Latex skinning defect online detection method and system based on machine vision

Abstract

The invention discloses a latex skinning defect online detection method and system based on machine vision, and relates to the technical field of machine vision, and the method comprises the following steps: obtaining image acquisition parameters, and constructing an industrial camera imaging model based on a latex production environment; based on the industrial camera imaging model, collecting target area image data of the latex surface; sequentially executing edge enhancement processing and connected region extraction operation on the target region image data to obtain a candidate defect region set; sequentially extracting geometric morphological features from the candidate defect region set, and screening suspected defect regions through a hierarchical judgment rule; and identifying a target defect area according to the number and spatial distribution characteristics of the suspected defect areas. The technical problem that the skinning defect in the latex production process is difficult to identify online can be effectively solved, and the method has good application prospects and popularization value.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and in particular to a method and system for online detection of latex skinning defects based on machine vision. Background Art

[0002] With the development of the rubber industry, latex products are widely used in various fields. During the latex production process, due to factors such as environmental temperature and humidity, and uneven drying rates, latex surfaces are prone to crusting, affecting product quality and subsequent processing performance. To address this, the industry has gradually introduced machine vision-based quality inspection technologies, attempting to automate the identification of latex surface defects. Compared to traditional manual sampling methods, machine vision methods offer the advantages of non-contact, high efficiency, and continuous detection, and have already been initially applied in some production lines.

[0003] However, the existing detection methods still have many shortcomings, which are mainly reflected in the following aspects: (1) Limited image acquisition quality: Due to factors such as the strong reflectivity of the latex surface and the fast running speed of the production line, the acquired images are prone to blurring and uneven brightness, which makes it difficult to accurately extract the defective areas; (2) The image processing process lacks specificity: Most systems use general image processing algorithms and fail to optimize them in combination with the morphological characteristics of latex crusting defects, resulting in low recognition accuracy; (3) Single defect recognition logic: Existing technologies mostly rely on a single feature (such as area size) for judgment, making it difficult to effectively distinguish between real defects and texture interference, and false detections and missed detections occur frequently; (4) Poor system adaptability: Image acquisition parameters are mostly fixed settings, and cannot be dynamically matched according to actual working conditions such as changes in on-site lighting and product type switching, affecting detection stability.

[0004] These issues directly restrict the practicality and reliability of online latex skinning defect detection systems. Therefore, a detection method with a rational structure, clear process, and strong adaptability is urgently needed. This method can achieve efficient and stable recognition of latex surface skinning defects by building an image acquisition model adapted to the latex production environment, combining a specific sequence of image processing processes with a hierarchical and progressive defect recognition mechanism, thereby improving detection accuracy and system robustness. Summary of the Invention

[0005] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a method and system for online detection of latex skinning defects based on machine vision, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an online detection method for latex skinning defects based on machine vision, which comprises obtaining image acquisition parameters and constructing an industrial camera imaging model based on a latex production environment; Based on the industrial camera imaging model, collecting target area image data of the latex surface; performing edge enhancement processing and connected region extraction operations on the target area image data in sequence to obtain a set of candidate defect areas; Extracting geometric features from the candidate defect area set in sequence, and screening suspected defect areas using hierarchical judgment rules; The target defect area is identified based on the number and spatial distribution characteristics of the suspected defect areas.

[0009] As a preferred solution of the online detection method for latex skinning defects based on machine vision of the present invention, the construction of an industrial camera imaging model based on a latex production environment includes: Obtaining operating parameters of the latex production line, wherein the operating parameters include conveyor belt speed, ambient light intensity, and latex surface reflective properties; Determining the installation position and shooting angle of the industrial camera according to the operating parameters so that the image acquisition area covers the area prone to latex skinning defects; Set the exposure time, gain factor, and white balance parameters of the industrial camera, configure the type and illumination angle of the fill light source, and establish an image acquisition control model.

[0010] As a preferred embodiment of the online detection method for latex skinning defects based on machine vision of the present invention, the acquisition of target area image data of the latex surface includes: Start the industrial camera and fill light system and enter the image acquisition preparation state; When the production line is in continuous operation, the image acquisition frequency is set according to the conveyor belt speed; The trigger signal synchronizes the industrial camera with the production line motion status to locate and capture the target area on the latex surface; Capture multiple frames of images and timestamp them.

[0011] As a preferred solution of the online detection method for latex skinning defects based on machine vision of the present invention, the step of obtaining a set of candidate defect regions includes: Grayscale the collected target area image to obtain a single-channel image matrix; Performing edge detection on the single-channel image matrix using the Canny operator to generate a binary edge image; Performing a morphological closing operation on the binary edge image to connect broken edges and remove isolated noise points; Based on the connected component analysis method, independent regions in the image are extracted and used as candidate defect regions.

