Decorative paper surface defect online detection method based on machine vision

Abstract

The invention discloses a machine vision-based decorative paper surface defect online detection method, and relates to the technical field of computer processing, and the method comprises the following steps: S01, carrying out image acquisition on moving decorative paper through a visual camera, and obtaining a surface image of the decorative paper; s02, extracting edge contour features of the surface image, and determining a first defect feature set of the decorative paper according to the edge contour features; s03, performing classification identification on the surface image based on the extracted surface image to obtain the category of the surface image, and identifying a second defect feature set of the decorative paper based on the category of the surface image; and S04, acquiring a third defect feature set of the decorative paper through the defect template set updated in real time according to the first defect feature set. According to the method, the defects of the decorative paper are obtained through multi-level feature extraction, the limitation of a single detection method can be effectively overcome, and the omission ratio and the misjudgment rate are greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer processing, in particular to a surface defect online detection method for decorative paper based on machine vision. BACKGROUND

[0002] In the field of decorative paper production and manufacturing, surface quality is a key factor determining product grade and value. Traditional defect detection mainly relies on manual visual inspection, which is not only inefficient and labor-intensive, but also affected by personnel subjective factors and fatigue, leading to missed detection and misjudgment, and unable to meet the online quality control needs of modern high-speed production lines.

[0003] With the development of machine vision technology, automated surface defect detection has become an industry trend. In the prior art, a common solution is to use a single visual processing algorithm. Referring to the Chinese authorized patent, publication number CN102984434B, publication date 2015-03-04, an image processing device and an image processing method are disclosed. The correction of the front-end blur or the correction of the scanning cage corresponding to the thinning is performed. The image processing device includes: a contour detection unit that analyzes image data, detects pixels of an outer edge of a contour, and determines the edge direction of the edge; a thinning processing unit that performs thinning processing on the detected pixels of the outer edge and reduces the pixel value of the pixels; a front-end blur correction value for the pixels of the outer edge is determined according to the paper conveying direction during printing processing of the image data and the edge direction of the pixels of the outer edge, and the correction value of the front-end blur is reduced according to the proportion of the pixel value of the pixels of the outer edge reduced by the thinning processing, and the reduced correction value is added to the pixel value of the pixels of the outer edge to correct the front-end blur.

[0004] The deficiencies of the prior art are as follows: first, the texture of decorative paper is diverse, such as wood grain, solid color, stone texture, etc., and different types of texture have different defect forms, so a single, universal detection model cannot achieve ideal accuracy on all categories. For specific defects such as knot scars on wood grain, color difference on solid color surface, and breakage in complex texture, the adaptability and accuracy are insufficient. Second, the conditions on the production line are complex, and image acquisition is easily disturbed by light, speed, etc. A single detection standard has poor robustness when facing these fluctuations, and is prone to false positives. In addition, the pattern of defects is not fixed, and new defect types may appear at any time, while the traditional static template library cannot be updated in real time, resulting in ineffective identification of new defect patterns. SUMMARY

[0005] To solve the above technical problems, the technical solution adopted by the present application is a surface defect online detection method for decorative paper based on machine vision, which includes the following steps: S01. The moving decorative paper is captured by a vision camera to obtain an image of the surface of the decorative paper; S02. Extract the edge contour features of the surface image, and determine the first defect feature set of the decorative paper based on the edge contour features; S03. Based on the extracted surface image, classify and identify the category of the surface image, and identify the second defect feature set of the decorative paper based on the category of the surface image; S04. Obtain the third defect feature set of the decorative paper through the real-time updated defect template set. Based on the acceptance strategies corresponding to the first defect feature set and the second defect feature set, determine the defect feature set that conforms to the acceptance strategy in the first defect feature set and the second defect feature set as the fourth defect feature set. S05. Identify the first defect feature set of multiple paper samples through the first feature extraction model, and use it as the first test result; S06. Identify the second defect feature set of the multiple paper samples using the second feature extraction model, and use it as the second test result; S07. Calculate the first confidence level of the first feature extraction model for each defect feature set contained in the first test result, and calculate the second confidence level of the second feature extraction model for each defect feature set contained in the second test result. S08. For the target defect feature sets that overlap in the first test results and the second test results, based on the first confidence level and the second confidence level of each target defect feature set, determine the confidence model corresponding to each target defect feature set from the first feature extraction model and the second feature extraction model, and use the correspondence between the target defect feature set and the confidence model as the acceptance strategy corresponding to the target defect feature set. S09. Determine the defect detection result of the decorative paper based on the fourth defect feature set and the third defect feature set.

