Condom cleaning and drying system based on machine vision and control method

Abstract

The invention discloses a condom cleaning and drying system based on machine vision and a control method, and relates to the technical field of automatic control of condom production and processing equipment. Based on a preprocessed image, color features, texture features and morphological features are extracted through a multi-feature fusion stain recognition algorithm; inputting into a pre-trained stain level judgment model, and analyzing based on a stain coverage area calculation result to obtain a stain level; step 3, the main control module executes a targeted cleaning and drying control strategy according to the judged stain level in combination with the real-time operation parameters collected by the feedback regulation module; the condom image is collected and analyzed in real time through machine vision, the stain level is accurately judged in combination with a multi-feature fusion algorithm, targeted adjustment of a cleaning and drying strategy is achieved, and the condom cleaning cleanliness is effectively improved; and by dynamically optimizing control parameters, excessive cleaning and drying are avoided, and energy consumption is remarkably reduced.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for condom production and processing equipment, and in particular to a condom cleaning and drying system and control method based on machine vision. Background Technology

[0002] In the condom manufacturing process, cleaning and drying are crucial steps to ensure product cleanliness and safety. Existing condom cleaning and drying systems mostly employ fixed-parameter control modes, meaning they complete the entire process according to preset parameters such as cleaning time, spray pressure, water temperature, drying temperature, and airflow. This makes it impossible to dynamically adjust the control strategy based on the actual soiling condition of the condoms.

[0003] The existing technology has the following shortcomings:

[0004] Using fixed cleaning and drying parameters can lead to over-cleaning and drying of condoms with light stains, resulting in energy waste and potential damage to the condom material. For condoms with heavy stains, fixed parameters may not guarantee cleanliness, affecting product quality.

[0005] The lack of precise criteria for stain assessment means that existing systems often rely on manual observation or simple sensor detection of cleaning fluid turbidity to indirectly determine stain levels. Manual observation is inefficient and highly subjective, while sensor detection is susceptible to interference and cannot accurately reflect the actual stain severity of a single condom, resulting in poorly targeted control strategies. Furthermore, the lack of a dynamic optimization mechanism prevents adjustments to control parameters based on real-time operating parameters and stain changes, leading to insufficient system adaptability and control precision, and making it difficult to balance improved cleanliness with energy efficiency. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a condom cleaning and drying system and control method based on machine vision. The technical solution adopted is as follows:

[0007] A machine vision-based control method for a condom washing and drying system includes the following steps:

[0008] Step 1: After starting the cleaning program, the vision acquisition module acquires images of the condoms inside the cleaning and drying machine at preset intervals, and preprocesses the acquired raw images.

[0009] Step 2: Based on the preprocessed image, extract color features, texture features and morphological features through a multi-feature fusion stain recognition algorithm, input them into a pre-trained stain level determination model, and analyze the stain level based on the stain coverage area calculation results.

[0010] Step 3: The main control module executes targeted cleaning and drying control strategies based on the determined stain level and the real-time operating parameters collected by the feedback adjustment module.

[0011] Step 4: When the stain level of the condom stabilizes at level 0 and the drying humidity reaches the preset threshold, stop all execution modules, complete the cleaning and drying process, and store key operating data.

[0012] Optionally, step 5 is also included: real-time acquisition of operating parameters and transmission to the main control module, and dynamic optimization of the control strategy based on image analysis results.

[0013] Optionally, the dynamic optimization control strategy in step 5 specifically includes: if the turbidity of the cleaning fluid exceeds the threshold, replacing the cleaning fluid or turning on the high-efficiency filter; if the drying temperature exceeds the upper limit, reducing the hot air temperature and increasing the wind speed; if the humidity decreases too slowly, increasing the wind speed; and optimizing the adjustment parameters and image acquisition interval time corresponding to each stain level through historical data statistics.

[0014] Optionally, in step 2, the stain grade assessment criteria are as follows: the stain grade is divided into 0-4 levels, with level 0 corresponding to no stain; level 1 corresponding to light dust, with a stain coverage area of ​​less than 5%; level 2 corresponding to a small amount of stain, with a stain coverage area of ​​5%-20%; level 3 corresponding to moderate stain, with a stain coverage area of ​​20%-50%; and level 4 corresponding to heavy stain, with a stain coverage area of ​​more than 50%.

[0015] Optionally, in step 3, the multi-feature fusion stain recognition algorithm specifically includes:

[0016] Step 31: Color feature extraction. Extract the color difference values ​​ΔR, ΔG, and ΔB in the RGB color space of the image to filter out suspected stain areas; convert to the HSV color space and extract saturation and brightness features to eliminate false stain areas.

[0017] Step 32, texture feature extraction: use gray-level co-occurrence matrix to extract contrast, correlation, energy and entropy feature parameters of suspected stain areas, and use support vector machine model to classify and distinguish different types of stains;

[0018] Step 33, Morphological feature extraction: Calculate the area, perimeter, and roundness parameters of the suspected stain area to determine the stain coverage ratio;

[0019] Step 34: The analysis results of color features, texture features and morphology features are fused and input into the stain level determination model to obtain the stain level. If the determination results of two consecutive images are consistent, the level is confirmed.

