with the high-resolution label; calculating quality grade data of the to-be-evaluated monitoring image marked with the resolution label in the mark; and calculating quality grade data of the to-be-evaluated monitoring image of the low-resolution label.

Description

Technical Field

[0001] This application relates to the field of electronic component monitoring technology, specifically to a method, system, and equipment for online monitoring and evaluation of electronic component quality. Background Technology

[0002] In the current standard process for inspecting the surface quality of electronic components, two image processing methods are typically used: one is to directly analyze and judge the acquired raw image, and the other is to first apply simple noise reduction preprocessing to the raw image and then further extract feature information. However, regardless of which method is used, both require high image resolution, meaning that high-resolution images of electronic components must be used as the input source.

[0003] This high dependence on image quality leads to a significant decrease in the recognition and detection capabilities of existing methods when faced with samples of insufficient resolution or low image quality. This, in turn, limits the adaptability of the detection system under different imaging conditions and equipment configurations. As a result, the applicable scenarios of existing detection schemes are relatively limited, making it difficult to cover the low-resolution image situations commonly encountered in diverse and complex industrial environments. Ultimately, this weakens the universality and potential for widespread application of electronic component surface quality inspection technology.

[0004] It should be noted that the information disclosed in this background section is only for understanding the background technology of the present invention, and therefore may include information that does not constitute prior art. Application content

[0005] To address the shortcomings of existing technologies, this application discloses an online monitoring and evaluation method, system, and device for electronic component quality, which can solve the problem that current electronic component surface quality inspection relies on high-resolution electronic component images.

[0006] To achieve the above objectives, this application provides the following technical solution: A method for online monitoring and evaluation of electronic component quality, comprising the following steps: Establish a historical monitoring image set, where each historical monitoring image has resolution data, edge curve data, grayscale data, and quality level data, and obtain the specific values ​​of resolution data, edge curve data, grayscale data, and quality level data; The historical monitoring image set is divided into a training set and a validation set. The training set and the validation set are input into an initial model for training. The parameters in the initial model are optimized, and the similarity of each group of data in the training set and the validation set is calculated. Subjective evaluation scores are calculated based on the similarity of each group of data in the training set and the validation set to construct a target evaluation model, which is used to evaluate historical monitoring images of different resolutions. Acquire the monitoring images to be evaluated, and label them with high-resolution, medium-resolution, and low-resolution tags based on their resolution data; Calculate the similarity between the edge curve data of the monitoring image to be evaluated labeled with high-resolution tags and the edge curve data of the historical monitoring image set. If the similarity meets the specified subjective evaluation score, the data are labeled with the same quality level. Calculate the similarity between the grayscale data of the monitoring image to be evaluated and the grayscale data of the historical monitoring image set, which are labeled with medium resolution labels. If the similarity meets the specified subjective evaluation score, the data are labeled with the same quality level. A coordinate system is established for the low-resolution labeled monitoring images to be evaluated according to the specified standards. Multiple monitoring images to be evaluated are obtained by segmentation based on the coordinate system. The similarity between the multiple monitoring images to be evaluated and all data of multiple historical monitoring image sets is calculated. If the similarity meets the specified subjective evaluation score, the data are labeled with the same quality level.

[0007] In a preferred embodiment, when the resolution data of the monitoring image to be evaluated is higher than a first threshold, the resolution data of the monitoring image to be evaluated is labeled as a high-resolution label; when the resolution data of the monitoring image to be evaluated is lower than the first threshold but higher than a second threshold, the resolution data of the monitoring image to be evaluated is labeled as a medium-resolution label; and when the resolution data of the monitoring image to be evaluated is lower than the second threshold, the resolution data of the monitoring image to be evaluated is labeled as a low-resolution label.

[0008] The preferred technical solution involves filtering and denoising the image to be evaluated, calculating the coordinates of each pixel in the image based on the coordinate system, determining and connecting edge points for each pixel, and generating an edge curve.