[0006] Preferably, step S02, which involves extracting the edge contour features of the surface image and determining the first defect feature set of the decorative paper based on the edge contour features, includes: Identify the edge contours of the surface image; The edge contour is enlarged by a predetermined factor to obtain an enlarged edge contour; Calculate the matching degree between the pixel points of the enlarged edge contour and the standard pixel points. If the matching degree is lower than a preset threshold, then determine that the first defect feature set of the decorative paper has a contour defect.

[0007] Preferably, in step S03, the surface image is classified and identified to obtain the category of the surface image, and the second defect feature set of the decorative paper is identified based on the category of the surface image. When a specific category of the surface image is determined through the first defect feature set, features that identify the specific category in the surface image are extracted; The second defect feature of the decorative paper is determined based on the characteristics of the specific category.

[0008] Preferably, determining the second defect feature of the decorative paper based on the features of the specific category includes: Identify specific categories of features in the surface image to determine defect labels for specific categories of features in the decorative paper; Obtain the corresponding specific type of defect model related to the defect label of the specific category of features through the knowledge graph model; Calculate the similarity between the surface image and the defect image corresponding to the specific type of defect model to determine the defects of a specific category of the decorative paper; The second defect feature is determined based on the defects of the specific category of features and the defect labels of the specific category of features.

[0009] Preferably, step S03, which involves identifying the second set of defect features of the decorative paper based on the category of the surface image, includes: When the surface image is classified as a solid color, it is matched with a color uniformity template, and the second defect feature set of the decorative paper is determined based on the matched color uniformity template.

[0010] Preferably, the color uniformity template includes a color distribution histogram, and the matching of the color uniformity template with the surface image and the determination of the second defect feature set of the decorative paper based on the matching result includes: The surface image is matched with multiple color distribution histogram templates to determine a target color distribution histogram that matches the surface image; Obtain the target color uniformity index associated with the target color distribution histogram; The second defect feature set is determined based on the target color uniformity index.

[0011] Preferably, the step S04 of obtaining the third defect feature set of the decorative paper through a real-time updated defect template set includes: Obtain a real-time updated set of defect templates, and obtain the defect image data corresponding to the set of defect templates; Calculate the similarity between the surface image and the defect image data, and filter out multiple target defect images whose similarity meets a preset threshold from the defect image data; The third defect feature set of the decorative paper is determined by using the defect labels corresponding to the multiple target defect images.

[0012] Preferably, step S04, which involves determining the defect feature sets in the first and second defect feature sets that conform to the acceptance strategies corresponding to the first and second defect feature sets, as the fourth defect feature set, includes: For each first defect feature set, if the model outputting the first defect feature set is the confidence model corresponding to the first defect feature set, then the first defect feature set is determined as a fourth defect feature set; wherein, the confidence level of the first defect feature set output by the confidence model corresponding to the first defect feature set conforms to the acceptance strategy corresponding to the first defect feature set. For each second defect feature set, if the model outputting the second defect feature set is the confidence model corresponding to the second defect feature set, then the second defect feature set is determined as a fourth defect feature set; wherein, the confidence level of the second defect feature set output by the confidence model corresponding to the second defect feature set conforms to the acceptance strategy corresponding to the second defect feature set.

[0013] The present invention has at least the following beneficial effects: By constructing a multi-level defect feature set, multi-dimensional capture of surface defects in decorative paper was achieved. The first defect feature set, based on edge contour analysis, effectively identifies contour shape defects. The second defect feature set, based on an image classification gating mechanism, uses a dedicated model for in-depth analysis of different texture categories, significantly improving the recognition accuracy of specific types of defects. The third defect feature set, through a real-time updated defect template library, ensures the timely identification of new defects.

[0014] By constructing a multi-level defect feature set, multi-dimensional capture of surface defects in decorative paper was achieved. The first defect feature set, based on edge contour analysis, effectively identifies contour shape defects. The second defect feature set, based on an image classification gating mechanism, uses a dedicated model for in-depth analysis of different texture categories, significantly improving the recognition accuracy of specific types of defects. The third defect feature set, through a real-time updated defect template library, ensures the timely identification of new defects. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of an online detection method for surface defects of decorative paper based on machine vision, provided in Embodiment 1 of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "including," "having," and any variations are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0019] Example 1

[0020] This embodiment provides an online detection method for surface defects in decorative paper based on machine vision. The method includes the following steps: Figure 1 As shown: S01. The moving decorative paper is captured by a vision camera to obtain an image of the surface of the decorative paper; Specifically, visual cameras are mounted on a conveyor belt, and the finished stamped or processed decorative paper is conveyed through the conveyor belt and then continuously photographed by the visual cameras at 1-second or 0.5-second intervals. At least two visual cameras are installed, and they are placed at the same angle and height.