[0020] Optionally, the stain grading model can calculate the stain coverage area using the following method:

[0021] The core parameters of color, texture, and morphology features are normalized to obtain feature values ​​within a unified quantization range of [0,1]. Weights are assigned based on the importance of each feature to stain identification, and the stain confidence score of suspected areas is calculated to filter out actual stain areas.

[0022] Only real stain areas with a confidence level greater than the set confidence level threshold are counted, and the percentage of stain coverage area is calculated.

[0023] Optionally, calculate the stain confidence level for the suspected area. The formula is:

[0024] ;

[0025] in It is the normalized value of color features. It is the weight of color features. It is the normalized value of texture features. It is the weight of texture features. It is a normalized value of morphological features. It is the weight of morphological features;

[0026] The formula for calculating the percentage of area covered by the stain, S, is:

[0027] ;

[0028] in It includes all confidence levels greater than a certain confidence threshold. The total number of pixels in the actual stained area; It is the total number of pixels in the effective area of ​​the condom obtained after preprocessing and segmentation, which is obtained by traversing and counting the pixels row by row and column by column in the condom area.

[0029] Optionally, the method for calculating the eigenvalues ​​of the unified quantization interval [0,1] is as follows:

[0030] ;

[0031] ;

[0032] ;

[0033] Where ΔR, ΔG, and ΔB represent the differences in red, green, and blue channel pixel values ​​between the suspected stain area and the condom body area in the RGB color space, respectively. It is the standard deviation of the saturation of the suspected stain area in the HSV color space; It is the preset maximum standard deviation of saturation; It is the entropy value extracted from the suspected stain area based on the gray-level co-occurrence matrix, reflecting the complexity of the area's texture; the larger the entropy value, the more complex the texture. It is the preset maximum entropy value; It is the contrast of the suspected stain area extracted based on the gray-level co-occurrence matrix, reflecting the degree of difference in pixel gray-level values ​​within the area; This is the preset maximum contrast value; It refers to the circularity of the suspected stain area; It is the preset maximum roundness value.

[0034] Optionally, in step 4, the specific cleaning and drying control strategies are as follows:

[0035] If the stain level is 0: Stop the cleaning program, enter the energy-saving drying mode, reduce the hot air temperature by 5-8℃ and the air speed by 20%, and stop drying when the humidity drops to 30%RH.

[0036] If the stain level is 1: maintain the initial cleaning parameters and clean for 1 cycle, then re-analyze. If it is still 1, reduce the drum speed by 10%, the spray pressure by 15%, and extend the cleaning cycle by 1.

[0037] If the stain level is 2: increase the spray pressure by 10% and the drum speed by 15%, and re-analyze after one cleaning cycle; if it is still 2, increase the water temperature by 3-5℃ and continue cleaning for one cycle.

[0038] If the stain level is 3: activate the enhanced cleaning mode, increase the spray pressure by 25%, increase the drum speed by 30%, increase the water temperature by 5-8℃, turn on the cleaning solution circulation filter, and analyze once after each cleaning cycle;

[0039] If the stain level is 4: Determine whether it is a stubborn stain. For non-stubborn stains, start the deep cleaning mode, increase the spray pressure by 40%, increase the drum speed by 40%, increase the water temperature by 8-10℃, and add special cleaning agent; for stubborn stains, additionally control the drum to run in both forward and reverse directions alternately until the stain level drops to 3 or below.

[0040] A condom cleaning and drying control system is used to implement a machine vision-based control method for a condom cleaning and drying system. The condom cleaning and drying control system includes a vision acquisition module, an image processing and analysis module, a main control module, a cleaning execution module, a drying execution module, and a feedback adjustment module.

[0041] The visual acquisition module is electrically connected to the image processing and analysis module, used to acquire condom images and transmit them to the image processing and analysis module; the image processing and analysis module is electrically connected to the main control module, used to analyze images and output stain levels; the main control module is electrically connected to the cleaning execution module and the drying execution module respectively, used to output control commands according to the stain levels; the feedback adjustment module is electrically connected to the main control module, used to acquire operating parameters and feed them back to the main control module.

[0042] In summary, the present invention has at least one of the following beneficial technical effects:

[0043] This invention provides a condom cleaning and drying system and control method based on machine vision. It uses machine vision to collect and analyze condom images in real time, and combines this with a multi-feature fusion algorithm to accurately determine the level of stains, enabling targeted adjustments to the cleaning and drying strategy and effectively improving the cleanliness of the condoms. By dynamically optimizing control parameters, it avoids over-cleaning and drying, significantly reducing energy consumption. Stain detection and control strategy adaptation can be completed without manual intervention, improving the system's automation level and ensuring consistent product quality. Confidence screening ensures accurate stain identification, reducing ineffective control caused by false stain misjudgments, and further improving system control precision and operational stability. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the control method of the condom cleaning and drying system based on machine vision according to the present invention. Detailed Implementation

[0045] The present invention will be further described in detail below with reference to the accompanying drawings.

[0046] This invention discloses a condom cleaning and drying system and control method based on machine vision.

[0047] Reference Figure 1 Example 1, a control method for a condom cleaning and drying system based on machine vision, includes the following steps:

[0048] Step 1: After starting the cleaning program, the vision acquisition module acquires images of the condoms inside the cleaning and drying machine at preset intervals, and preprocesses the acquired raw images.