[0012] As a preferred embodiment of the online detection method for latex skinning defects based on machine vision of the present invention, the screening of suspected defect areas includes: Obtain the pixel coordinate set of the candidate defect area and calculate its minimum bounding rectangle and contour boundary; Calculating the area, perimeter, and equivalent diameter of the candidate defect region based on the pixel coordinate set and contour boundary information; Calculating the aspect ratio, circularity, and edge closure of the candidate defect region based on the area and perimeter; The above geometric feature parameters are combined into a defect feature parameter set.

[0013] As a preferred embodiment of the online detection method for latex skinning defects based on machine vision of the present invention, the identification of target defect areas includes: Loading a preset defect determination threshold set, wherein the defect determination threshold set includes an area threshold, an aspect ratio threshold, and an edge closure threshold; The geometric morphological feature parameters of each candidate defect area are sequentially compared with the corresponding defect judgment threshold to determine whether the defect recognition conditions are met; Mark the candidate defect areas that meet the identification conditions and count their quantity and distribution information; Output the marked target defect area and its identification results to the subsequent processing module or control system.

[0014] As a preferred embodiment of the online detection method for latex skinning defects based on machine vision of the present invention, the multi-level comparison includes: Determine whether the area is greater than the minimum threshold, determine whether the aspect ratio falls into the abnormal range, and verify whether the edge closure exceeds the set standard. When all three items are met, it is identified as a valid defect area.

[0015] In a second aspect, the present invention provides an online detection system for latex skinning defects based on machine vision, which includes: a model building module for obtaining image acquisition parameters and building an industrial camera imaging model based on a latex production environment; A data acquisition module, for collecting image data of a target area on the latex surface based on the industrial camera imaging model; A candidate defect module is used to sequentially perform edge enhancement processing and connected region extraction operations on the target area image data to obtain a set of candidate defect areas; A defect screening module is used to sequentially extract geometric features from the candidate defect area set and screen suspected defect areas using hierarchical judgment rules; The defect identification module is used to identify the target defect area according to the number and spatial distribution characteristics of the suspected defect areas.

[0016] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements the steps of an online detection method for latex skinning defects based on machine vision.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, the steps of an online detection method for latex skinning defects based on machine vision are implemented.

[0018] Compared with the existing technology, the beneficial effect of the present invention is that by constructing an industrial camera imaging model adapted to the latex production environment and combining image processing and geometric morphological feature analysis methods, it can achieve efficient and stable identification of latex surface crusting defects in the production line. The method sets image acquisition parameters by obtaining operating parameters, thereby improving image contrast and edge clarity and enhancing the system's adaptability under different lighting and motion conditions. An image processing process of first enhancing edges and then extracting connected regions is adopted to improve the integrity and recognizability of defect region boundaries. Geometric features such as area, aspect ratio, circularity, and edge closure are extracted from candidate defect regions, and multi-level screening is performed in combination with hierarchical judgment rules to effectively distinguish between real defects and texture interference, thereby improving recognition accuracy. A preset defect judgment threshold set and a multi-level comparison mechanism are introduced to ensure good repeatability and consistency of the entire recognition process, facilitating its promotion and application. At the same time, the image acquisition and recognition process supports timestamp marking and result output, providing a reliable basis for subsequent control systems to perform rejection, alarms, or process adjustments, thus meeting the real-time and intelligent requirements of online detection. This method does not rely on complex AI models, but only uses traditional image processing and feature analysis techniques to achieve high-precision detection, reducing system deployment costs and maintenance difficulties. It has the advantages of reasonable structure, simple operation, high detection efficiency, and strong applicability. It can effectively solve the technical problem of difficult online identification of skinning defects in latex production, and has good application prospects and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 A flowchart of a method and system for online detection of latex skinning defects based on machine vision provided by one embodiment of the present invention; Figure 2 An internal structural diagram of a computer device for an online detection method and system for latex skinning defects based on machine vision provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0021] To make the above-mentioned objects, features, and advantages of the present invention more easily understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Unless specifically stated otherwise, the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure.

[0024] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not necessarily drawn according to the actual proportional relationship.

[0025] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0026] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0027] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0028] In addition, in the description of the present disclosure, the terms "first," "second," "third," etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance or order. Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order, or that all illustrated operations be performed to achieve the desired result. In certain circumstances, multitasking and parallel processing can be advantageous.

