Tourniquet disinfection parameter adaptive control method based on visual identification and deep learning

Abstract

The invention relates to a tourniquet disinfection parameter adaptive control method based on visual identification and deep learning, and the method comprises the steps: scanning the volume distribution characteristics of a tourniquet in a net frame through a visual identification module, calculating the volume fraction parameter of the tourniquet through a deep residual network, and obtaining the volume and volume fraction of the tourniquet in the tourniquet net frame; the liquid level information of the liquid required for disinfection is obtained; generating an optimal disinfectant proportioning scheme, stirring rotating speed, soaking time, stirring time, clear water washing time and drying time, and executing proportioning injection through a disinfectant feeding pump; the tourniquet net frame is upwards pulled up from the disinfection container through the electric cylinder; and sequentially washing the tourniquet with clean water and drying the tourniquet with hot air on the basis of the obtained clean water washing time and drying time. Compared with the prior art, through accurate volume measurement and intelligent parameter optimization, the intelligent and self-adaptive parameter control of the whole tourniquet disinfection process is realized, and the reliability and stability of the disinfection effect and the resource utilization efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical tourniquet disinfection, and in particular to a method for adaptively controlling tourniquet disinfection parameters based on visual recognition and deep learning. Background Art

[0002] The traditional tourniquet disinfection process generally has the problem of relying on manual experience and judgment, and the problem of fixed disinfection parameters. Existing disinfection equipment mostly uses preset programs to control parameters such as disinfectant dosage and immersion time, and cannot be dynamically adjusted according to the actual load, which can easily lead to waste of disinfectant or insufficient effective concentration. In terms of tourniquet volume estimation, conventional methods mostly use direct manual experience operations, which are difficult to accurately reflect the actual accumulation volume and spatial distribution characteristics of the tourniquet in the mesh frame, affecting the accuracy of liquid level calculations. In addition, existing equipment lacks the ability to intelligently analyze variables such as material properties and disinfectant penetration efficiency, and the stirring intensity and drying parameter settings lack data support, which can easily lead to incomplete disinfection or equipment damage.

[0003] While some recent studies have attempted to incorporate image recognition technology, these techniques are often limited to two-dimensional image processing, resulting in insufficient accuracy in calculating volume fractions of three-dimensional stacking states. Existing neural network models struggle with optimizing complex disinfection parameters, including insufficient feature extraction and difficulty modeling multivariate coupling relationships. This makes it difficult to achieve coordinated optimization of parameters such as immersion time and stirring speed. Particularly during the tourniquet leaching stage, traditional processes lack intelligent leaching duration control, which can easily lead to disinfectant residue and inefficiency.

[0004] Existing technologies, such as the tourniquet disinfection device provided by patent CN119548654A, achieve a certain degree of automation in the disinfection, rinsing, and drying processes. However, these disinfection parameter controls rely on preset programs and cannot be adjusted in real time based on the actual tourniquet load, material properties, and other factors. This leads to problems such as wasted disinfectant or insufficient effective concentration. The fully automatic tourniquet disinfection device described in patent CN209092253U, which integrates rinsing, disinfection, rinsing, and drying, also suffers from the problem of fixed disinfection parameters and inaccurate tourniquet volume estimation, making it difficult to accurately reflect the actual accumulated volume and spatial distribution characteristics of the tourniquet within the disinfection chamber, thus affecting disinfection effectiveness and efficiency. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned existing technologies and provide a tourniquet disinfection parameter adaptive control method based on visual recognition and deep learning. Through precise volume measurement and intelligent parameter optimization, intelligent and adaptive parameter control of the entire tourniquet disinfection process is realized, thereby improving the reliability, stability and resource utilization efficiency of the disinfection effect.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] The present invention provides a method for adaptively controlling tourniquet disinfection parameters based on visual recognition and deep learning, comprising the following steps:

[0008] S1. Scan the volume distribution characteristics of the tourniquet in the frame through the visual recognition module, calculate the tourniquet volume fraction parameters using the deep residual network, and obtain the volume and volume fraction of the tourniquet in the tourniquet frame;

[0009] S2. obtaining liquid level information of the liquid required for disinfection based on the volume and volume fraction occupied by the tourniquet;

[0010] S3. Based on the liquid level information, a convolutional neural network is used to generate the optimal disinfectant ratio, stirring speed, soaking time, stirring time, water rinsing time, and drying time. The disinfectant dosing pump is then used to perform the ratio injection.