[0049] Step 2: Based on the preprocessed image, extract color features, texture features and morphological features through a multi-feature fusion stain recognition algorithm, input them into a pre-trained stain level determination model, and analyze the stain level based on the stain coverage area calculation results.

[0050] Step 3: The main control module executes targeted cleaning and drying control strategies based on the determined stain level and the real-time operating parameters collected by the feedback adjustment module.

[0051] Step 4: When the stain level of the condom stabilizes at level 0 and the drying humidity reaches the preset threshold, stop all execution modules, complete the cleaning and drying process, and store key operating data.

[0052] Example 2 also includes step 5, which involves collecting operating parameters in real time and transmitting them to the main control module, and dynamically optimizing the control strategy based on the image analysis results.

[0053] In Example 3, the dynamic optimization control strategy in step 5 specifically includes: if the turbidity of the cleaning fluid exceeds the threshold, replace the cleaning fluid or turn on the high-efficiency filter; if the drying temperature exceeds the upper limit, reduce the hot air temperature and increase the wind speed; if the humidity decreases too slowly, increase the wind speed; and optimize the adjustment parameters and image acquisition interval time corresponding to each stain level through historical data statistics.

[0054] By adopting the above technical solution, real-time perception and level-matching control of stain status are achieved based on machine vision. The vision acquisition module periodically captures images of condoms inside the washing and drying chamber. In the preprocessing stage, image noise is removed, effective areas are segmented, and distortion is corrected to improve image quality and ensure the accuracy of subsequent feature extraction. Using a multi-feature fusion recognition algorithm, stain features are comprehensively captured from three dimensions: color, texture, and shape, overcoming the limitations of single-feature recognition. A stain level determination model transforms feature information into precise stain levels, enabling an objective quantitative assessment of the condom's stain status. The main control module matches the corresponding washing and drying strategy according to the stain level, ensuring precise adaptation of control actions to the actual stain situation and avoiding over- or under-processing. When the stain level is consistently within the acceptable range and the drying humidity meets requirements, the process stops and the data is stored, ensuring product cleanliness while achieving process closure and quality traceability, balancing cleaning effectiveness with basic energy-saving requirements.

[0055] The addition of a dynamic optimization component is based on the core principle of constructing a closed-loop control system encompassing "perception-decision-execution-feedback-optimization." Throughout the cleaning and drying process, system operating parameters are collected in real-time and combined with the analysis results of stain images acquired through machine vision, forming dual data support. Through the dynamic optimization mechanism, the control strategy can be corrected in real time, compensating for the shortcomings of fixed control strategies in adapting to fluctuations in operating conditions. When operating parameters deviate from the preset range or the stain state changes beyond expectations, the control parameters are adjusted promptly, improving the system's adaptability to complex operating conditions, further optimizing cleanliness control accuracy and energy utilization efficiency, and achieving dynamic improvement in control effectiveness.

[0056] Based on threshold control of key operating parameters and iterative optimization using historical data, the system addresses situations where the cleaning fluid turbidity exceeds a threshold. This is achieved by replacing the cleaning fluid or activating the filter to prevent secondary stain adhesion caused by impurities, ensuring stable cleaning results. For abnormal temperature and humidity during the drying stage, hot air temperature and speed are adjusted to balance drying efficiency with condom material protection, preventing high-temperature deformation or incomplete drying. Statistical analysis of historical operating data identifies the optimal control parameter range for different stain levels. Continuous iterative optimization of adjustment parameters and image acquisition intervals gradually adapts the control strategy to the processing needs of condoms of different materials and stain types, achieving long-term improvements in system control accuracy and energy efficiency.

[0057] In Example 4, step 2, the stain grade evaluation criteria are as follows: the stain grade is divided into 0-4 levels, with level 0 corresponding to no stain; level 1 corresponding to light dust, with a stain coverage area of ​​less than 5%; level 2 corresponding to a small amount of stain, with a stain coverage area of ​​5%-20%; level 3 corresponding to moderate stain, with a stain coverage area of ​​20%-50%; and level 4 corresponding to heavy stain, with a stain coverage area of ​​more than 50%.

[0058] By adopting the above technical solution, a unified and traceable stain level assessment benchmark is constructed. By clearly defining stain levels from 0 to 4 and matching precise stain coverage area thresholds for each level, the previously ambiguous stain status is transformed into a quantifiable and comparable objective indicator, eliminating discrepancies in level determination caused by traditional subjective judgments or vague standards. This grading standard directly adapts to the stain coverage area calculation results output by the multi-feature fusion recognition algorithm mentioned earlier, providing clear data support for stain level determination and ensuring consistency in stain level assessment across different working conditions and batches of condoms. Simultaneously, the clear grading standard provides a foundation for the precise matching of subsequent targeted control strategies, making the cleaning and drying efforts corresponding to different stain levels more targeted, avoiding insufficient or excessive control due to ambiguous level definitions, and further ensuring a balance between cleanliness control accuracy and energy utilization efficiency.

[0059] Example 5, in step 3, the multi-feature fusion stain recognition algorithm specifically includes:

[0060] Step 31: Color feature extraction. Extract the color difference values ​​ΔR, ΔG, and ΔB in the RGB color space of the image to filter out suspected stain areas; convert to the HSV color space and extract saturation and brightness features to eliminate false stain areas.