[0029] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an online detection method for latex skinning defects based on machine vision, comprising: Figure 1 The following is a flowchart of a method and system for online detection of latex skinning defects based on machine vision, including: S1: Obtain image acquisition parameters and build an industrial camera imaging model based on the latex production environment; S2: Based on the industrial camera imaging model, the target area image data of the latex surface is collected; S3: performing edge enhancement processing and connected region extraction operations on the target area image data in sequence to obtain a set of candidate defect areas; S4: Extract geometric features from the candidate defect area set in sequence, and screen suspected defect areas using hierarchical judgment rules; S5: Identify target defect areas based on the number and spatial distribution characteristics of suspected defect areas.

[0030] It should be noted that during the operation of the latex production line, its surface is prone to crusting defects due to factors such as environmental temperature and humidity, and uneven drying speed. This type of defect has the characteristics of blurred edges, random distribution, and diverse shapes. It is easy to be confused with normal textures during the image acquisition process, resulting in recognition errors. Due to the strong reflectivity of the latex surface and the continuous operation of the production line, the image acquisition process is susceptible to interference such as changes in lighting and motion blur, making it difficult for traditional image processing methods to achieve stable detection. In addition, existing detection systems mostly use a single feature to determine the defect type, lack a comprehensive analysis of the geometric morphology of the defect area, and it is difficult to accurately distinguish between real defects and false detection interference.

[0031] Therefore, in order to address the above-mentioned problems of large image acquisition interference and low defect recognition accuracy, the present invention constructs an industrial camera imaging model suitable for the latex production environment through steps S1 to S5 to improve image contrast; adopts a specific order of image processing flow to first perform edge enhancement and then extract connected areas to enhance the defect boundary recognition ability; further based on the geometric morphological features of the candidate defect area, a hierarchical and progressive defect recognition structure is adopted, and different feature parameters are used in turn for screening and confirmation, thereby improving the accuracy and stability of latex skinning defect recognition.

[0032] Example 2, reference Figure 1 , which is the second embodiment of the present invention, this embodiment further provides a specific step of an online detection method for latex skinning defects based on machine vision, including: In the embodiment of the present application, in step S1, an industrial camera imaging model based on a latex production environment is constructed by the following steps, specifically: S101: Acquiring operating parameters of a latex production line, including conveyor belt speed, ambient light intensity, and latex surface reflective properties; S102: Determine the installation position and shooting angle of the industrial camera according to the operating parameters so that the image acquisition area covers the area prone to latex skinning defects; S103: Setting the exposure time, gain factor, and white balance parameters of the industrial camera to obtain high-contrast images under dynamic production conditions; S104: configuring the type and illumination angle of the fill light source to ensure uniform illumination of the latex surface and reduce shadow interference; S105: An image acquisition control model is established based on the above parameters, and the improvement effect of the model on edge clarity and texture contrast is verified by simulating image quality evaluation indicators.

[0033] In one optional embodiment, operating parameters include conveyor belt speed, production line layout, and information about typical distribution areas of latex skinning defects. Based on these operating parameters, geometric modeling is used to calculate the optimal installation height and pitch angle of the industrial camera, ensuring that the image acquisition area fully covers areas with a high incidence of latex skinning defects while avoiding image distortion or blind spots caused by improper installation angles.

[0034] In one optional embodiment, actual operational data from the latex production line is incorporated into the geometric modeling process, including conveyor belt movement direction, latex flow state, and the spatial distribution of defect frequencies. The industrial camera is installed above an area where the probability of defect occurrence exceeds a preset threshold, with a pitch angle of 15° to 30°. This ensures that the image capture clearly captures the latex surface texture while minimizing image oversaturation caused by reflections.

[0035] In an optional embodiment, a laser rangefinder is used to assist in positioning the industrial camera, maintaining it between 40cm and 60cm from the latex surface. A rotating bracket is used to adjust the angle between the camera lens axis and the conveyor belt's direction of travel to 10° to 20°, thereby achieving higher edge clarity and defect recognition rates under dynamic shooting conditions. This setup is particularly suitable for inspection scenarios involving fast production lines and subtle defects.

[0036] In an optional embodiment, the exposure time and gain factor of the industrial camera are set based on the latex production line's operating speed and ambient lighting conditions, ensuring that the captured image retains sufficient detail without overexposure. Furthermore, manually setting the white balance parameter eliminates the impact of on-site light source color temperature differences on image color, improving overall image contrast and the visibility of defect areas.

[0037] In an optional embodiment, when setting the exposure time, the distance the latex surface moves per unit time is calculated in combination with the conveyor belt running speed, and the exposure time is controlled to a value that makes the image motion blur no more than one pixel; the gain coefficient is adaptively adjusted according to the ambient light intensity to ensure that the image brightness remains stable under different lighting conditions; the white balance parameters are set based on the standard gray plate calibration results to obtain accurate color reproduction effects.