[0021] S02. Extract the edge contour features of the surface image and determine the first defect feature set of the decorative paper based on the edge contour features; Further, in step S02, edge contour features of the surface image are extracted, and the first defect feature set of the decorative paper is determined based on the edge contour features, including: Identify the edge contours of a surface image; The edge contour is enlarged by a predetermined factor to obtain the enlarged edge contour; Calculate the matching degree between the pixel points of the enlarged edge contour and the standard pixel points. If the matching degree is lower than the preset threshold, the first defect feature set of the decorative paper is determined to have contour defects.

[0022] Specifically, the system first uses an edge detection algorithm (such as the Canny operator) to accurately identify the edge contours of the surface image. Then, the identified contour image is magnified by a predetermined factor (usually 5-10 times). This magnification significantly enhances the features of defects such as tiny burrs and jagged unevenness that are difficult to detect at the original resolution. Next, the system performs point-by-point matching calculations between the magnified contour pixels and pre-stored standard qualified product pixels. By setting a quantized matching threshold (such as 98% or 95%), when the actual matching degree is lower than this threshold, the system automatically determines that the decorative paper has contour defects and includes its features in the first defect feature set. The purpose is that by magnifying the contour and then performing pixel-level comparison, the system can effectively capture micron-level contour anomalies that are difficult for the human eye to detect and easily overlooked by conventional algorithms.

[0023] S03. Based on the extracted surface image, classify and identify the category of the surface image, and identify the second defect feature set of the decorative paper based on the category of the surface image; Furthermore, the above-mentioned classification and recognition of surface images obtains the category of the surface images, and the second defect feature set of the decorative paper is identified based on the category of the surface images: When a specific category of a surface image is determined using a first defect feature set, features that identify that specific category in the surface image are extracted. The second defect characteristics of decorative paper are determined based on specific categories of features, including: Identify specific categories of features in surface images to determine defect labels for specific categories of features in decorative paper; Obtain a corresponding defect model of a specific type that is related to the defect label of a specific category of features through a knowledge graph model; Calculate the similarity between the surface image and the defect image corresponding to the specific type of defect model to determine the defects of a specific category in the decorative paper; The second defect feature is determined based on the defect of a specific category and the defect label of that specific category.

[0024] Specifically, the system first determines the material category (such as wood grain, solid color, or stone grain) through preliminary analysis of the surface image, and then initiates a deep feature extraction process for that specific category. When identified as wood grain, the system uses a pre-trained specialized model to extract features such as wood grain knots and texture breaks, and generates preliminary defect labels. Next, through a pre-built knowledge graph model, this type of defect label is associated and mapped with typical wood grain defect patterns (such as misalignments, dead knots, and texture irregularities) stored in the knowledge base to obtain the corresponding specialized defect recognition model. Finally, by calculating the similarity matrix between the image to be detected and the standard defect samples in the knowledge graph, the system completes the accurate identification and verification of defects of a specific category, and finally outputs a structured second defect feature set.

[0025] S04. Obtain the third defect feature set of decorative paper through the real-time updated defect template set. Based on the acceptance strategies corresponding to the first and second defect feature sets, determine the defect feature sets in the first and second defect feature sets that conform to the acceptance strategies as the fourth defect feature set, including: For each first defect feature set, if the model that outputs the first defect feature set is the confidence model corresponding to the first defect feature set, then the first defect feature set is determined as a fourth defect feature set; wherein, the confidence level of the first defect feature set output by the confidence model corresponding to the first defect feature set conforms to the acceptance strategy corresponding to the first defect feature set. For each second defect feature set, if the model that outputs the second defect feature set is the confidence model corresponding to the second defect feature set, then the second defect feature set is determined as a fourth defect feature set; wherein, the confidence level of the second defect feature set output by the confidence model corresponding to the second defect feature set conforms to the acceptance strategy corresponding to the second defect feature set.

[0026] Furthermore, the third defect feature set of decorative paper obtained through the real-time updated defect template set includes: Obtain the real-time updated set of defect templates and the corresponding defect image data; Calculate the similarity between surface image and defect image data, and select multiple target defect images whose similarity meets a preset threshold from the defect image data; The third defect feature set of decorative paper is determined by using the defect labels corresponding to multiple target defect images.