[0011] S4. When the operation time reaches the sum of the soaking time and the stirring time, the tourniquet frame is pulled upward from the disinfection container by the electric cylinder to separate the tourniquet frame from the disinfectant;

[0012] S5. Based on the water flushing time and drying time obtained in S3, the tourniquet is rinsed with water and dried with hot air in sequence.

[0013] Furthermore, in S1, obtaining the volume of the tourniquet in the tourniquet frame specifically includes the following process:

[0014] Multiple visual recognition modules are used to scan the tourniquet in the frame at multiple angles to obtain multiple tourniquet images from different perspectives, and these images are pre-processed to improve image quality.

[0015] The preprocessed image is input into a pre-trained deep residual network. The volume distribution features of the tourniquet are extracted through the convolutional layer, pooling layer, and fully connected layer of the deep residual network. The pre-trained deep residual network is trained based on existing sample data labeled with tourniquet volume information. The mapping relationship between the tourniquet volume and image features is learned through training.

[0016] The volume occupied by the tourniquet in the mesh frame is calculated based on the extracted volume distribution characteristics.

[0017] Furthermore, in S1, obtaining the volume fraction of the tourniquet in the tourniquet frame specifically includes the following process:

[0018] The volume fraction of the tourniquet is determined based on the ratio of the volume occupied by the tourniquet in the mesh frame to the known total volume of the mesh frame.

[0019] Furthermore, in S2, obtaining the liquid level information of the liquid required for disinfection based on the volume and volume fraction occupied by the tourniquet specifically includes the following process:

[0020] Estimating the liquid level: Dividing the total volume of the tourniquet obtained by the effective cross-sectional area of ​​the sterilization container to obtain a preliminary liquid level. The sterilization container is a right cylindrical or prismatic structure.

[0021] Determine the liquid level adjustment coefficient: select the corresponding coefficient from the pre-established liquid level adjustment coefficient database based on the material properties of the tourniquet and the properties of the disinfectant;

[0022] Calculate the calibrated liquid level: multiply the initial liquid level by the liquid level adjustment coefficient to obtain the calibrated liquid level. The liquid level is the liquid level information of the liquid required for disinfection.

[0023] Furthermore, in S3, the convolutional neural network is used to generate the optimal disinfectant ratio scheme, which specifically includes the following process:

[0024] Training a convolutional neural network model: collecting disinfection experimental data, including the volume and volume fraction of the tourniquet within the tourniquet frame, disinfectants of different concentrations and compositions, and corresponding disinfection effect evaluation data, to construct training samples;

[0025] The training samples are input into the convolutional neural network. By adjusting the parameters of the network's convolutional layer, pooling layer, and fully connected layer, the network learns the mapping relationship between the tourniquet characteristics and the disinfectant ratio scheme, thus obtaining a trained convolutional neural network model.

[0026] The volume and volume fraction of the tourniquet obtained by the visual recognition module and the disinfectant with preset composition and concentration are used as inputs to the convolutional neural network model, and the corresponding optimal disinfectant ratio scheme is output.