[0038] In an optional embodiment, a high-speed backlight compensation light source is used in combination with a short exposure strategy, the exposure time is set to 500μs~1ms, and the gain coefficient is controlled between 1.0 and 2.0 times to avoid the introduction of noise due to excessive gain; the white balance is calibrated by selecting multiple regions of interest in the defect-free area and performing average color temperature analysis, thereby effectively improving the grayscale difference between the edge of the crusting defect and the normal area, and enhancing the segmentation accuracy in the subsequent image processing stage.

[0039] In the embodiment of the present application, image data is collected based on the imaging model in step S2, specifically: S201: Start the industrial camera and the fill light system to prepare for image acquisition; S202: When the production line is in continuous operation, the image acquisition frequency is set according to the conveyor belt speed to ensure that there are no missed inspection areas between adjacent images; S203: Synchronizing the industrial camera with the motion state of the production line through trigger signals to achieve accurate positioning and acquisition of the target area on the latex surface; S204: Capturing multiple frames of images and marking them with time stamps to support timing analysis in subsequent image processing; S205: The collected image data is transmitted to the image processing module, and the corresponding collection parameter information is recorded for subsequent defect recognition process.

[0040] In an optional embodiment, the system automatically initializes the industrial camera and the fill light system after receiving the start detection instruction, and confirms whether the equipment operating status is normal through the communication interface.

[0041] In an optional embodiment, during the initialization process, the system performs self-inspections on the image sensor, lens driver module, and light source controller of the industrial camera in sequence, and issues an early warning signal when an abnormality is found, prompting the operator to handle it in a timely manner.

[0042] In an optional embodiment, the system automatically loads the preset image acquisition parameters after completing the device self-test and adjusts the brightness of the fill light system to the calibration value, so that the entire image acquisition system enters a stable standby state, waiting for an external trigger signal to start the formal acquisition process.

[0043] In an optional embodiment, the system obtains the current conveyor belt running speed through an encoder and calculates the number of image frames that should be collected per unit time based on this to match the movement rhythm of the latex product.

[0044] In an optional embodiment, when the conveyor belt speed is 0.8 m / s, the system sets the image acquisition frequency to 40 frames per second; when the speed changes, the system uses linear interpolation to dynamically adjust the acquisition frequency to ensure that adjacent image coverage areas are not missed.

[0045] In an optional embodiment, the system is set with a minimum image overlap rate threshold, for example, requiring at least 10% overlap area between adjacent images to ensure the continuity and integrity of defect identification and prevent missed detection due to improper acquisition frequency.

[0046] In an optional embodiment, the system uses a pulse signal generated by an external photoelectric sensor or encoder as a trigger source for image acquisition to ensure that each image acquisition moment corresponds to the position of the latex product.

[0047] In an optional embodiment, the image acquisition trigger signal is linked to the production line PLC control system. When the latex product reaches the predetermined acquisition position, the system automatically triggers the camera to shoot, thereby achieving precise matching of spatial coordinates.

[0048] In an optional embodiment, the system also has a trigger delay compensation function, which can dynamically adjust the trigger timing according to the mechanical transmission and image acquisition response time, so that the image acquisition error is controlled within ±1 pixel, thereby improving the image positioning accuracy.

[0049] In an optional embodiment, the system records the corresponding timestamp information of each frame of image while capturing it, so that the subsequent image processing module can perform sequence analysis and determine the defect evolution trend.

[0050] In an optional embodiment, the timestamp information includes year, month, day, hour, minute, second, and millisecond-level accuracy, and is synchronized with the time base of the production line control system to ensure accurate association of image data with production events.

[0051] In an optional embodiment, the image data and its timestamp information are stored together in a local cache area and can be exported for quality traceability and process optimization analysis, thereby improving the manageability and traceability of the image data.

[0052] In an optional embodiment, after image acquisition is completed, the system transmits the image data to the image processing module at high speed via a Gigabit Ethernet or Camera Link interface, and simultaneously attaches acquisition parameter information.

[0053] In an optional embodiment, additional acquisition parameters include exposure time, gain factor, white balance setting, light source intensity, etc. The image processing module can dynamically adjust the sensitivity and threshold of the defect recognition algorithm according to these parameters.

[0054] In an optional embodiment, after receiving the image data, the image processing module first verifies whether the acquisition parameters are within the preset range. If a parameter is found to exceed the allowable deviation, the image is marked and weighted processing is performed during the defect recognition process to improve the reliability of the recognition results.