[0061] Step 32, texture feature extraction: use gray-level co-occurrence matrix to extract contrast, correlation, energy and entropy feature parameters of suspected stain areas, and use support vector machine model to classify and distinguish different types of stains;

[0062] Step 33, Morphological feature extraction: Calculate the area, perimeter, and roundness parameters of the suspected stain area to determine the stain coverage ratio;

[0063] Step 34: The analysis results of color features, texture features and morphology features are fused and input into the stain level determination model to obtain the stain level. If the determination results of two consecutive images are consistent, the level is confirmed.

[0064] By employing the above technical solution, a comprehensive stain feature description system is constructed by extracting three types of features—color, texture, and morphology—step by step, overcoming the limitations of single-feature recognition. In the color feature extraction stage, suspected stain areas are first screened using RGB color space differences, and then false stains such as reflections are eliminated using saturation and brightness features of the HSV color space, achieving preliminary and accurate identification of suspected areas. In the texture feature extraction stage, the gray-level co-occurrence matrix is ​​used to mine subtle texture information in the stain area, combined with a support vector machine model to effectively distinguish different types of stains, providing a basis for subsequent targeted cleaning strategies. In the morphology feature extraction stage, parameters such as the area, perimeter, and roundness of the stain area are calculated to quantify the spatial shape and coverage ratio of the stain, providing key quantitative indicators for stain level determination. Finally, the analysis results of the three types of features are integrated, fully consolidating the advantages of each feature, and inputting them into the stain level determination model to achieve accurate level output. Simultaneously, a confirmation mechanism ensuring consistency between the determination results of two consecutive image acquisitions avoids misjudgments caused by interference from a single image acquisition, ensuring the stability and reliability of stain level determination.

[0065] Example 6: The method for calculating the stain coverage area using the stain grading model is as follows:

[0066] The core parameters of color, texture, and morphology features are normalized to obtain feature values ​​within a unified quantization range of [0,1]. Weights are assigned based on the importance of each feature to stain identification, and the stain confidence score of suspected areas is calculated to filter out actual stain areas.

[0067] Only real stain areas with a confidence level greater than the set confidence level threshold are counted, and the percentage of stain coverage area is calculated.

[0068] Example 7: Calculating the confidence level of stains in suspected areas The formula is:

[0069] ;

[0070] in It is the normalized value of color features. It is the weight of color features. It is the normalized value of texture features. It is the weight of texture features. It is a normalized value of morphological features. It is the weight of morphological features;

[0071] The formula for calculating the percentage of area covered by the stain, S, is:

[0072] ;

[0073] in It includes all confidence levels greater than a certain confidence threshold. The total number of pixels in the actual stained area; It is the total number of pixels in the effective area of ​​the condom obtained after preprocessing and segmentation, which is obtained by traversing and counting the pixels row by row and column by column in the condom area.

[0074] Example 8: The method for calculating the eigenvalues ​​of the unified quantization interval [0,1] is as follows:

[0075] ;

[0076] ;

[0077] ;

[0078] Where ΔR, ΔG, and ΔB represent the differences in red, green, and blue channel pixel values ​​between the suspected stain area and the condom body area in the RGB color space, respectively. It is the standard deviation of the saturation of the suspected stain area in the HSV color space; It is the preset maximum standard deviation of saturation; It is the entropy value extracted from the suspected stain area based on the gray-level co-occurrence matrix, reflecting the complexity of the area's texture; the larger the entropy value, the more complex the texture. It is the preset maximum entropy value; It is the contrast of the suspected stain area extracted based on the gray-level co-occurrence matrix, reflecting the degree of difference in pixel gray-level values ​​within the area; This is the preset maximum contrast value; It refers to the circularity of the suspected stain area; It is the preset maximum roundness value.

[0079] By employing the aforementioned technical solution, the core parameters of the three major features—color, texture, and morphology—are normalized to eliminate dimensional differences between different feature parameters, unifying them into the same quantification range and laying the foundation for subsequent multi-feature fusion calculations. Weights are assigned according to the varying importance of different features in stain determination, making the fusion calculation results more closely aligned with the actual needs of stain identification. The stain confidence score of suspected areas is obtained through weighted fusion, serving as the basis for screening genuine stain areas. This effectively eliminates false stain interference, ensuring that all areas included in the area calculation are genuine stains. Only genuine stain areas with confidence scores meeting the threshold requirements are statistically analyzed for area, ultimately yielding a precise stain coverage percentage, providing core quantitative support for accurate stain level determination.

[0080] The system employs mathematical formulas to achieve feature fusion quantification and precise conversion of coverage area. The confidence score calculation formula uses a linear weighted summation model, multiplying the normalized three types of feature values ​​by their corresponding weights and then summing the results. This intuitively quantifies the comprehensive contribution of different features to stain determination, and the resulting confidence score can be directly used to distinguish between real and fake stains, ensuring the objectivity and operability of stain screening. The stain coverage area ratio formula converts the statistical results at the image pixel level into an intuitive area ratio indicator by comparing the total number of pixels in the real stain area to the total number of pixels in the effective area of ​​the condom. This indicator is directly compatible with stain level assessment standards, enabling a quantitative evaluation of stain coverage and providing clear data for stain level determination. Furthermore, the formula's simple structure facilitates rapid system calculation and real-time response.