[0027] Furthermore, in S3, determining the stirring speed, soaking time, stirring time, and drying time specifically includes the following process:

[0028] Obtaining historical disinfection data, the historical disinfection data including the volume and volume fraction of the tourniquet in the tourniquet frame under different disinfectant ratios, the corresponding reasonable stirring speed range, the optimal soaking time, the stirring time, the clean water rinsing time, and the drying time;

[0029] Use convolutional neural networks to mine and analyze historical disinfection data and establish a multi-parameter prediction model of time and speed;

[0030] The time and speed multi-parameter prediction model is used to predict the appropriate stirring speed, soaking time, stirring time, water rinsing time, and drying time according to the current tourniquet volume and volume fraction and the selected disinfectant ratio.

[0031] Furthermore, the optimal soaking time, stirring time, water rinsing time, and drying time can be set to minimum values, such as the minimum soaking time being 30 minutes.

[0032] Furthermore, after S3, the following processes are also included:

[0033] The liquid level of the disinfectant is monitored in real time by a liquid level sensor, and the monitored liquid level is compared with the liquid level information obtained in S2. If the deviation exceeds the preset threshold, an alarm is issued and subsequent disinfection operations are suspended.

[0034] Furthermore, after S4, the following process is also included:

[0035] After the electric cylinder pulls the tourniquet frame upward from the disinfection container, the tourniquet frame is separated from the disinfectant and drained for 1 to 10 minutes.

[0036] Furthermore, S5 specifically includes the following processes:

[0037] Clean the tourniquet inside the mesh frame by spraying clean water;

[0038] Obtain the material of the tourniquet based on the visual recognition module;

[0039] Based on the material of the tourniquet and its state after disinfection, the corresponding drying temperature and wind speed are selected from a pre-established drying parameter database to perform hot air drying on the tourniquet. The database is constructed based on experimental data of tourniquets of different materials under different drying conditions.

[0040] Furthermore, the visual recognition module includes multiple industrial cameras and a microprocessor connected to each industrial camera.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1) This invention achieves intelligent parameter control throughout the entire tourniquet disinfection process through the collaborative innovation of visual recognition and deep learning technologies. Based on an algorithm fused with multi-angle stereo scanning and a deep residual network, it breaks through the dimensional limitations of traditional volume estimation, accurately analyzing the spatial distribution characteristics and volume fraction of the tourniquet, providing high-precision data support for disinfectant volume calculation. Combined with a multivariate decision model constructed using a convolutional neural network, it can adaptively optimize disinfectant ratios and dynamically match stirring and drying parameters, significantly improving disinfectant penetration uniformity and equipment operating efficiency. At the same time, through real-time liquid level monitoring and drainage status feedback mechanisms, it effectively avoids resource waste and operational risks.

[0043] 2) This method innovatively establishes a correlation system between material properties and disinfection processes, achieving full-link adaptive control from liquid level calibration, ratio optimization to drying regulation. It not only overcomes the fluctuations in disinfection quality caused by reliance on manual experience, but also dynamically adjusts the operation sequence according to the accumulation state of the tourniquet, ensuring the sterilization effect while reducing the risk of device damage. The entire solution takes into account the accuracy, reliability and energy efficiency optimization of the disinfection process through data-driven closed-loop control logic, providing an innovative technical path for the intelligent disinfection of medical devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning in the present invention. DETAILED DESCRIPTION

[0045] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. Component models, material names, connection structures, circuit structures, control methods, algorithms, and other features not explicitly described in this technical solution are considered common technical features disclosed in the prior art.

[0046] Example 1

[0047] The disinfection hardware equipment used in the present invention mainly includes a disinfection container, a tourniquet frame used in conjunction with the disinfection container, a visual recognition module, a disinfectant feeding pump, an electric cylinder, a liquid level sensor, a hot air drying device for drying, etc. The visual recognition module is composed of a plurality of industrial cameras arranged around the disinfection container and a microprocessor connected thereto, which is used to scan and obtain information such as the volume distribution characteristics of the tourniquet; the disinfectant feeding pump is responsible for executing the proportioned injection of the disinfectant according to the instructions; the electric cylinder connected to the tourniquet frame is used to pull the tourniquet frame up from the disinfection container at the right time; the liquid level sensor can monitor the liquid level of the disinfectant in real time to ensure the accuracy and safety of the disinfection process; the hot air drying device is used to perform hot air drying operations on the tourniquet after disinfection. The above-mentioned hardware equipment are all existing technologies and are not the innovation points of the present invention, so they will not be described here.

