Intelligent classification and closed-loop management system for medical waste in public health scenarios

Abstract

The application discloses a medical waste intelligent classification and closed-loop management system in a public health scenario, relates to the technical field of medical waste treatment, and comprises a data training module, which obtains various medical waste image data and trains and optimizes an intelligent classification model; the intelligent classification model adopts an improved deep learning model as a basic model; a classification module, in which a worker places medical waste in a camera recognition area of the classification module, and the current medical waste is cooperatively recognized based on the intelligent classification model in combination with a camera and a spectrometer, the type and ingredient characteristics of the current medical waste are judged, and corresponding dropping position information is given; the classification module is integrated with a light supplementing unit to adaptively adjust the brightness under the current environment, thereby increasing the camera recognition efficiency. The application simplifies the operation process, reduces the contact risk and the wrong dropping probability of medical staff, realizes digital recording of the whole process, traceability, supports remote supervision and cooperative scheduling, and provides a basis for resource allocation optimization and cost control.

Description

Technical Field

[0001] This invention relates to the field of medical waste treatment technology, specifically to an intelligent classification and closed-loop management system for medical waste in public health settings. Background Technology

[0002] Medical waste, generated during medical institutions' diagnosis, treatment, nursing, and testing activities, possesses hazardous characteristics such as infectiousness, injury, chemical activity, and pharmaceutical properties. Its standardized classification and closed-loop management throughout the entire process are core aspects of public health safety control, ecological environmental protection, and occupational health protection for medical personnel, directly impacting public health, environmental safety, and the overall completeness of the public health system. With the rapid development of my country's medical industry and the continuous expansion of medical institutions' treatment scale, the amount of medical waste generated has been increasing year by year. Simultaneously, the promotion of new diagnostic and treatment technologies has brought about new types of medical waste, placing higher demands on the accuracy of medical waste classification, the efficiency of management, and the traceability of the entire process. Traditional medical waste management models are no longer adequate to meet the management needs of modern public health scenarios.

[0003] Currently, my country's medical waste management still has many weaknesses, with the lack of standardization in the classification process being particularly prominent, becoming a key bottleneck restricting the improvement of management quality. Existing medical waste classification mainly relies on manual judgment and disposal by staff. Limited by factors such as varying levels of staff familiarity with the "Medical Waste Classification Catalog (2021 Edition)," workload, operational negligence, and insufficient professional training, problems such as classification confusion and misdisposal are prone to occur. For example, infectious waste is mixed with pathological waste, pharmaceutical waste is not strictly separated from chemical waste, and even recyclable non-medical waste is mistakenly mixed with infectious waste. This not only increases subsequent disposal costs and wastes resources but also poses serious public health and safety hazards and environmental pollution risks. According to surveys, approximately 35% of medical waste bags in some primary healthcare institutions have mixed waste disposal issues, and the misclassification rate for specimen waste is as high as 40%. The accuracy rate of manual classification is generally only around 65%, far below the precision requirements of modern management.

[0004] To address the shortcomings of manual sorting, some medical institutions have attempted to introduce simple automated sorting equipment. However, existing equipment still has significant technical deficiencies, making it difficult to achieve accurate and efficient sorting and closed-loop management throughout the entire process. It is also difficult to effectively distinguish between medical wastes with similar appearances but different components, and cannot meet the diverse sorting needs of medical waste in complex public health scenarios. In response, we propose an intelligent sorting and closed-loop management system for medical waste in public health scenarios. Summary of the Invention

[0005] To address the aforementioned technical challenges, this technical solution provides an intelligent classification and closed-loop management system for medical waste in public health scenarios. This solution resolves the difficulties in achieving accurate and efficient classification and full-process closed-loop management.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: an intelligent classification and closed-loop management system for medical waste in public health scenarios, comprising: The data training module acquires image data of various types of medical waste and trains and optimizes the intelligent classification model; the intelligent classification model uses an improved deep learning model as the base model. In the sorting module, staff place medical waste in the camera recognition area. Based on the intelligent sorting model, combined with the camera and spectrometer, the module collaboratively identifies the medical waste, determines its type and composition characteristics, and provides corresponding placement location information. The sorting module also integrates a supplementary lighting unit to adaptively adjust the brightness in the current environment, increasing the camera's recognition efficiency. The corresponding classification warning module generates corresponding light warnings in the corresponding disposal area based on the disposal location information, so as to remind staff of the current disposal area for medical waste; The control module receives data transmitted from each module in real time, summarizes, analyzes, and verifies it; generates management instructions; and manages the coordinated operation of each module.