[0009] In a preferred technical solution, the similarity between the edge curve data of the high-resolution labeled monitoring image to be evaluated and the edge curve data of the historical monitoring image set is set as the overlap degree between the edge curves of the high-resolution labeled monitoring image to be evaluated and the edge curves of the historical monitoring image set. When the overlap degree is greater than 90%, the data are labeled with the same quality level.

[0010] In a preferred technical solution, the similarity between the edge curve data of the low-resolution labeled monitoring image to be evaluated and the edge curve data of the historical monitoring image set is set as the overlap degree between the edge curves of the low-resolution labeled monitoring image to be evaluated and the edge curves of the historical monitoring image set. When the overlap degree is greater than 70%, the data are labeled with the same quality level.

[0011] In a preferred technical solution, multiple monitoring images to be evaluated are segmented and labeled as Xn, and multiple historical monitoring images are segmented and labeled as Yn. The center coordinates of the nth monitoring image to be evaluated Xn and the nth historical monitoring images Yn are the same.

[0012] In addition, this application also discloses an online monitoring and evaluation system for the quality of electronic components, including: The image monitoring unit is used to acquire the monitoring image to be evaluated and the historical monitoring images used to establish a historical monitoring image set, and to acquire specific values ​​of resolution data, edge curve data, grayscale data and quality level data for each of the historical monitoring images; The image preprocessing unit is used to label the monitoring image to be evaluated with high-resolution, medium-resolution, and low-resolution labels based on the resolution data of the monitoring image to be evaluated; The image evaluation unit is used to optimize the parameters in the initial model, construct the target evaluation model, evaluate historical monitoring images of different resolutions based on the target evaluation model, and obtain quality level data.

[0013] In addition, this application also discloses a computer device, including a processor and a memory, wherein the memory is used to store a computer program, and the computer program, when executed by the processor, implements the online monitoring and evaluation method for the quality of electronic components as described in any of the above technical solutions.

[0014] This application discloses a method, system, and device for online monitoring and evaluation of electronic component quality, which has the following advantages: A systematic approach was adopted to collect historical monitoring images of electronic component surfaces from different production batches and under varying imaging conditions, forming an initial image set of sufficient size and diversity. For each image in this set, multimodal data annotation was performed to construct a structured dataset. This addresses the current problem of relying on high-resolution images for electronic component surface quality inspection. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0016] Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without any creative effort.

[0017] Figure 1 This is a flowchart illustrating Embodiment 1 of this application; Figure 2 This is a system schematic diagram of Embodiment 2 of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.

[0019] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Example 1 like Figure 1 As shown, the online monitoring and evaluation method for electronic component quality described in this application includes the following steps: A systematic collection of historical monitoring images of electronic component surfaces from different production batches and imaging conditions was developed to form an initial image set of sufficient scale and diversity. For each image in this set, multimodal data annotation was performed to construct a structured dataset. The data annotation specifically included: resolution data, edge curve data, grayscale data, and quality grade data, obtaining specific values ​​for these data. Among these, the quality grade data was assigned a specific quality grade label to each image based on established standards by experienced quality inspectors, including acceptable, minor defects, and severe defects.

[0021] The historical monitoring image set is randomly divided into three subsets according to a preset ratio: a training set, a validation set, and a test set. The training set is used for model learning, the validation set is used to evaluate model performance and guide parameter adjustment during training, and the test set is used to finally evaluate the model's generalization ability. The training set and the validation set are input into an initial model for training, the parameters in the initial model are optimized, and the similarity of each data set in the training set and the validation set is calculated. A subjective evaluation score is calculated based on the similarity of each data set in the training set and the validation set; the higher the score, the more accurate the constructed target evaluation model. The target evaluation model is used to evaluate historical monitoring images at different resolutions.

[0022] The system acquires the monitoring images to be evaluated and labels them as high-resolution, medium-resolution, and low-resolution images based on their resolution data. The system presets two resolution thresholds: a first threshold and a second threshold, with the first threshold being greater than the second threshold. The specific values ​​of these two thresholds can be preset based on statistical analysis of a large amount of historical image data, the image accuracy requirements of the specific detection task, or by verifying the best results through experiments. The thresholds can also be dynamically adjustable to adapt to different production lines and detection standards.