[0048] See also Figure 1 In this embodiment, the method for adaptively controlling tourniquet disinfection parameters based on visual recognition and deep learning includes the following steps:

[0049] S1. Scan the volume distribution characteristics of the tourniquet in the frame through the visual recognition module, calculate the tourniquet volume fraction parameters using the deep residual network, and obtain the volume and volume fraction of the tourniquet in the tourniquet frame;

[0050] In specific implementation, in S1, obtaining the volume of the tourniquet in the tourniquet frame specifically includes the following process:

[0051] Multiple visual recognition modules are used to scan the tourniquet in the frame at multiple angles to obtain multiple tourniquet images from different perspectives, and these images are pre-processed to improve image quality.

[0052] The preprocessed image is input into a pre-trained deep residual network. The volume distribution features of the tourniquet are extracted through the convolutional layer, pooling layer, and fully connected layer of the deep residual network. The pre-trained deep residual network is trained based on existing sample data labeled with tourniquet volume information. The mapping relationship between the tourniquet volume and image features is learned through training.

[0053] The volume occupied by the tourniquet in the mesh frame is calculated based on the extracted volume distribution characteristics.

[0054] In specific implementation, in S1, obtaining the volume fraction of the tourniquet in the tourniquet frame specifically includes the following process:

[0055] The volume fraction of the tourniquet is determined based on the ratio of the volume occupied by the tourniquet in the mesh frame to the known total volume of the mesh frame.

[0056] In specific implementation, multiple visual recognition modules are first used to scan the tourniquet within the mesh frame from multiple angles, acquiring multiple tourniquet images from different perspectives. The goal of acquiring images from different angles is to comprehensively capture the three-dimensional spatial distribution of the tourniquet within the mesh frame, as images from a single angle cannot accurately reflect the tourniquet's three-dimensional stacking morphology and its actual volumetric proportion in space. After acquiring these images, they undergo preprocessing operations. This step primarily aims to improve image quality, including noise removal, image contrast enhancement, and image normalization, ensuring high-quality image data input into the deep residual network.

[0057] In specific implementation, the preprocessed image is input into a pre-trained deep residual network. The deep residual network has a unique architecture, which includes structures such as convolutional layers, pooling layers, and fully connected layers. The function of the convolutional layer is to extract features from the image. By sliding the convolution kernel on the image, it can automatically learn various local features of the tourniquet in the image, such as edges, textures, and other volume-related feature information; the pooling layer is used to downsample the features extracted by the convolutional layer, reducing the amount of data while retaining key features, which helps to reduce computational complexity and prevent overfitting; the fully connected layer integrates the features extracted by the previous layers, combining the scattered local features into a global feature representation with semantic meaning. The deep residual network is trained based on existing sample data labeled with tourniquet volume information. Through a large amount of sample training, the network learns the mapping relationship between tourniquet volume and image features, that is, it can understand which feature combinations in the image correspond to specific tourniquet volume sizes. When fed an image of a tourniquet to be inspected, the network extracts the tourniquet's volume distribution based on the learned mapping relationship, and then calculates the volume occupied by the tourniquet within the frame. Finally, the tourniquet's volume fraction is intuitively determined based on the ratio of the tourniquet's volume within the frame to the known total volume of the frame. This ratio reflects the tourniquet's density within the frame, or the proportion of space it occupies.

[0058] S2. obtaining liquid level information of the liquid required for disinfection based on the volume and volume fraction occupied by the tourniquet;

[0059] In specific implementation, in S2, obtaining the liquid level information of the liquid required for disinfection based on the volume and volume fraction occupied by the tourniquet specifically includes the following process:

[0060] Estimating the liquid level: Dividing the total volume of the tourniquet obtained by the effective cross-sectional area of ​​the sterilization container to obtain a preliminary liquid level. The sterilization container is a right cylindrical or prismatic structure.