[0007] Preferably, the medical waste image data includes infectious, pathological, traumatic, pharmaceutical, and chemical medical waste. Various types of medical waste image data are collected from multiple dimensions using cameras, macro lenses, and 3D image acquisition devices from different departments. The acquired image data is then standardized, enhanced, and normalized.

[0008] Preferably, the intelligent classification model is trained and optimized by selecting an improved MobileNetV2 lightweight convolutional neural network model and introducing a lightweight attention mechanism to strengthen the attention weight of key feature regions of medical waste; the model framework includes an input layer, a feature extraction layer, an attention fusion layer and a classification layer. The intelligent classification model inputs the preprocessed training dataset into the improved deep learning model and uses the backpropagation algorithm to adjust the model's weight parameters, bias parameters, and network layer parameters through multiple rounds of iterative training. The model training effect is verified in real time using a validation dataset. When the model performance reaches a recognition accuracy of ≥95%, training is stopped, and the initial training optimization of the model is completed.

[0009] Preferably, the classification module includes a supplementary lighting unit, which adaptively adjusts the brightness in the current environment by collecting ambient lighting parameters of the current identification area in real time through a built-in light sensor, including brightness value and light uniformity, and comparing them with the set optimal light parameter threshold to determine whether the light is sufficient in the current environment. If the light is insufficient, the supplementary lighting unit automatically and adaptively adjusts the supplementary lighting power and angle. After the supplementary lighting adjustment is completed, the camera starts to collect images of the medical waste from multiple angles and transmits them to the intelligent classification model in real time.

[0010] Preferably, the collaborative identification steps for current medical waste are based on an intelligent classification model combined with cameras and spectrometers: The camera captures images, and the spectrometer scans them, covering the surface and shallow interior of the medical waste. The wavelength range is set to 400-1000nm to obtain the spectral characteristic curves of the medical waste. The spectral curves are analyzed to extract the core component characteristics of the medical waste. After the extracted component characteristic data is standardized, it is synchronously transmitted to the intelligent classification model. The intelligent classification model receives image feature data and component feature data, processes the image data through a feature extraction layer to extract the visual features of medical waste, and performs bidirectional fusion of image visual features and spectral component features, eliminating the recognition bias between the two features based on a feature matching algorithm. The intelligent classification model, based on the fusion features after collaborative recognition and combined with the classification logic optimized by training, determines the category of medical waste and divides it into corresponding categories. Combining spectral composition features, it further determines the core components and potential risks of medical waste, generates category determination results and component feature descriptions, and transmits them to the control module for data verification.

[0011] Preferably, after receiving the verification pass signal from the control module, the classification module matches the preset delivery area database based on the category determination result to generate delivery location information; and synchronously outputs the delivery location information to the classification warning module for light warning.

[0012] Preferably, after receiving the placement location information, the corresponding category warning module parses the information, matches the corresponding category of light warning parameters in the warning parameter database, and sends a start command to the LED warning lights in the corresponding placement area to activate the lights and light them up according to the matched colors.

[0013] Preferably, the control module includes: The data processing unit is used to summarize, analyze, and verify the data transmitted from each module. The instruction generation unit is used to generate management instructions based on the data processing results; The module coordination unit coordinates and schedules various modules based on management commands; The data storage unit is used to store the data transmitted and analyzed between the various modules.

[0014] Preferably, the data processing unit uses an abnormal data tracing and verification algorithm to locate identification errors, warning failures, and delivery anomalies, and generates investigation results.