[0027] Specifically, during runtime, the system performs high-speed similarity calculations between the currently acquired surface image and the massive defect image data in the dynamic template library (for example, using perceptual hashing algorithms or deep learning feature similarity comparison), quickly filters out multiple target defect images with similarity exceeding a preset threshold, and then summarizes the defect labels corresponding to these target images to generate a third defect feature set that reflects the latest defect trends.

[0028] Furthermore, the system pre-evaluates the first feature extraction model (general contour model) and the second feature extraction model (specific category model) offline, calculating their confidence levels for identifying various specific defects. Based on this, an acceptance strategy is established for each defect—that is, it explicitly stipulates which model (confidence model) is most reliable for judging this type of defect. In online detection, after the first and second feature sets are generated, the system does not simply merge all results, but acts as an arbitrator: it examines each reported defect, and if the model outputting the defect happens to be its pre-defined corresponding confidence model, then the defect is accepted and included in the final fourth defect feature set; otherwise, the result is considered unreliable and discarded.

[0029] S05. Identify the first defect feature set of multiple paper samples using the first feature extraction model, and use it as the first test result; S06. Identify the second defect feature set of multiple paper samples using the second feature extraction model, and use it as the second test result; S07. Calculate the first confidence level of the first feature extraction model for each defect feature set contained in the first test result, and calculate the second confidence level of the second feature extraction model for each defect feature set contained in the second test result. S08. For the target defect feature sets that overlap in the first test results and the second test results, based on the first confidence level and the second confidence level of each target defect feature set, determine the confidence model corresponding to each target defect feature set from the first feature extraction model and the second feature extraction model, and use the correspondence between the target defect feature set and the confidence model as the acceptance strategy corresponding to the target defect feature set. S09. Determine the defect detection results of the decorative paper based on the fourth defect feature set and the third defect feature set.

[0030] Specifically, the system is pre-trained by collecting a standard paper sample library containing a large number of known defect types. This standard paper sample library serves as the basis for defect judgment. During operation, firstly, the entire sample library is tested using a first feature extraction model (general contour model) and a second feature extraction model (various specialized models), generating first and second test results. Subsequently, the system performs refined analysis on the two test results: For each identified defect, the confidence level of the respective model output is calculated. Specifically, for overlapping defects identified by both models (i.e., the target defect feature set), the system compares the first and second confidence levels. Through pre-defined rules (e.g., selecting the model with consistently higher confidence, or using a weighted average method), the system explicitly assigns an optimal confidence model to each overlapping defect. Ultimately, all these defect type-confidence model correspondences are structured and stored, forming the final confidence strategy library.

[0031] When the system performs online inspection of new decorative paper, it makes a final judgment by integrating three aspects of information: first, the most reliable fourth defect feature set (derived from the first and second models) selected through an acceptance strategy; second, a third defect feature set reflecting the latest defect trends from a dynamic template library. The final defect detection result is a fusion of these three pieces of information, and the system will give a comprehensive judgment based on weights or logical rules (such as union or intersection).

[0032] In summary, by constructing a multi-level defect feature set, multi-dimensional capture of surface defects in decorative paper was achieved. The first defect feature set, based on edge contour analysis, effectively identifies contour shape defects. The second defect feature set, based on an image classification gating mechanism, employs a dedicated model for deep analysis of different texture categories, significantly improving the recognition accuracy for specific types of defects. The third defect feature set, through a real-time updated defect template library, ensures the timely identification of novel defects.

[0033] Example 2

[0034] Based on Embodiment 2, this embodiment identifies the second set of defect features of the decorative paper in step S03 based on the category of the surface image, including: When the surface image is classified as a solid color, a color uniformity template is matched against the surface image, and the second defect feature set of the decorative paper is determined based on the matched color uniformity template, including: The surface image is matched with multiple color distribution histogram templates to determine the target color distribution histogram that matches the surface image. Obtain the target color uniformity index associated with the target color distribution histogram; The second defect feature set is determined based on the target color uniformity index.