[0061] Determine the liquid level adjustment coefficient: select the corresponding coefficient from the pre-established liquid level adjustment coefficient database based on the material properties of the tourniquet and the properties of the disinfectant;

[0062] Calculate the calibrated liquid level: multiply the initial liquid level by the liquid level adjustment coefficient to obtain the calibrated liquid level. The liquid level is the liquid level information of the liquid required for disinfection.

[0063] During specific implementation, the preliminary liquid level height is estimated. This is based on a simple geometric relationship. The total volume of the tourniquet is divided by the effective cross-sectional area of ​​the disinfection container to obtain the height occupied by the liquid in the container. Since the disinfection container is a right cylindrical or prismatic structure, this calculation method can quickly obtain a theoretical liquid level height. However, relying solely on this preliminary height is not accurate enough because it does not take into account the impact of the tourniquet material and the disinfectant properties on the actual required liquid level. Therefore, it is necessary to determine the liquid level adjustment coefficient next. This coefficient is selected from a pre-established liquid level adjustment coefficient database based on the material properties of the tourniquet and the properties of the disinfectant. Tourniquets of different materials have different absorption and permeability to disinfectants, and disinfectants with different properties (such as density, surface tension, etc.) will also affect their distribution in the tourniquet and the required immersion effect. Therefore, the liquid level adjustment coefficient can comprehensively consider these factors to correct the preliminary liquid level height. Finally, multiply the initial liquid level by the liquid level adjustment coefficient to obtain the calibrated liquid level. This height is the liquid level information required for disinfection. It can accurately reflect the actual height that the disinfectant should reach in the container in order to achieve a good disinfection effect, ensuring that the tourniquet can be fully immersed and disinfected.

[0064] S3. Based on the liquid level information, a convolutional neural network is used to generate the optimal disinfectant ratio, stirring speed, soaking time, stirring time, water rinsing time, and drying time. The disinfectant dosing pump is then used to perform the ratio injection.

[0065] In the specific implementation, in S3, the convolutional neural network is used to generate the optimal disinfectant ratio scheme, which specifically includes the following process:

[0066] Training a convolutional neural network model: collecting disinfection experimental data, including the volume and volume fraction of the tourniquet within the tourniquet frame, disinfectants of different concentrations and compositions, and corresponding disinfection effect evaluation data, to construct training samples;

[0067] The training samples are input into the convolutional neural network. By adjusting the parameters of the network's convolutional layer, pooling layer, and fully connected layer, the network learns the mapping relationship between the tourniquet characteristics and the disinfectant ratio scheme, thus obtaining a trained convolutional neural network model.

[0068] The volume and volume fraction of the tourniquet obtained by the visual recognition module and the disinfectant with preset composition and concentration are used as inputs to the convolutional neural network model, and the corresponding optimal disinfectant ratio scheme is output.

[0069] In specific implementation, in S3, determining the stirring speed, soaking time, stirring time, and drying time specifically includes the following process:

[0070] Obtaining historical disinfection data, the historical disinfection data including the volume and volume fraction of the tourniquet in the tourniquet frame under different disinfectant ratios, the corresponding reasonable stirring speed range, the optimal soaking time, the stirring time, the clean water rinsing time, and the drying time;

[0071] Use convolutional neural networks to mine and analyze historical disinfection data and establish a multi-parameter prediction model of time and speed;

[0072] The time and speed multi-parameter prediction model is used to predict the appropriate stirring speed, soaking time, stirring time, water rinsing time, and drying time according to the current tourniquet volume and volume fraction and the selected disinfectant ratio.

[0073] During specific implementation, the optimal soaking time, stirring time, water rinsing time, and drying time can be set to minimum values, such as the minimum soaking time is 30 minutes.