[0015] Preferably, when a single module fails, the module collaboration unit automatically switches to backup mode, calls historical data for troubleshooting, and sends fault warnings and maintenance prompts to management personnel.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention can significantly improve the intelligence, standardization, and safety of medical waste management, achieving closed-loop control throughout the entire process. The data training module continuously optimizes and improves the deep learning model, enhancing recognition accuracy and generalization ability while reducing reliance on manual labor. The classification module uses cameras and spectrometers for collaborative recognition, coupled with an adaptive supplementary lighting unit, to reduce environmental interference, simplify the operation process, and reduce the risk of contact and misdisposal for medical staff. It also enables full-process digital recording and traceability, supporting remote supervision and collaborative scheduling, while providing a basis for resource allocation optimization and cost control. Attached Figure Description

[0017] Figure 1 Framework diagram for an intelligent classification and closed-loop management system. Detailed Implementation

[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0019] Reference Figure 1 As shown, the intelligent classification and closed-loop management system for medical waste in a public health setting includes: The data training module acquires image data of various types of medical waste and trains and optimizes the intelligent classification model; the intelligent classification model uses an improved deep learning model as the base model. In the sorting module, staff place medical waste in the camera recognition area. Based on the intelligent sorting model, combined with the camera and spectrometer, the module collaboratively identifies the medical waste, determines its type and composition characteristics, and provides corresponding placement location information. The sorting module also integrates a supplementary lighting unit to adaptively adjust the brightness in the current environment, increasing the camera's recognition efficiency. The corresponding classification warning module generates corresponding light warnings in the corresponding disposal area based on the disposal location information, so as to remind staff of the current disposal area for medical waste; The control module receives data transmitted from each module in real time, summarizes, analyzes, and verifies it; generates management instructions; and manages the coordinated operation of each module.

[0020] Medical waste image data includes infectious, pathological, traumatic, pharmaceutical, and chemical medical waste. Various types of medical waste image data are collected from multiple dimensions using cameras, macro lenses, and 3D image acquisition equipment from different departments. The acquired image data undergoes standardization, image enhancement, and normalization processing.

[0021] The intelligent classification model is trained and optimized by introducing a lightweight attention mechanism on the basis of the improved MobileNetV2 model. By adaptively allocating attention weights, the focus on key feature regions of medical waste is strengthened, while the influence of background interference and irrelevant regions is weakened. This effectively solves the problem of identification error caused by the similar appearance and indistinct features of different types of medical waste. The intelligent classification model framework is clearly divided into four layers, each working together to achieve medical waste feature extraction and classification: The input layer is responsible for receiving medical waste image data that has undergone standardization, enhancement, and normalization preprocessing, and converting it into a tensor format that the model can recognize; the feature extraction layer is based on an improved MobileNetV2 deep separable convolutional structure, extracting shallow texture features and deep semantic features of medical waste layer by layer; the attention fusion layer fuses the key features processed by the attention mechanism with the features output by the feature extraction layer, further enhancing effective features and suppressing redundant information; the classification layer uses the Softmax activation function to map the fused features to the category space of the five major categories of medical waste, outputting the recognition probability of each category, and completing the medical waste type judgment.

[0022] The model training process strictly follows a standardized procedure. The preprocessed medical waste image dataset is divided into training, validation, and test datasets in a 7:2:1 ratio. During the training phase, the training dataset is input into the improved deep learning model. The backpropagation algorithm is used to dynamically adjust the model's weight parameters, bias parameters, and network layer parameters through multiple rounds of iterative training. After each iteration, the model's loss function value is calculated, and the parameter combination is continuously optimized using gradient descent to reduce model recognition errors. Simultaneously, the model training effect is verified in real time using the validation dataset. After each iteration, core performance indicators such as recognition accuracy and recall on the validation set are output to continuously monitor the model's generalization ability and avoid overfitting and underfitting. When the model's recognition accuracy on the validation dataset stably reaches ≥95%, and auxiliary indicators such as recall and F1 score meet preset standards, iterative training is stopped, and the initial model training optimization is completed. This ensures that the model can accurately identify various types of medical waste, providing reliable support for the collaborative recognition of subsequent classification modules.

[0023] The classification module includes a supplementary lighting unit based on a built-in high-precision light sensor. This sensor is embedded at the edge of the camera's recognition area in the classification module, enabling real-time and accurate acquisition of ambient light parameters for the current recognition area. The core parameters include two key indicators: ambient brightness (unit: lux) and light uniformity. The acquisition frequency is set to twice per second to ensure timely capture of light changes. Simultaneously, the system presets optimal light parameter thresholds for medical waste recognition. The brightness threshold is set to 300-500 lux, and the light uniformity threshold is no less than 85%. These thresholds were determined through multiple tests and calibrations, taking into account the characteristic recognition requirements of different types of medical waste, ensuring both image clarity and preventing interference from strong light reflections on camera acquisition.