[0055] In the embodiment of the present application, step S3 implements the processing of the target area image data through the following steps, specifically: S301: grayscale the collected target area image to obtain a single-channel image matrix; S302: Using the Canny operator to perform edge detection on the single-channel image matrix to generate a binary edge image; S303: performing a morphological closing operation on the binary edge image to connect broken edges and remove isolated noise points; S304: Extracting independent regions from the image based on the connected component analysis method, and outputting them as candidate defect regions to the next processing stage.

[0056] In an optional embodiment, the system uses a weighted averaging method to convert a color image into a grayscale image to reduce the computational complexity of subsequent image processing and retain the main structural information of the image.

[0057] In an optional embodiment, the system dynamically adjusts the weight coefficients in the grayscale algorithm according to the light source intensity and white balance parameters during image acquisition, so that images acquired under different lighting conditions can obtain consistent grayscale representation, thereby improving the robustness of image processing.

[0058] In an optional embodiment, the system uses the Canny edge detection algorithm to process the grayscale image and extracts the image edge information through steps such as Gaussian filtering noise reduction, gradient calculation, non-maximum suppression and double threshold processing.

[0059] In an optional embodiment, the high and low thresholds used in the Canny algorithm are automatically set based on the overall grayscale distribution of the image, for example, the 75% and 25% quantiles of the image grayscale histogram are taken as initial thresholds to adapt to image inputs under different contrast conditions.

[0060] In an optional embodiment, the system introduces local contrast enhancement preprocessing before performing Canny edge detection to highlight the boundary difference between the latex skinning defect area and the normal area, thereby improving the accuracy and completeness of the edge detection results.

[0061] In an optional embodiment, the system sequentially performs dilation and erosion operations on the binary image after edge detection, namely, morphological closing operations, to repair edge breaks and eliminate small-area noise interference in the image.

[0062] In an optional embodiment, the structural element used in the morphological operation is a 3×3 rectangular kernel, the dilation operation expands the edge area, and the erosion operation restores the original contour size, thereby effectively connecting adjacent but unclosed edges and improving the integrity of the defect area.

[0063] In an optional embodiment, the system dynamically adjusts the size of the structural element according to the typical size of the defect area in the image. For example, when the defect diameter is expected to be large, a 5×5 structural element is used to obtain stronger edge connectivity to prevent region mis-breakage due to over-refinement.

[0064] In an optional embodiment, the system uses a connected domain labeling algorithm based on pixel eight-neighborhood search to identify all interconnected foreground regions in the image and assign a unique label to each region.

[0065] In an optional embodiment, after completing the connected domain labeling, the system counts the number of pixels in each region and filters out small regions with an area smaller than a preset threshold to eliminate image noise or non-defective texture interference.

[0066] In an optional embodiment, the system further screens the candidate regions in combination with regional shape factors (such as aspect ratio and circularity), and retains only the regions that meet the morphological characteristics of latex crusting defects as the final candidate defect regions for subsequent classification and recognition module processing.

[0067] In the embodiment of the present application, in step S4, feature extraction and parameter set construction of the candidate defect area are achieved through the following steps, specifically: S401: Obtain a set of pixel coordinates of a candidate defect area and calculate its minimum bounding rectangle and contour boundary; S402: Calculating the area, perimeter, and equivalent diameter of the candidate defect region based on the pixel coordinate set and contour boundary information; S403: Calculating the aspect ratio, circularity, and edge closure of the candidate defect region based on the area and perimeter; S404: Combining the above-mentioned geometric morphological feature parameters into a defect feature parameter set, and outputting it to the defect recognition module for subsequent classification and judgment.

[0068] In an optional embodiment, the system obtains all pixel coordinates of each candidate defect area by traversing the image label matrix, and calculates the minimum bounding rectangle and closed contour boundary of the area based on these coordinate points.

[0069] In an optional embodiment, when obtaining the minimum bounding rectangle, the system uses the boundingRect function in OpenCV to achieve fast calculation, and records the center position and direction angle of the rectangle for subsequent analysis of the spatial distribution characteristics of the defect.

[0070] In an optional embodiment, the system performs polygonal approximation on the boundary points after extracting the contour boundary to reduce the amount of boundary data and improve the efficiency of subsequent feature calculations, while retaining key shape information to ensure the accuracy of defect identification.

[0071] In an optional embodiment, the system counts the total number of pixels in the candidate defect area as the area value based on the pixel coordinate set, and calculates the perimeter and equivalent diameter of the area in combination with the contour boundary points.

[0072] In an optional embodiment, the system introduces an image resolution factor when calculating the above parameters to convert pixel units into actual physical dimensions, thereby improving measurement accuracy and facilitating horizontal comparative analysis between different batches of products.