[0074] In specific implementation, the following processes are also included after S3:

[0075] The liquid level of the disinfectant is monitored in real time by a liquid level sensor, and the monitored liquid level is compared with the liquid level information obtained in S2. If the deviation exceeds the preset threshold, an alarm is issued and subsequent disinfection operations are suspended.

[0076] In specific implementation, a convolutional neural network (CNN) is used to build an intelligent decision-making system to generate the optimal disinfectant ratio and other disinfection parameters. First, a large amount of disinfection experimental data is collected as training samples. This data covers key information such as tourniquet volume and volume fraction, disinfectants of different concentrations and compositions, and corresponding disinfection efficacy evaluations. These samples are then fed into the CNN model. The model's convolutional layers automatically extract features from the input data, such as the correlation between tourniquet volume and disinfection efficacy, and the impact of different disinfectant compositions on disinfection efficacy. The pooling layer then performs dimensionality reduction on these features, further highlighting key features and reducing computational complexity. The fully connected layer integrates these extracted features, establishing a mapping between the input data (tourniquet volume and volume fraction, disinfectant characteristics, etc.) and the output target (disinfection efficacy). By continuously adjusting the parameters of the convolutional, pooling, and fully connected layers, the model gradually learns the inherent relationships between tourniquet characteristics and disinfectant ratios. After training is completed, when the tourniquet volume and volume fraction obtained by the visual recognition module, as well as the disinfectant information of preset composition and concentration, are input, the CNN model can output an optimal disinfectant ratio scheme based on the learned mapping relationship that can meet the disinfection effect requirements and make rational use of resources. At the same time, using historical disinfection data, CNN can also mine the relationship between the tourniquet volume, volume fraction and disinfection parameters (such as stirring speed, soaking time, etc.), establish a time and speed multi-parameter prediction model, and then predict the appropriate stirring speed, soaking time, stirring time, clean water rinsing time and drying time that match the current situation, so as to achieve coordinated optimization of various parameters of the disinfection process, improve disinfection efficiency and quality, and ensure the accuracy and reliability of the disinfection process. In addition, in order to ensure the bottom line requirements of disinfection effect, a minimum value of the disinfection parameter is also set, such as a minimum soaking time of 30 minutes.

[0077] During specific implementation, after generating the disinfection parameters and executing the proportioned injection, the liquid level of the disinfectant is monitored in real time by the liquid level sensor and compared with the liquid level information obtained in S2. The liquid level sensor uses technologies such as float, pressure measurement or ultrasound to obtain the liquid level data of the disinfectant in real time. The data reflects whether the actual injection amount of the disinfectant meets the ideal liquid level requirements determined in step S2. If the deviation between the actual liquid level and the preset liquid level exceeds the preset threshold, it means that there may be an abnormal injection of the disinfectant, such as insufficient or excessive injection, which will affect the disinfection effect or cause waste of resources. At this time, the system will immediately issue an alarm to remind the operator or the automatic control system to take measures, and suspend subsequent disinfection operations to avoid disinfection failure or resource loss due to liquid level problems, ensure that the entire disinfection process is carried out under the correct liquid level conditions, and ensure the accuracy and safety of the disinfection process, forming a closed-loop monitoring and feedback control mechanism.

[0078] S4. When the operation time reaches the sum of the soaking time and the stirring time, the tourniquet frame is pulled upward from the disinfection container by the electric cylinder to separate the tourniquet frame from the disinfectant;

[0079] In specific implementation, the following processes are also included after S4:

[0080] After the electric cylinder pulls the tourniquet frame upward from the disinfection container, the tourniquet frame is separated from the disinfectant and drained for 1 to 10 minutes.