[0023] When the resolution data of the monitoring image to be evaluated is higher than a first threshold, the resolution data of the monitoring image to be evaluated is labeled as high resolution. Such images typically contain rich and clear details, with sharp edge contours, making them suitable for precise microscopic defect detection. When the resolution data of the monitoring image to be evaluated is lower than the first threshold but higher than a second threshold, the resolution data of the monitoring image to be evaluated is labeled as medium resolution. The basic structure and contours of such images are still recognizable, but some detail information is lost, requiring analysis using features that are insensitive to resolution degradation. When the resolution data of the monitoring image to be evaluated is lower than the second threshold, the resolution data of the monitoring image to be evaluated is labeled as low resolution. Such images are typically blurry, with severely lost details, making them difficult to process using traditional methods, requiring the implementation of a specially designed robust evaluation strategy.

[0024] To ensure the accuracy of edge detection, the high-resolution monitoring images to be evaluated are first preprocessed. The images are filtered and denoised to suppress noise interference that may be introduced during image acquisition, while preserving as much realistic edge information as possible. A two-dimensional pixel coordinate system is established with the top-left corner of the image as the origin. Each pixel in the image is traversed, and its gradient magnitude and direction in the x and y directions are calculated to initially identify pixel-level high-gradient regions. Then, the similarity between the edge curve data of the monitoring image to be evaluated labeled with high-resolution tags and the edge curve data of the historical monitoring image set is calculated. If the similarity meets the specified subjective evaluation score, the data is labeled with the same quality level.

[0025] The similarity between the edge curve data of the high-resolution labeled monitoring image to be evaluated and the edge curve data of the historical monitoring image set is set as the overlap degree between the edge curves of the high-resolution labeled monitoring image to be evaluated and the edge curves of the historical monitoring image set. The edge curve of the image to be evaluated is superimposed and aligned with the edge curve of a standard quality sample in the image space, and the percentage of overlapping pixels is calculated out of the total number of edge pixels. When the calculated overlap degree is greater than a preset 90%, the system determines that the two have extremely high similarity. At this time, the similarity meets the standard corresponding to the specified subjective evaluation score. If the historical monitoring image is qualified, the monitoring image to be evaluated is also rated as qualified.

[0026] For monitoring images labeled as medium resolution, their detail clarity is lower than that of high resolution images, and direct, precise edge curve comparison may introduce significant errors due to feature blurring. Therefore, instead of relying on fine edges, we use grayscale data, which is less sensitive to resolution changes and better reflects the macroscopic statistical characteristics of images, for similarity analysis and quality evaluation.

[0027] After acquiring the monitoring image to be evaluated, labeled with a medium-resolution tag, the system first extracts its grayscale data. This grayscale data is not the grayscale value of a single pixel, but rather a statistical feature that characterizes the overall brightness and darkness distribution of the image or key areas. The system then performs a similarity calculation between the extracted grayscale data of the monitoring image to be evaluated and the grayscale data of images with known quality levels in the historical monitoring image set. When the similarity between the grayscale data of the monitoring image to be evaluated and the grayscale data of the historical monitoring images reaches or exceeds a preset threshold of 80%, it is determined that the image meets the specified subjective evaluation score requirements.

[0028] For monitoring images labeled as low-resolution, the information they carry is severely insufficient, and their features are highly blurred. Using edge comparison methods for high-resolution images or grayscale statistical methods for medium-resolution images makes it difficult to obtain reliable evaluation results.