[0073] In an optional embodiment, the system calculates the aspect ratio using the width and height of the minimum circumscribed rectangle, calculates the circularity by combining the area and perimeter, and determines the edge closure by the contour closure to generate multiple derived features describing the defect morphology.

[0074] In an optional embodiment, the system further introduces contour curvature analysis to identify whether there are breaks or depressions on the edge, and refines the quantitative expression of edge closure accordingly to more accurately distinguish between crusting defects and other surface texture interferences.

[0075] In an optional embodiment, the system organizes the extracted geometric features into feature vectors in a preset order, and appends acquisition timestamp and image number information to form a structured defect feature parameter set.

[0076] In an optional embodiment, the defect feature parameter set is encapsulated as a JSON format data packet and transmitted to the defect identification module through an internal communication interface to facilitate subsequent rule-based or model-based classification processing.

[0077] In an optional embodiment, the system also supports standardization of feature parameter sets to eliminate feature fluctuations caused by changes in image acquisition conditions and improve the consistency and stability of subsequent defect recognition results.

[0078] In the embodiment of the present application, in step S5, defect identification based on the defect feature parameter set is achieved through the following steps, specifically: S501: Loading a preset defect determination threshold set, the threshold set including an area threshold, an aspect ratio threshold, and an edge closure threshold; S502: Compare the geometric morphological feature parameters of each candidate defect area with the corresponding defect judgment threshold in turn to determine whether the defect recognition condition is met; S503: Marking candidate defect areas that meet the identification conditions, and counting their number and distribution information; S504: Output the marked target defect area and its recognition result to a subsequent processing module or a control system.

[0079] In an optional embodiment, before performing defect identification, the system loads a set of preset defect determination thresholds from a local configuration file or a database for classifying and determining candidate defect areas.

[0080] In an optional embodiment, the defect judgment threshold set is set according to the typical size and morphological characteristics of latex skinning defects, for example, the area threshold is 50 to 200 pixels², the aspect ratio threshold is 1.2 to 3.0, and the edge closure threshold is above 0.8.

[0081] In an optional embodiment, the system supports dynamic switching of multiple groups of judgment thresholds according to different product batches or production environments, and provides a threshold editing function through the user interface to adapt to different detection requirements and enhance the flexibility of the system.

[0082] In an optional embodiment, the system sequentially reads characteristic parameters such as the area, aspect ratio, and edge closure of the candidate defect area, and compares them item by item with the corresponding judgment thresholds to comprehensively determine whether it is a target defect.

[0083] In an optional embodiment, a multi-level decision logic is adopted: first, it is determined whether the area is greater than the minimum threshold, then it is determined whether the aspect ratio falls into the abnormal range, and finally it is verified whether the edge closure exceeds the set standard. Only when all three items are met can it be identified as a valid defect area.

[0084] In an optional embodiment, the system introduces a weighted scoring mechanism to assign different weights to various feature parameters, calculate a comprehensive score and compare it with the total score threshold, thereby improving the robustness and adaptability of the recognition process.

[0085] In an optional embodiment, the system visually marks the identified target defect area, draws a bounding box or a color-filled layer on the original image, and records its position coordinates and feature parameters.

[0086] In an optional embodiment, the system automatically counts the number of target defect areas, their total area percentage, and spatial distribution density, and generates a defect distribution heat map to assist in quality analysis and process improvement.

[0087] In an optional embodiment, the system saves the defect statistics of each inspection to a database and supports query and comparative analysis by time, batch or product type to facilitate long-term quality trend tracking.

[0088] In an optional embodiment, the system packages the marked target defect area image data and the recognition results and sends them to the image display module or the alarm control module for the operator to view or trigger corresponding actions.

[0089] In an optional embodiment, the recognition result includes the number of defects, grade identification, location information and feature parameter set, and supports multiple output formats, such as JSON, XML or structured text files, to facilitate integration with other systems.

[0090] In an optional embodiment, the system automatically selects an output channel according to the defect level: if it is a minor defect, only a log is recorded; if it is a serious defect, a rejection mechanism or an alarm device is immediately triggered to achieve closed-loop control.

[0091] Example 3 is an embodiment of the present invention. This embodiment also provides an online detection system for latex skinning defects based on machine vision. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0092] To verify the practical application of the proposed "machine vision-based online latex skinning defect detection method," a continuous latex production line at a rubber products company was selected as the experimental subject. This line primarily produces medical latex gloves. During the drying process, the latex coating is susceptible to skinning defects due to temperature and humidity fluctuations, which impacts the subsequent curing quality of the product. Existing detection systems rely on manual spot checks combined with image processing using a single area threshold, resulting in high missed detection rates and unstable recognition.