[0081] In specific implementation, when the disinfection operation time reaches the sum of the pre-set soaking time and stirring time, it means that the tourniquet has completed the predetermined soaking and stirring disinfection process. At this time, the electric cylinder, a linear motion actuator, is used to smoothly pull the tourniquet-loaded mesh frame upward in the vertical direction from the disinfection container according to the preset control instructions, so that the tourniquet mesh frame is completely separated from the disinfectant liquid surface. Because the tourniquet will absorb a certain amount of disinfectant during the soaking and stirring process, if subsequent operations are performed directly, it may cause waste of disinfectant and have adverse effects on subsequent rinsing, drying and other steps. Therefore, after the mesh frame is pulled up, gravity is used to allow the disinfectant on the tourniquet to naturally drip back into the disinfection container for drainage. The drainage time is set to 1 to 10 minutes. This can fully recover the disinfectant and reduce waste, and prepare for the subsequent clean water rinsing and hot air drying steps, avoiding excessive residual disinfectant affecting the rinsing and drying effects.

[0082] S5. Based on the water flushing time and drying time obtained in S3, the tourniquet is rinsed with water and dried with hot air in sequence.

[0083] In the specific implementation, S5 specifically includes the following processes:

[0084] Clean the tourniquet inside the mesh frame by spraying clean water;

[0085] Obtain the material of the tourniquet based on the visual recognition module;

[0086] Based on the material of the tourniquet and its state after disinfection, the corresponding drying temperature and wind speed are selected from a pre-established drying parameter database to perform hot air drying on the tourniquet. The database is constructed based on experimental data of tourniquets of different materials under different drying conditions.

[0087] In a specific implementation, the visual recognition module includes multiple industrial cameras and a microprocessor connected to each industrial camera.

[0088] During specific implementation, the tourniquet within the mesh frame is cleaned with a jet of clean water according to the clean water flushing time determined in S3 to remove residual disinfectant. Subsequently, the visual recognition module acquires information about the tourniquet's material. An industrial camera captures images of the tourniquet from different angles, and a microprocessor analyzes and processes the images to accurately identify the tourniquet's material. Based on the tourniquet's material and its actual state after disinfection, the system consults a pre-established drying parameter database, selects a matching drying temperature and wind speed, and then uses a hot air drying device to perform hot air drying on the tourniquet. This database is constructed based on experimental data from tourniquets of various materials under different drying conditions. This ensures that the drying process is both efficient and harmless, achieving thorough cleaning and drying of the tourniquet.

[0089] The above description of the embodiments is intended to facilitate understanding and use of the invention by those skilled in the art. It will be apparent that those skilled in the art can readily make various modifications to these embodiments and apply the general principles described herein to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above-described embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention, without departing from the scope of the present invention, should be within the scope of protection of the present invention.

Claims

1. A tourniquet disinfection parameter adaptive control method based on visual recognition and deep learning, characterized in that: The following steps are involved: S1. Scan the volume distribution characteristics of the tourniquet in the frame through the visual recognition module, calculate the tourniquet volume fraction parameters using the deep residual network, and obtain the volume and volume fraction of the tourniquet in the tourniquet frame; S2. obtaining liquid level information of the liquid required for disinfection based on the volume and volume fraction occupied by the tourniquet; S3. Based on the liquid level information, a convolutional neural network is used to generate the optimal disinfectant ratio, stirring speed, soaking time, stirring time, water rinsing time, and drying time. The disinfectant dosing pump is then used to perform the ratio injection. S4. When the operation time reaches the sum of the soaking time and the stirring time, the tourniquet frame is pulled upward from the disinfection container by the electric cylinder to separate the tourniquet frame from the disinfectant; S5. Based on the water flushing time and drying time obtained in S3, the tourniquet is rinsed with water and dried with hot air in sequence.

2. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: In S1, the specific volume of the tourniquet in the tourniquet frame is obtained. The following processes are included: Multiple visual recognition modules are used to scan the tourniquet in the frame at multiple angles to obtain multiple tourniquet images from different perspectives, and these images are pre-processed to improve image quality. The preprocessed image is input into a pre-trained deep residual network. The volume distribution features of the tourniquet are extracted through the convolutional layer, pooling layer, and fully connected layer of the deep residual network. The pre-trained deep residual network is trained based on existing sample data labeled with tourniquet volume information. The mapping relationship between the tourniquet volume and image features is learned through training. The volume occupied by the tourniquet in the mesh frame is calculated based on the extracted volume distribution characteristics.

3. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 2, characterized in that: In S1, obtaining the volume fraction of the tourniquet in the tourniquet frame specifically includes the following steps: The volume fraction of the tourniquet is determined based on the ratio of the volume occupied by the tourniquet in the mesh frame to the known total volume of the mesh frame.

4. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: In S2, the liquid level information of the liquid required for disinfection is obtained based on the volume and volume fraction occupied by the tourniquet. The following processes are included: Estimating the liquid level: Dividing the total volume of the tourniquet obtained by the effective cross-sectional area of ​​the sterilization container to obtain a preliminary liquid level. The sterilization container is a right cylindrical or prismatic structure. Determine the liquid level adjustment coefficient: select the corresponding coefficient from the pre-established liquid level adjustment coefficient database based on the material properties of the tourniquet and the properties of the disinfectant; Calculate the calibrated liquid level: multiply the initial liquid level by the liquid level adjustment coefficient to obtain the calibrated liquid level. The liquid level is the liquid level information of the liquid required for disinfection.

5. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: In S3, a convolutional neural network is used to generate the optimal disinfectant ratio scheme. The following processes are included: Training a convolutional neural network model: collecting disinfection experimental data, including the volume and volume fraction of the tourniquet within the tourniquet frame, disinfectants of different concentrations and compositions, and corresponding disinfection effect evaluation data, to construct training samples; The training samples are input into the convolutional neural network. By adjusting the parameters of the network's convolutional layer, pooling layer, and fully connected layer, the network learns the mapping relationship between the tourniquet characteristics and the disinfectant ratio scheme, thus obtaining a trained convolutional neural network model. The volume and volume fraction of the tourniquet obtained by the visual recognition module and the disinfectant with preset composition and concentration are used as inputs to the convolutional neural network model, and the corresponding optimal disinfectant ratio scheme is output.

6. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: In S3, the stirring speed, soaking time, stirring time and drying time are determined. The following processes are included: Obtaining historical disinfection data, the historical disinfection data including the volume and volume fraction of the tourniquet in the tourniquet frame under different disinfectant ratios, the corresponding reasonable stirring speed range, the optimal soaking time, the stirring time, the clean water rinsing time, and the drying time; Use convolutional neural networks to mine and analyze historical disinfection data and establish a multi-parameter prediction model of time and speed; The time and speed multi-parameter prediction model is used to predict the appropriate stirring speed, soaking time, stirring time, water rinsing time, and drying time according to the current tourniquet volume and volume fraction and the selected disinfectant ratio.

7. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: The following steps are also included after S3: The liquid level of the disinfectant is monitored in real time by a liquid level sensor, and the monitored liquid level is compared with the liquid level information obtained in S2. If the deviation exceeds the preset threshold, an alarm is issued and subsequent disinfection operations are suspended.

8. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: After S4 The following processes are included: After the electric cylinder pulls the tourniquet frame upward from the disinfection container, the tourniquet frame is separated from the disinfectant and drained for 1 to 10 minutes.

9. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: S5 specifically includes the following processes: Clean the tourniquet inside the mesh frame by spraying clean water; Obtain the material of the tourniquet based on the visual recognition module; Based on the material of the tourniquet and its state after disinfection, the corresponding drying temperature and wind speed are selected from a pre-established drying parameter database to perform hot air drying on the tourniquet. The database is constructed based on experimental data of tourniquets of different materials under different drying conditions.

10. The method for adaptive control of tourniquet disinfection parameters based on visual recognition and deep learning according to claim 1, characterized in that: The visual recognition module includes a plurality of industrial cameras and a microprocessor connected to each industrial camera.

Images