[0029] To perform structured analysis on the low-resolution image to be evaluated, a standardized two-dimensional pixel coordinate system is established with the top-left corner of the low-resolution monitoring image as the origin, the width direction as the X-axis, and the height direction as the Y-axis. The image is uniformly divided into M×N grids, resulting in multiple sub-images. For ease of processing, these sub-images are labeled X1, X2, ..., Xn. The system records the center coordinates (Cxn, Cyn) of each sub-image Xn in the original low-resolution image coordinate system. To ensure the effectiveness and fairness of the comparison, the system processes all samples in the historical monitoring image set identically. A sample image from the historical monitoring image set is selected and segmented into multiple sub-images using the same grid division rules and coordinate definitions. These sub-images from the historical samples are labeled Y1, Y2, ..., Yn. Crucially, through design, the nth sub-image to be evaluated, Xn, and the nth historical sub-image, Yn, have exactly the same center coordinates in their respective image coordinate systems. This means that Xn and Yn originate from the exact same relative positions in two different images, thus ensuring spatial correspondence in subsequent comparisons.

[0030] In this step, the system initiates a comprehensive similarity calculation strategy. Instead of relying solely on edge or grayscale data, it calculates the comprehensive similarity of all data between the sub-image Xn to be evaluated and its corresponding historical sub-image Yn, including but not limited to edge curve data and grayscale data. Specifically, regarding the similarity calculation of edge curve data, the judgment criteria are appropriately relaxed due to the ambiguity of edges in low-resolution images. The similarity of the edge curve data is quantified as the degree of overlap between the corresponding edge curves of Xn and Yn. Since it is a local comparison and the image quality is low, when the calculated overlap is greater than a preset 70%, it is considered that the edge features of the two images have reached an acceptable similarity standard in that local area. The system integrates the similarity results of all sub-image pairs (Xn, Yn) across all data dimensions, and obtains a global, fused comprehensive similarity score through a weighted average or voting mechanism.

[0031] When the calculated overall similarity score meets the specified subjective evaluation score set for the low-resolution image, the system determines that the low-resolution monitoring image to be evaluated is highly similar to the historical sample image in overall quality. Based on this determination, the system will assign the low-resolution monitoring image to be evaluated the same quality level data as the historical sample image, thereby achieving a reliable evaluation of the low-quality image.

[0032] Example 2 like Figure 2 As shown, the online monitoring and evaluation system for electronic component quality described in this application includes: an image monitoring unit, used to acquire the monitoring image to be evaluated and historical monitoring images for establishing a historical monitoring image set, and to acquire specific values ​​of resolution data, edge curve data, grayscale data, and quality grade data for each historical monitoring image; an image preprocessing unit, used to label the monitoring image to be evaluated with high-resolution, medium-resolution, and low-resolution labels according to the resolution data of the monitoring image to be evaluated; and an image evaluation unit, used to optimize the parameters in the initial model, construct a target evaluation model, and evaluate historical monitoring images with different resolution data according to the target evaluation model to obtain quality grade data. The online monitoring and evaluation system for electronic component quality described in this application is used to implement the online monitoring and evaluation method for electronic component quality described in Embodiment 1.

[0033] Example 3 The computer device described in this application includes a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements the online monitoring and evaluation method for the quality of electronic components as described in Embodiment 1.

[0034] It should be noted that, in this article, relational terms are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0035] Moreover, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0036] In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0037] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices, which can be centralized on a single computing device or distributed on a network of multiple computing devices.

[0038] Alternatively, they can be implemented using computer-executable program code, so that they can be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module.

[0039] This invention is not limited to any particular combination of hardware and software.