[0024] The light sensor compares and analyzes the parameters collected in real time with preset thresholds to quickly determine whether the current ambient light meets the collection requirements: if the detected brightness value is below 300 lux or the light uniformity is below 85%, it is determined that the light is insufficient or uneven, and the supplementary lighting unit immediately starts the adaptive adjustment mode; during the adjustment process, the supplementary lighting unit automatically adapts to the ambient light gap through the built-in power adjustment module, and gradually adjusts the supplementary lighting power to avoid the problem of strong light reflection caused by excessive power and insufficient supplementary lighting caused by insufficient power; at the same time, the supplementary lighting unit is equipped with a rotatable and adjustable supplementary lighting angle component, which can automatically fine-tune the supplementary lighting angle within the range of 0-45°, and accurately supplement the light for weak light areas to ensure that the entire recognition area is uniformly lit without obvious shadows or light spots.

[0025] After the supplementary lighting is adjusted, the system will confirm through a signal feedback mechanism that the illumination parameters of the current recognition area have reached the preset optimal threshold. At this time, the camera will be triggered to start the acquisition process. The camera will acquire images of the medical waste placed in the recognition area from multiple angles according to the preset acquisition logic, acquiring 2-3 clear images from each angle. After the acquisition is completed, the image data will be transmitted to the intelligent classification model in real time through the internal data transmission channel, providing high-quality image support for the subsequent identification of the type and composition characteristics of medical waste, and ensuring the accuracy and efficiency of classification and identification.

[0026] The following steps are used for the collaborative identification of medical waste based on an intelligent classification model combined with cameras and spectrometers: The camera captures images, and the spectrometer scans them, covering the surface and shallow interior of the medical waste. The wavelength range is set to 400-1000nm to obtain the spectral characteristic curves of the medical waste. The spectral curves are analyzed to extract the core component characteristics of the medical waste. After the extracted component characteristic data is standardized, it is synchronously transmitted to the intelligent classification model. The intelligent classification model receives image feature data and component feature data, processes the image data through a feature extraction layer to extract the visual features of medical waste, and performs bidirectional fusion of image visual features and spectral component features, eliminating the recognition bias between the two features based on a feature matching algorithm. The intelligent classification model, based on the fusion features after collaborative recognition and combined with the classification logic optimized by training, determines the category of medical waste and divides it into corresponding categories. Combining spectral composition features, it further determines the core components and potential risks of medical waste, generates category determination results and component feature descriptions, and transmits them to the control module for data verification.

[0027] The camera acquires images, and the spectrometer scans the medical waste, covering both the surface and shallow interior layers. The wavelength range is set to 400-1000 nm to obtain the spectral characteristic curves of the medical waste. The spectral curves are analyzed to extract the core component characteristics of the medical waste. After standardization, the extracted component characteristic data is synchronously transmitted to the intelligent classification model. The spectral characteristic curve extraction is achieved through: The spectrometer collects light intensity signals from medical waste within the wavelength range λ (400nm ≤ λ ≤ 1000nm), corrects for this signal using Lambert-Beer's law, and obtains a standardized spectral characteristic curve, as shown in the following formula: ; in The actual spectral reflectance intensity of medical waste at wavelength λ (unit: cd / m²) 2 );