[0035] Specifically, when the system identifies the decorative paper to be tested as belonging to the solid color category through image classification, it immediately activates a dedicated color uniformity analysis module. This module divides the acquired surface image into multiple analysis regions and generates corresponding color distribution histograms. It then calculates the matching degree between these histograms and multiple pre-stored standard color distribution histogram templates, determining the closest target color distribution histogram through similarity comparison. The system then acquires a series of color uniformity quantification indicators associated with this target template, including but not limited to key parameters such as color difference value, color distribution standard deviation, and maximum color deviation value. By comparing these measured indicators with preset thresholds, a second defect feature set for solid color decorative paper is finally generated.

[0036] The above embodiments solve the most prominent quality control problem of color uniformity in the production of solid-color decorative paper. Compared with traditional manual visual inspection, this solution achieves accurate identification and objective evaluation of color unevenness defects such as color difference, color spots, and gradations through digital matching of color distribution histograms and quantitative index analysis.

[0037] Example 3

[0038] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps: Images of the moving decorative paper are captured using a visual camera to obtain an image of the paper's surface. Extract edge contour features from the surface image and determine the first defect feature set of the decorative paper based on the edge contour features; Based on the extracted surface images, classification and recognition are performed to obtain the categories of the surface images, and the second defect feature set of the decorative paper is identified based on the categories of the surface images. The third defect feature set of decorative paper is obtained by using a real-time updated defect template set. Based on the acceptance strategies corresponding to the first and second defect feature sets, the defect feature sets that conform to the acceptance strategies in the first and second defect feature sets are determined as the fourth defect feature set. The first defect feature set of multiple paper samples is identified by the first feature extraction model and used as the first test result. The second feature extraction model is used to identify a second set of defect features from multiple paper samples, which are then used as the second test results. Calculate the first confidence level of the first feature extraction model for each defect feature set contained in the first test result, and calculate the second confidence level of the second feature extraction model for each defect feature set contained in the second test result; For the target defect feature sets that overlap in the first test results and the second test results, based on the first confidence level and the second confidence level of each target defect feature set, the confidence model corresponding to each target defect feature set is determined from the first feature extraction model and the second feature extraction model, and the correspondence between the target defect feature set and the confidence model is used as the acceptance strategy corresponding to the target defect feature set. The defect detection results of the decorative paper are determined based on the fourth defect feature set and the third defect feature set.

[0039] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0040] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0041] Example 4

[0042] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps: Images of the moving decorative paper are captured using a visual camera to obtain an image of the paper's surface. Extract edge contour features from the surface image and determine the first defect feature set of the decorative paper based on the edge contour features; Based on the extracted surface images, classification and recognition are performed to obtain the categories of the surface images, and the second defect feature set of the decorative paper is identified based on the categories of the surface images. The third defect feature set of decorative paper is obtained by using a real-time updated defect template set. Based on the acceptance strategies corresponding to the first and second defect feature sets, the defect feature sets that conform to the acceptance strategies in the first and second defect feature sets are determined as the fourth defect feature set. The first defect feature set of multiple paper samples is identified by the first feature extraction model and used as the first test result. The second feature extraction model is used to identify a second set of defect features from multiple paper samples, which are then used as the second test results. Calculate the first confidence level of the first feature extraction model for each defect feature set contained in the first test result, and calculate the second confidence level of the second feature extraction model for each defect feature set contained in the second test result; For the target defect feature sets that overlap in the first test results and the second test results, based on the first confidence level and the second confidence level of each target defect feature set, the confidence model corresponding to each target defect feature set is determined from the first feature extraction model and the second feature extraction model, and the correspondence between the target defect feature set and the confidence model is used as the acceptance strategy corresponding to the target defect feature set. The defect detection results of the decorative paper are determined based on the fourth defect feature set and the third defect feature set.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for online detection of surface defects of decorative paper based on machine vision, characterized in that, The method comprises the following steps: S01, image acquisition of the moving decorative paper by a visual camera, to obtain a surface image of the decorative paper; S02, extracting edge contour features of the surface image, and determining a first defect feature set of the decorative paper according to the edge contour features; S03, classifying and identifying the surface image based on the extracted features, to obtain a category of the surface image, and identifying a second defect feature set of the decorative paper based on the category of the surface image; S04, obtaining a third defect feature set of the decorative paper through a real-time updated defect template set, and determining a fourth defect feature set from the first defect feature set and the second defect feature set that conforms to a corresponding acquisition strategy according to the first defect feature set and the second defect feature set; S05, identifying a first defect feature set of a plurality of paper samples through the first feature extraction model as a first test result; S06, identifying a second defect feature set of the plurality of paper samples through the second feature extraction model as a second test result; S07, calculating a first confidence degree of the first feature extraction model for each defect feature set contained in the first test result, and calculating a second confidence degree of the second feature extraction model for each defect feature set contained in the second test result; S08, for a target defect feature set that coincides in the first test result and the second test result, determining a corresponding confidence model of each target defect feature set from the first feature extraction model and the second feature extraction model according to the first confidence degree and the second confidence degree of each target defect feature set, and taking the corresponding relationship between the target defect feature set and the confidence model as an acquisition strategy corresponding to the target defect feature set; S09, determining a defect detection result of the decorative paper according to the fourth defect feature set and the third defect feature set.