[0040] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for on-line monitoring and evaluation of the quality of electronic components, wherein, The method comprises the following steps: establishing a historical monitoring image set, each of the historical monitoring images having resolution data, edge curve data, grayscale data and quality level data, and obtaining specific values of the resolution data, the edge curve data, the grayscale data and the quality level data; dividing the historical monitoring image set into a training set and a validation set, inputting the training set and the validation set into an initial model for training, optimizing parameters in the initial model, and calculating the similarity of each group of data in the training set and the validation set; performing subjective evaluation scoring according to the similarity of each group of data in the training set and the validation set, and constructing a target evaluation model, which is used for evaluating the historical monitoring images with different resolution data; obtaining a to-be-evaluated monitoring image, and marking the to-be-evaluated monitoring image with a high-resolution label, a medium-resolution label and a low-resolution label according to the resolution data of the to-be-evaluated monitoring image; calculating the similarity of the edge curve data of the to-be-evaluated monitoring image marked with the high-resolution label and the edge curve data of the historical monitoring image set, and marking the same quality level data when the similarity meets a specified subjective evaluation score; calculating the similarity of the grayscale data of the to-be-evaluated monitoring image marked with the medium-resolution label and the grayscale data of the historical monitoring image set, and marking the same quality level data when the similarity meets a specified subjective evaluation score; establishing a coordinate system for the to-be-evaluated monitoring image marked with the low-resolution label according to a specified standard, segmenting the to-be-evaluated monitoring image according to the coordinate system to obtain a plurality of to-be-evaluated monitoring images, calculating the similarity of the plurality of to-be-evaluated monitoring images and all data of the plurality of historical monitoring image sets, and marking the same quality level data when the similarity meets a specified subjective evaluation score.

2. The method of claim 1, wherein, When the resolution data of the to-be-evaluated monitoring image is higher than a first threshold value, the resolution data of the to-be-evaluated monitoring image is marked as the high-resolution label; when the resolution data of the to-be-evaluated monitoring image is lower than the first threshold value and higher than a second threshold value, the resolution data of the to-be-evaluated monitoring image is marked as the medium-resolution label; and when the resolution data of the to-be-evaluated monitoring image is lower than the second threshold value, the resolution data of the to-be-evaluated monitoring image is marked as the low-resolution label.

3. The method of claim 1, wherein, The to-be-evaluated monitoring image is filtered and denoised, the coordinates of each pixel point in the image are calculated based on the coordinate system, edge point determination and connection are performed on each pixel point, and an edge curve is generated.

4. The method of claim 1, wherein, The similarity of the edge curve data of the to-be-evaluated monitoring image marked with the high-resolution label and the edge curve data of the historical monitoring image set is set as the coincidence degree of the edge curve of the to-be-evaluated monitoring image marked with the high-resolution label and the edge curve of the historical monitoring image set, and the same quality level data is marked when the coincidence degree is greater than 90%.

5. The method for online monitoring and evaluation of the quality of electronic components according to claim 1, wherein, The similarity of the edge curve data of the to-be-evaluated monitoring image marked with the low-resolution label and the edge curve data of the historical monitoring image set is set as the coincidence degree of the edge curve of the to-be-evaluated monitoring image marked with the low-resolution label and the edge curve of the historical monitoring image set, and the same quality level data is marked when the coincidence degree is greater than 70%.

6. The method for online monitoring and evaluation of the quality of electronic components according to claim 1, wherein, The segmented plurality of to-be-evaluated monitoring images are marked as Xn, the segmented plurality of historical monitoring images are marked as Yn, and the center coordinates of the nth to-be-evaluated monitoring image Xn and the n historical monitoring images Yn are the same.

7. The system for on-line monitoring and evaluating the quality of electronic components according to any one of claims 1 to 6, wherein, Comprise: An image monitoring unit is configured to acquire to-be-evaluated monitoring images and historical monitoring images for establishing a set of historical monitoring images, and acquire specific values of resolution data, edge curve data, grayscale data and quality level data of each historical monitoring image; An image preprocessing unit is configured to mark to-be-evaluated monitoring images with high-resolution labels, medium-resolution labels and low-resolution labels according to resolution data of the to-be-evaluated monitoring images; An image evaluation unit is configured to optimize parameters in the initial model, construct a target evaluation model, and evaluate historical monitoring images with different resolution data according to the target evaluation model to obtain quality level data.

8. Computer device, wherein, Comprise a processor and a memory, the memory is used to store a computer program, the computer program is executed by the processor to realize the electronic element quality online monitoring and evaluation method as claimed in any one of claims 1 to 6.

Images