[0093] In this experiment, an online inspection system based on industrial cameras was established without changing the existing production line structure. First, an imaging model adapted to the latex production environment was constructed by collecting operating parameters such as conveyor speed, ambient light intensity, and the reflective properties of the latex surface. The industrial camera was set at a mounting height of 50 cm, a shooting angle of 20°, an exposure time of 800 μs, a gain factor of 1.5, and a ring LED fill light to enhance image contrast.

[0094] Subsequently, image acquisition was performed while the production line was running continuously, with a frequency of 40 frames per second to ensure that no adjacent images were missed. The captured target area images were then subjected to grayscale conversion, Canny edge enhancement, morphological closing operations, and connected domain analysis to extract candidate defect regions. Furthermore, geometric features such as area, perimeter, aspect ratio, circularity, and edge closure were extracted for each candidate region, and a feature parameter set was constructed. Finally, the target defect regions were screened using pre-set multi-level judgment rules (area > 60 pixels², aspect ratio > 1.5, edge closure > 0.85), and the recognition results were output to the control system.

[0095] The experiment lasted seven days, collecting approximately 50,000 sample images, covering production conditions from different batches and under varying temperature and humidity conditions. All image data was synchronously recorded with timestamps, acquisition parameters, and recognition results for subsequent statistical analysis and effectiveness evaluation.

[0096] Table 1: Comparison of defect recognition accuracy of samples in each batch under different detection methods (unit: %)

[0097] Note: Method A: Traditional single area threshold method; Method B: a method that introduces edge enhancement but does not use hierarchical decision rules; Method C: The method proposed by the present invention.

[0098] Table 2: Comparison of key performance indicators under different detection methods (unit: % or pixel)

[0099] It can be seen from the data in Table 1 and Table 2 that the method proposed in the present invention is significantly superior to existing detection methods in multiple key performance indicators, reflecting good innovation and practicality.

[0100] First, in terms of recognition accuracy, the proposed method C achieved an average of 93.1%, significantly higher than the 73.5% of traditional method A and the 82.7% of improved method B. This demonstrates that the introduction of geometric feature extraction and hierarchical judgment rules effectively enhances the ability to distinguish between real defects and texture interference, thereby significantly improving detection accuracy.

[0101] Secondly, the missed detection rate and false detection rate of the present invention are controlled at 4.1% and 2.8%, respectively, which are significantly lower than the 18.7% and 7.8% of Method A, and better than the 12.3% and 5.0% of Method B. This advantage is mainly due to the optimization of the image processing process, specifically the sequence of edge enhancement followed by connected region extraction, which enhances the integrity of defect boundaries and reduces the probability of false positives.

[0102] Thirdly, in terms of image acquisition stability and system adaptability, the proposed method demonstrated greater robustness under varying lighting and working conditions, with an image acquisition stability score of 92.4% and a system adaptability score of 90.1%, both outperforming the other two methods. This is due to an imaging model customized for the latex production environment, including dynamically adjusted light source configuration and camera parameter settings, which enables the system to maintain consistent image quality in complex field environments.

[0103] In addition, the recognition response time of the three methods is not much different, all controlled within 110ms, meeting the real-time requirements of online detection. However, the present invention achieves higher recognition accuracy while maintaining response speed, indicating that its overall performance is better.

[0104] In summary, the technical solution proposed in this paper not only addresses the problems of poor image acquisition quality, simplistic recognition logic, and insufficient system adaptability found in existing detection methods, but also significantly improves the accuracy and stability of defect recognition by introducing a specific image processing process, a geometric feature extraction mechanism, and multi-level judgment rules. These technological improvements are significantly creative and novel, suitable for quality monitoring scenarios in various latex product production lines, and possess broad application prospects and promotional value.

[0105] Example 4, with reference to Figure 2 , which is an embodiment of the present invention, this embodiment also provides an online detection system for latex skinning defects based on machine vision, comprising: A model building module is used to obtain image acquisition parameters and build an industrial camera imaging model based on the latex production environment; A data acquisition module, used for collecting image data of a target area on the latex surface based on an industrial camera imaging model; The candidate defect module is used to perform edge enhancement processing and connected region extraction operations on the target area image data in sequence to obtain a set of candidate defect areas; The defect screening module is used to sequentially extract geometric features from the set of candidate defect areas and screen suspected defect areas using hierarchical judgment rules; The defect recognition module is used to identify the target defect area based on the number and spatial distribution characteristics of the suspected defect areas.