[0081] Dedicated normalization algorithms were designed for different feature parameters to achieve standardization and uniformity. For color features, a weighted summation method was used to calculate the normalized value by combining the channel difference values ​​of the RGB color space and the saturation standard deviation of the HSV color space. This method retains the core discriminative information of color differences while further enhancing the distinguishability of color features through saturation characteristics. For texture features, entropy and contrast were selected as two key texture parameters, and their contributions to stain recognition were weighted to calculate the normalized value, accurately quantifying the texture difference between the stain area and the body. For morphological features, roundness was used as the core parameter, and the normalized value was obtained through reverse calculation, highlighting the characteristics of irregularly shaped stains and eliminating the interference of pseudo-stains with regular shapes. The dedicated normalization algorithms for the three types of features all ensure that the output values ​​are within a unified quantization range, providing high-quality feature data support for the subsequent accurate calculation of confidence.

[0082] In Example 9, step 4, the specific targeted cleaning and drying control strategy is as follows:

[0083] If the stain level is 0: Stop the cleaning program, enter the energy-saving drying mode, reduce the hot air temperature by 5-8℃ and the air speed by 20%, and stop drying when the humidity drops to 30%RH.

[0084] If the stain level is 1: maintain the initial cleaning parameters and clean for 1 cycle, then re-analyze. If it is still 1, reduce the drum speed by 10%, the spray pressure by 15%, and extend the cleaning cycle by 1.

[0085] If the stain level is 2: increase the spray pressure by 10% and the drum speed by 15%, and re-analyze after one cleaning cycle; if it is still 2, increase the water temperature by 3-5℃ and continue cleaning for one cycle.

[0086] If the stain level is 3: activate the enhanced cleaning mode, increase the spray pressure by 25%, increase the drum speed by 30%, increase the water temperature by 5-8℃, turn on the cleaning solution circulation filter, and analyze once after each cleaning cycle;

[0087] If the stain level is 4: Determine whether it is a stubborn stain. For non-stubborn stains, start the deep cleaning mode, increase the spray pressure by 40%, increase the drum speed by 40%, increase the water temperature by 8-10℃, and add special cleaning agent; for stubborn stains, additionally control the drum to run in both forward and reverse directions alternately until the stain level drops to 3 or below.

[0088] By adopting the above technical solution, a balance between cleaning and drying effects and energy-saving requirements is achieved through graded adaptation and dynamic feedback adjustment based on stain levels. Differentiated control intensities are matched to different stain levels to establish a precise correspondence between stain severity and control intensity: Level 0 (no stains) uses an energy-saving drying mode to avoid excessive energy consumption and material damage; Level 1 (light stains) prioritizes maintaining initial parameters and dynamically observes them, slightly lowering parameters as needed to save energy; Levels 2-3 (light to moderate stains) gradually increase spray pressure, drum speed, and water temperature, activating the filtration device as needed to enhance cleaning power while avoiding resource waste; Level 4 (heavy stains) distinguishes between stubborn and non-stubborn types, specifically activating a deep cleaning mode, increasing mechanical friction as needed to ensure effective removal of stubborn stains. Each level strategy incorporates a periodic image re-analysis mechanism to track stain changes in real time and dynamically adjust control actions, ensuring cleaning cleanliness under different stain levels while minimizing energy and consumable consumption, thus improving the targeting and economy of the control strategy.

[0089] Example 10: Condom cleaning and drying control system, used to implement a machine vision-based control method for a condom cleaning and drying system. The condom cleaning and drying control system includes a vision acquisition module, an image processing and analysis module, a main control module, a cleaning execution module, a drying execution module, and a feedback adjustment module.

[0090] The visual acquisition module is electrically connected to the image processing and analysis module, used to acquire condom images and transmit them to the image processing and analysis module; the image processing and analysis module is electrically connected to the main control module, used to analyze images and output stain levels; the main control module is electrically connected to the cleaning execution module and the drying execution module respectively, used to output control commands according to the stain levels; the feedback adjustment module is electrically connected to the main control module, used to acquire operating parameters and feed them back to the main control module.

[0091] The implementation principle of the present invention is illustrated below through specific embodiments:

[0092] This invention relates to the mass production cleaning and drying process for natural latex condoms, aiming to achieve automated and precise control based on machine vision, while balancing cleaning cleanliness with energy efficiency. The condom cleaning and drying system and control process strictly adhere to the aforementioned technical solution, as detailed below.

[0093] System composition and selection:

[0094] The condom cleaning and drying control system consists of a vision acquisition module, an image processing and analysis module, a main control module, a cleaning execution module, a drying execution module, and a feedback adjustment module. The specific selection and configuration of each module are as follows:

[0095] The visual acquisition module uses a high-definition industrial camera, paired with a ring-shaped LED uniform light source. The light source power is 20W, the color temperature is 5500K, the camera frame rate is set to 30fps, and the resolution is 2592×1944. The anti-fog and dustproof protective cover has a built-in 5W miniature hot air device with a hot air temperature set to 40℃ and a continuous airflow rate of 0.5m / s to ensure that the acquired images are free from water mist interference.

[0096] The image processing and analysis module uses the NX edge computing terminal, which integrates a GPU acceleration unit and has a computing speed of up to 21 TOPS. It is used to process image data in real time and run multi-feature fusion recognition algorithms.