2. The method for online detection of surface defects of decorative paper based on machine vision according to claim 1, characterized in that, The step S02 of extracting edge contour features of the surface image and determining a first defect feature set of the decorative paper according to the edge contour features comprises: recognizing the edge contour of the surface image; enlarging the edge contour by a predetermined multiple to obtain an enlarged edge contour; calculating a matching degree of a pixel point position of the enlarged edge contour and a standard pixel point position, and if the matching degree is lower than a preset threshold, determining that the first defect feature set of the decorative paper has a contour defect.

3. The method for online detection of surface defects of decorative paper based on machine vision according to claim 1, characterized in that, The step S03 of classifying and identifying the surface image to obtain a category of the surface image, and identifying a second defect feature set of the decorative paper based on the category of the surface image comprises: when a specific category of the surface image is determined through the first defect feature set, extracting and identifying features of the specific category in the surface image; determining a second defect feature of the decorative paper based on the features of the specific category.

4. The method for online detection of surface defects of decorative paper based on machine vision according to claim 3, characterized in that, The step of determining a second defect feature of the decorative paper based on the features of the specific category comprises: identifying the features of the specific category in the surface image to determine a defect label of the features of the specific category of the decorative paper; acquire a corresponding specific type defect model related to the defect label of the specific category feature of the surface image through a knowledge graph model; calculate the similarity between the surface image and the defect image corresponding to the corresponding specific type defect model, and determine the defect of the specific category feature of the decorative paper; determine the second defect feature of the decorative paper according to the defect of the specific category feature and the defect label of the specific category feature.

5. The method for online detection of surface defects of decorative paper based on machine vision according to claim 1, characterized in that, The step S03 includes: When the category of the surface image is a solid color category, match the surface image with a color uniformity template, and determine the second defect feature set of the decorative paper according to the matched color uniformity template.

6. A machine vision-based method for online detection of surface defects of decorative paper according to claim 5, characterized in that, The color uniformity template includes a color distribution histogram, and the matching of the surface image with the color uniformity template and the determination of the second defect feature set of the decorative paper according to the matching result include: matching the surface image with a plurality of color distribution histogram templates to determine a target color distribution histogram matching the surface image; acquiring a target color uniformity index associated with the target color distribution histogram; determining the second defect feature set according to the target color uniformity index.

7. The method for online detection of surface defects of decorative paper based on machine vision according to claim 1, characterized in that, The step S04 includes: acquiring a real-time updated defect template set and acquiring defect image data corresponding to the defect template set; calculating the similarity between the surface image and the defect image data, and screening a plurality of target defect images from the defect image data, the similarity of which meets a preset threshold; determining the third defect feature set of the decorative paper through the defect labels corresponding to the plurality of target defect images.

8. The method for online detection of surface defects of decorative paper based on machine vision according to claim 1, characterized in that, The step S04 includes: For each first defect feature set, if the model outputting the first defect feature set is a confidence model corresponding to the first defect feature set, the first defect feature set is determined as a fourth defect feature set; wherein the confidence degree of the first defect feature set output by the confidence model corresponding to the first defect feature set meets the acquisition strategy corresponding to the first defect feature set; For each second defect feature set, if the model outputting the second defect feature set is a confidence model corresponding to the second defect feature set, the second defect feature set is determined as a fourth defect feature set; wherein the confidence degree of the second defect feature set output by the confidence model corresponding to the second defect feature set meets the acquisition strategy corresponding to the second defect feature set. 9.A non-transitory computer-readable storage medium having stored therein at least one instruction or at least one piece of program, characterized in that, The at least one instruction or the at least one program is loaded and executed by the processor to implement the machine vision-based online detection method of the surface defect of the decorative paper.

10. An electronic device, comprising: The machine vision-based surface defect detection method for decorative paper comprises a processor and a memory, the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program is loaded and executed by the processor to realize the machine vision-based surface defect detection method for decorative paper as claimed in any one of claims 1-8.

Images