[0106] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 2 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for online detection of latex skinning defects based on machine vision is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0107] This embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the following steps are implemented: Obtain image acquisition parameters and build an industrial camera imaging model based on the latex production environment; Based on the industrial camera imaging model, the target area image data of the latex surface is collected; Perform edge enhancement processing and connected region extraction operations on the target area image data in sequence to obtain a set of candidate defect areas; Extract geometric features from the candidate defect area set in sequence, and screen suspected defect areas using hierarchical judgment rules; Identify target defect areas based on the number and spatial distribution characteristics of suspected defect areas.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0109] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a fully hardware embodiment, a fully software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application may be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0110] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0111] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0113] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0114] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A method for online detection of latex skinning defects based on machine vision, characterized by: include, Obtain image acquisition parameters and build an industrial camera imaging model based on the latex production environment; Based on the industrial camera imaging model, collecting target area image data of the latex surface; performing edge enhancement processing and connected region extraction operations on the target area image data in sequence to obtain a set of candidate defect areas; Extracting geometric features from the candidate defect area set in sequence, and screening suspected defect areas using hierarchical judgment rules; The target defect area is identified based on the number and spatial distribution characteristics of the suspected defect areas.

2. The method for online detection of latex skinning defects based on machine vision according to claim 1, wherein: The construction of an industrial camera imaging model based on a latex production environment includes: Obtaining operating parameters of the latex production line, wherein the operating parameters include conveyor belt speed, ambient light intensity, and latex surface reflective properties; Determining the installation position and shooting angle of the industrial camera according to the operating parameters so that the image acquisition area covers the area prone to latex skinning defects; Set the exposure time, gain factor, and white balance parameters of the industrial camera, configure the type and illumination angle of the fill light source, and establish an image acquisition control model.

3. The online detection method for latex skinning defects based on machine vision according to claim 2, characterized in that: The target area image data of the latex surface collected includes: Start the industrial camera and fill light system and enter the image acquisition preparation state; When the production line is in continuous operation, the image acquisition frequency is set according to the conveyor belt speed; The trigger signal synchronizes the industrial camera with the production line motion status to locate and capture the target area on the latex surface; Capture multiple frames of images and timestamp them.

4. The method for online detection of latex skinning defects based on machine vision according to claim 3, wherein: The obtaining of a set of candidate defect regions comprises: Grayscale the collected target area image to obtain a single-channel image matrix; Performing edge detection on the single-channel image matrix using the Canny operator to generate a binary edge image; Performing a morphological closing operation on the binary edge image to connect broken edges and remove isolated noise points; Based on the connected component analysis method, independent regions in the image are extracted and used as candidate defect regions.

5. The online detection method for latex skinning defects based on machine vision according to claim 4, characterized in that: The screening of suspected defect areas includes: Obtain the pixel coordinate set of the candidate defect area and calculate its minimum bounding rectangle and contour boundary; Calculating the area, perimeter, and equivalent diameter of the candidate defect region based on the pixel coordinate set and contour boundary information; Calculating the aspect ratio, circularity, and edge closure of the candidate defect region based on the area and perimeter; The above geometric feature parameters are combined into a defect feature parameter set.

6. The method for online detection of latex skinning defects based on machine vision according to claim 5, wherein: The identifying target defect area includes: Loading a preset defect determination threshold set, wherein the defect determination threshold set includes an area threshold, an aspect ratio threshold, and an edge closure threshold; The geometric morphological feature parameters of each candidate defect area are sequentially compared with the corresponding defect judgment threshold to determine whether the defect recognition conditions are met; Mark the candidate defect areas that meet the identification conditions and count their quantity and distribution information; Output the marked target defect area and its identification results to the subsequent processing module or control system.

7. The online detection method for latex skinning defects based on machine vision according to claim 6, characterized in that: The multi-level comparison includes: Determine whether the area is greater than the minimum threshold, determine whether the aspect ratio falls into the abnormal range, and verify whether the edge closure exceeds the set standard. When all three items are met, it is identified as a valid defect area.

8. A machine vision-based online detection system for latex skinning defects, based on the machine vision-based online detection method for latex skinning defects according to any one of claims 1 to 7, characterized in that: include, A model building module is used to obtain image acquisition parameters and build an industrial camera imaging model based on the latex production environment; A data acquisition module, for collecting image data of a target area on the latex surface based on the industrial camera imaging model; A candidate defect module is used to sequentially perform edge enhancement processing and connected region extraction operations on the target area image data to obtain a set of candidate defect areas; A defect screening module is used to sequentially extract geometric features from the candidate defect area set and screen suspected defect areas using hierarchical judgment rules; The defect identification module is used to identify the target defect area according to the number and spatial distribution characteristics of the suspected defect areas.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the online detection method for latex skinning defects based on machine vision according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the online detection method for latex skinning defects based on machine vision according to any one of claims 1 to 7 are implemented.

Images