[0097] The main control module uses an STM32H743ZI embedded microprocessor and establishes communication with each module via a CAN bus at a rate of 500kbps to ensure rapid transmission and response of control commands.

[0098] The cleaning execution module includes an adjustable pressure spray pump with a maximum pressure of 0.8MPa; the drum is driven by a variable frequency motor with a speed adjustment range of 10-60r / min; the water temperature adjustment unit uses a stainless steel heating tube with a heating power of 3kW and a water temperature adjustment range of 20-60℃.

[0099] The drying execution module adopts a centrifugal hot air circulation device with a fan power of 1.5kW and a wind speed adjustment range of 1-5m / s; it is equipped with a PT100 temperature sensor with a temperature measurement range of 0-100℃; the humidity sensor is a capacitive humidity sensor with a measurement range of 0-100%RH and an accuracy of ±2%RH.

[0100] The feedback adjustment module includes a turbidity sensor with a measurement range of 0-100 NTU and an accuracy of ±1 NTU; a drum speed sensor with an accuracy of ±1 r / min; and a spray pressure sensor with a measurement range of 0-1 MPa and an accuracy of ±0.01 MPa, which collects various operating parameters in real time.

[0101] Specific control process:

[0102] This embodiment describes the cleaning and drying process for a batch of 1000 natural latex condoms. The specific control procedure is as follows:

[0103] Step 1, System preprocessing and parameter initialization:

[0104] The system preprocessing program is initiated. The visual acquisition module completes self-calibration using a standard grayscale plate. White balance parameters are set to R255, G255, and B255, exposure is adjusted to 1 / 500s, and color reproduction error is calibrated to within ±2%. The cleaning and drying machine is then cleaned under no-load conditions. The drum rotates forward for 30 seconds, followed by a 10-second water spray to remove residual impurities. The operating status of each module is checked to confirm that the spray system, drum, and hot air device are functioning correctly.

[0105] Initialize the parameters as follows: initial cleaning parameters: drum speed 30 r / min, spray pressure 0.3 MPa, water temperature 35℃. Initial drying parameters: hot air temperature 50℃, air velocity 3 m / s. Initial image acquisition interval is set to 30 seconds / time. The stain level assessment criteria are set as follows: Level 0: no stains; Level 1: slight dust, stain coverage area less than 5%; Level 2: small amount of stains, stain coverage area 5%-20%; Level 3: moderate stains, stain coverage area 20%-50%; Level 4: heavy stains, stain coverage area greater than 50%.

[0106] Step 2, Image Acquisition and Preprocessing:

[0107] The cleaning program is initiated, and the visual acquisition module captures images of the condoms inside the cleaning and drying machine at 30-second intervals. During the acquisition process, the ring LED light source maintains a constant brightness, and the miniature hot air device in the anti-fog and dustproof protective cover operates continuously.

[0108] The original image is preprocessed as follows: a 3×3 core Gaussian filter algorithm is used to remove noise; an adaptive threshold segmentation algorithm is used to separate the condom area from the background, with the threshold range set to 120-200; geometric correction is performed on the segmented image, with the correction error controlled within 1 pixel; and a histogram equalization algorithm is used to enhance the image contrast, making the difference between the stained area and the condom body more significant.

[0109] Step 3, Multi-feature fusion recognition and stain level determination:

[0110] Color feature extraction: Calculate the RGB color difference values ​​ΔR, ΔG, and ΔB between the suspected stain area and the condom body. Set the color difference threshold to 30 and filter out areas where any one of ΔR, ΔG, or ΔB is greater than 30 as suspected stain areas. Convert to HSV color space and extract the saturation standard deviation Sstd of the suspected area. Set Smax to 50 and exclude false stain areas where Sstd is less than 10.

[0111] Texture feature extraction: Gray-level co-occurrence matrix is ​​used to extract feature parameters of suspected stain areas. The distance of the gray-level co-occurrence matrix is ​​set to 2 pixels, and the orientation is set to 0°, 45°, 90°, and 135°. Four parameters are extracted: contrast, correlation, energy, and entropy. The model is classified by support vector machine, and the radial basis function is selected as the kernel function to distinguish between two common stains: dust and oil.

[0112] Morphological feature extraction: Calculate the area, perimeter, and circularity of the suspected stain area. The formula for calculating circularity is Cs=4πA / P. 2 The default value for Cmax is 0.9.

[0113] Feature fusion and level determination: The three major feature parameters are normalized. In the calculation of the color feature normalization value Cnorm, the weight allocation is 0.5 for RGB difference and 0.5 for saturation standard deviation; in the calculation of the texture feature normalization value Tnorm, the entropy value accounts for 0.6 and the contrast value accounts for 0.4; the morphological feature normalization value Mnorm is 1 - Cs / Cmax. Color feature weights are set. =0.4, texture feature weight =0.3, morphological feature weight =0.3, the stain confidence threshold is 0.6. (Using the formula...) ; Calculate the confidence level of suspected regions, only statistically analyzing Areas with a value ≥0.6 are considered actual stain areas. The percentage of the stain-covered area is calculated using the formula S=ΣNs,i / Nn×100%, where Nn is the total number of pixels in the effective area of ​​the condom, which is 4,838,400 after traversal counting. The S value is obtained by calculating ΣNs,i, and then combined with the stain level assessment criteria to output the level. If the levels of two consecutive images are consistent, the level is confirmed; otherwise, the acquisition interval is shortened to 10 seconds / image and re-analyzed.

[0114] Step 4: Implement targeted cleaning and drying control strategies.

[0115] If the stain level is determined to be 0, the cleaning program will be stopped immediately and the energy-saving drying mode will be entered. The hot air temperature will be reduced to 42℃ and the air speed will be reduced to 2.4m / s. When the humidity sensor detects that the humidity inside the cavity has dropped to 30%RH, the drying will be stopped.

[0116] If the stain level is determined to be Level 1, maintain the initial cleaning parameters and clean for 30 seconds before re-analyzing; if it is still Level 1, reduce the drum speed to 27 r / min, the spray pressure to 0.255 MPa, and extend the cleaning time by 30 seconds.

[0117] If the stain level is determined to be level 2, keep the water temperature at 35℃, increase the spray pressure to 0.33MPa, increase the drum speed to 34.5r / min, and re-analyze after 30 seconds of cleaning; if it is still level 2, increase the water temperature to 38℃ and continue cleaning for 30 seconds.

[0118] If the stain level is determined to be level 3, the enhanced cleaning mode is activated, the spray pressure is increased to 0.375MPa, the drum speed is increased to 39r / min, the water temperature is increased to 40℃, and the cleaning fluid circulation filtration device is turned on; the image is analyzed every 30 seconds of cleaning until the level drops to level 2 or below.

[0119] If the stain level is determined to be level 4, stubborn stains are identified by texture feature entropy value. The preset entropy threshold is 7; an entropy value higher than 7 indicates a stubborn stain. For non-stubborn stains, a deep cleaning mode is activated, increasing the spray pressure to 0.42 MPa, the drum speed to 42 r / min, and the water temperature to 43℃. A special neutral cleaning agent is added, and the stain is cleaned for 60 seconds before re-analysis. For stubborn stains, the drum is additionally rotated clockwise for 30 seconds and counterclockwise for 30 seconds alternately until the stain level drops to level 3 or below.

[0120] Step 5, Real-time Feedback and Dynamic Optimization:

[0121] The feedback adjustment module collects various operating parameters in real time and transmits them to the main control module. If the turbidity of the cleaning solution exceeds 20 NTU, the cleaning solution is replaced immediately; if the temperature during the drying stage exceeds 55℃, the hot air temperature is reduced to 45℃ and the air velocity is increased to 3.5 m / s; if the humidity decrease rate is less than 0.5%RH / second, the air velocity is increased to 3.5 m / s. Based on historical data, if five consecutive batches of natural latex condoms show Grade 1 stains and a 15% reduction in spray pressure still maintains cleanliness, the spray pressure adjustment threshold for Grade 1 stains on this material is updated to a 15% reduction.

[0122] Step 6, Program End and Data Storage:

[0123] Once the condom stain level remains consistently at level 0 and the drying humidity reaches 30% RH, the system stops all execution modules. Key data from this cleaning and drying process is automatically stored, including initial parameters, image analysis results at each stage, stain level change curves, control strategy adjustment records, and energy consumption data, providing support for subsequent system optimization and quality traceability.

[0124] This embodiment, through the above configuration and process, can achieve precise cleaning and drying of natural latex condoms, with a cleaning cleanliness compliance rate of ≥99%; compared with the traditional fixed parameter control mode, energy consumption is reduced by more than 18%, with a high degree of automation and no need for manual intervention.

[0125] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A control method for a condom cleaning and drying system based on machine vision, characterized in that, Includes the following steps: Step 1: After starting the cleaning program, the vision acquisition module acquires images of the condoms inside the cleaning and drying machine at preset intervals, and preprocesses the acquired raw images. Step 2: Based on the preprocessed image, extract color features, texture features and morphological features through a multi-feature fusion stain recognition algorithm, input them into a pre-trained stain level determination model, and analyze the stain level based on the stain coverage area calculation results. Step 3: The main control module executes targeted cleaning and drying control strategies based on the determined stain level and the real-time operating parameters collected by the feedback adjustment module. Step 4: When the stain level of the condom stabilizes at level 0 and the drying humidity reaches the preset threshold, stop all execution modules, complete the cleaning and drying process, and store key operating data.

2. The control method for a condom cleaning and drying system based on machine vision according to claim 1, characterized in that, It also includes step 5, which involves collecting operating parameters in real time and transmitting them to the main control module, and dynamically optimizing the control strategy based on the image analysis results.

3. The control method for a condom cleaning and drying system based on machine vision according to claim 2, characterized in that, The dynamic optimization control strategy in step 5 specifically includes: if the turbidity of the cleaning fluid exceeds the threshold, replace the cleaning fluid or turn on the high-efficiency filter; if the drying temperature exceeds the upper limit, reduce the hot air temperature and increase the wind speed; if the humidity decreases too slowly, increase the wind speed; and optimize the adjustment parameters and image acquisition interval time corresponding to each stain level through historical data statistics.

4. The control method for a condom cleaning and drying system based on machine vision according to claim 3, characterized in that, In step 2, the specific criteria for judging stain levels are as follows: stain levels are divided into 0-4 levels, with level 0 corresponding to no stain; level 1 corresponding to light dust, with stain coverage area less than 5%; level 2 corresponding to a small amount of stain, with stain coverage area 5%-20%; level 3 corresponding to moderate stain, with stain coverage area 20%-50%; and level 4 corresponding to heavy stain, with stain coverage area greater than 50%.

5. The control method for a condom cleaning and drying system based on machine vision according to claim 4, characterized in that, In step 3, the multi-feature fusion stain recognition algorithm specifically includes: Step 31: Color feature extraction. Extract the color difference values ​​ΔR, ΔG, and ΔB in the RGB color space of the image to filter out suspected stain areas; convert to the HSV color space and extract saturation and brightness features to eliminate false stain areas. Step 32, texture feature extraction: use gray-level co-occurrence matrix to extract contrast, correlation, energy and entropy feature parameters of suspected stain areas, and use support vector machine model to classify and distinguish different types of stains; Step 33, Morphological feature extraction: Calculate the area, perimeter, and roundness parameters of the suspected stain area to determine the stain coverage ratio; Step 34: The analysis results of color features, texture features and morphology features are fused and input into the stain level determination model to obtain the stain level. If the determination results of two consecutive images are consistent, the level is confirmed.

6. The control method for a condom cleaning and drying system based on machine vision according to claim 5, characterized in that, The method for calculating the stain coverage area using the stain grading model is as follows: The core parameters of color, texture, and morphology features are normalized to obtain feature values ​​within a unified quantization range of [0,1]. Weights are assigned based on the importance of each feature to stain identification, and the stain confidence score of suspected areas is calculated to filter out actual stain areas. Only real stain areas with a confidence level greater than the set confidence level threshold are counted, and the percentage of stain coverage area is calculated.

7. The control method for a condom cleaning and drying system based on machine vision according to claim 6, characterized in that, Calculate the confidence level of stains in suspected areas The formula is: ； in It is the normalized value of color features. It is the weight of color features. It is the normalized value of texture features. It is the weight of texture features. It is a normalized value of morphological features. It is the weight of morphological features; The formula for calculating the percentage of area covered by the stain, S, is: ; in It includes all confidence levels greater than a certain confidence threshold. The total number of pixels in the actual stained area; It is the total number of pixels in the effective area of ​​the condom obtained after preprocessing and segmentation, which is obtained by traversing and counting the pixels row by row and column by column in the condom area.

8. The control method for a condom cleaning and drying system based on machine vision according to claim 7, characterized in that, The method for calculating the eigenvalues ​​of the unified quantization interval [0,1] is as follows: ; ; ; Where ΔR, ΔG, and ΔB represent the differences in red, green, and blue channel pixel values ​​between the suspected stain area and the condom body area in the RGB color space, respectively. It is the standard deviation of the saturation of the suspected stain area in the HSV color space; It is the preset maximum standard deviation of saturation; It is the entropy value extracted from the suspected stain area based on the gray-level co-occurrence matrix, reflecting the complexity of the area's texture; the larger the entropy value, the more complex the texture. It is the preset maximum entropy value; It is the contrast of the suspected stain area extracted based on the gray-level co-occurrence matrix, reflecting the degree of difference in pixel gray-level values ​​within the area; This is the preset maximum contrast value; It refers to the circularity of the suspected stain area; It is the preset maximum roundness value.

9. The control method for a condom cleaning and drying system based on machine vision according to claim 8, characterized in that, In step 4, the specific cleaning and drying control strategies are as follows: If the stain level is 0: stop the cleaning program, enter the energy-saving drying mode, reduce the hot air temperature by 5-8℃ and the air speed by 20%, and stop drying when the humidity drops to 30%RH. If the stain level is 1: maintain the initial cleaning parameters and clean for 1 cycle, then re-analyze. If it is still 1, reduce the drum speed by 10%, the spray pressure by 15%, and extend the cleaning cycle by 1. If the stain level is 2: increase the spray pressure by 10% and the drum speed by 15%, and re-analyze after one cleaning cycle; if it is still 2, increase the water temperature by 3-5℃ and continue cleaning for one cycle. If the stain level is 3: activate the enhanced cleaning mode, increase the spray pressure by 25%, increase the drum speed by 30%, increase the water temperature by 5-8℃, turn on the cleaning solution circulation filter, and analyze once after each cleaning cycle; If the stain level is 4: Determine whether it is a stubborn stain. For non-stubborn stains, start the deep cleaning mode, increase the spray pressure by 40%, increase the drum speed by 40%, increase the water temperature by 8-10℃, and add special cleaning agent; for stubborn stains, additionally control the drum to run in both forward and reverse directions alternately until the stain level drops to 3 or below.

10. A condom cleaning and drying control system, characterized in that, To implement the machine vision-based condom cleaning and drying system control method of claim 9, the condom cleaning and drying control system includes a vision acquisition module, an image processing and analysis module, a main control module, a cleaning execution module, a drying execution module, and a feedback adjustment module; The visual acquisition module is electrically connected to the image processing and analysis module, used to acquire condom images and transmit them to the image processing and analysis module; the image processing and analysis module is electrically connected to the main control module, used to analyze images and output stain levels; the main control module is electrically connected to the cleaning execution module and the drying execution module respectively, used to output control commands according to the stain levels; the feedback adjustment module is electrically connected to the main control module, used to acquire operating parameters and feed them back to the main control module.

Images