[0028] The intensity of incident light on the spectrometer at wavelength λ (unit: cd / m²) 2 ); Absorption coefficient of medical waste at wavelength λ (unit: m) -1 ), which is directly related to the composition of medical waste, is the core parameter for component feature extraction; The shallow depth of the spectral scan (unit: mm) is set to 0.1-0.5 mm based on the characteristics of medical waste, covering the shallow interior. Environmental correction factor (range 0.95-1.05), used to eliminate the interference of ambient temperature and humidity on spectral acquisition; Based on the spectral characteristic curve, the characteristic peaks corresponding to the core components of medical waste are extracted using a peak detection algorithm, as shown in the following formula: ; in: : Characteristic values ​​of core components of medical waste, corresponding characteristic peak values ​​of the components; Wavelength weighting coefficient: Differentiated weights are set for the characteristic wavelengths of different components (such as the characteristic wavelength of infectious waste protein at 280nm and the characteristic wavelength of chemical reagents at 550nm) to enhance the characteristics of core components and suppress interference from irrelevant wavelengths. The intelligent classification model receives image feature data and component feature data. It processes the image data through a feature extraction layer to extract the visual features of medical waste. The model then performs bidirectional fusion of the image visual features and spectral component features, eliminating the recognition bias between the two features based on a feature matching algorithm. The visual feature extraction, bidirectional feature fusion, and bias elimination utilize the following algorithm formulas: The feature extraction layer of the intelligent classification model (improved MobileNetV2) extracts visual features from medical waste images through depthwise separable convolution, as shown in the following formula: ; In the formula: : Extracted visual feature vector of medical waste image (dimension N×1, where N is the feature dimension); Depthwise separable convolution operations can extract features from the input image X, reducing computational cost. Batch normalization is used to stabilize model training and accelerate convergence. : Activation function, restricting the output range to [0,6], suitable for lightweight model requirements; A weighted fusion algorithm is used to combine the visual features of the image. With spectral composition characteristics Perform bidirectional fusion to obtain fusion features The formula is as follows: ; In the formula: Feature fusion weights (range 0.4-0.6) can be dynamically adjusted based on recognition confidence, improving performance when recognition is ambiguous. Weighting, boosted when the image is clear Weight; Image visual feature vector; : Standardized spectral component feature vector; Based on the feature matching algorithm, the deviation value between the two features is calculated, and the recognition bias is eliminated by a deviation correction factor, as shown in the following formula: ; In the formula: The deviation between the visual features and the spectral composition features of an image; The final fusion feature after bias correction; The maximum value of the two features is used to normalize the deviation value and avoid overcorrection of the deviation. Intelligent classification models are based on fused features after collaborative recognition. Combining the optimized classification logic, medical waste is categorized and classified accordingly. Furthermore, based on spectral composition characteristics, the core components and potential risks of the medical waste are determined, generating category determination results and component characteristic descriptions, which are then transmitted to the control module for data verification. The following algorithm formula is used for category determination and risk assessment: The Softmax classification function is used to calculate the category probability based on the fused features to determine the category of medical waste. The formula is as follows: ; In the formula: : The probability that the medical waste corresponding to the fusion feature belongs to the i-th category (i=1,2,3,4,5, corresponding to the five major categories of medical waste respectively). The weight matrix of the i-th class is obtained by optimizing the model during training; : The bias term for the i-th category; Total number of medical waste categories (K=5 here); The category corresponding to the highest probability is taken as the final category determination result for medical waste, that is: ; Combining spectral component characteristic values The risk assessment value is calculated to determine the potential risk level, using the following formula: ; In the formula: : Potential risk assessment value for medical waste (range [0,1]), R≥0.8 is high risk, 0.5≤R<0.8 is medium risk, and R<0.5 is low risk; Component feature weighting coefficient (value 1.2) strengthens the impact of core components on risk; The variance of component eigenvalues ​​reflects the uniformity of component distribution; the larger the variance, the higher the risk. : Variance weighting coefficient (value 0.3), used to assist in assessing risk level.

[0029] Model generates category determination results Including ingredient descriptions (including) The data (R) is transmitted to the control module for data verification to ensure that the judgment result is accurate and reliable.

[0030] After receiving the verification pass signal from the control module, the classification module matches the preset delivery area database based on the category determination result and generates delivery location information; the delivery location information is then synchronously output to the classification warning module for light warning.

[0031] After receiving the placement location information, the corresponding category warning module parses the information. Based on the parsing results, it matches the corresponding category of light warning parameters in the warning parameter database and sends a start command to the LED warning lights in the corresponding placement area to activate the lights and illuminate them according to the matched colors.

[0032] The control module includes: The data processing unit is used to summarize, analyze, and verify the data transmitted from each module. The instruction generation unit is used to generate management instructions based on the data processing results; The module coordination unit coordinates and schedules various modules based on management commands; The data storage unit is used to store the data transmitted and analyzed between the various modules.

[0033] The data processing unit uses an abnormal data tracing and verification algorithm to locate identification errors, warning failures, and delivery anomalies, and generates investigation results.

[0034] When a single module fails, the module collaboration unit automatically switches to backup mode, retrieves historical data for troubleshooting, and sends fault warnings and maintenance prompts to management personnel.

[0035] Based on the medical waste classification results determined by the intelligent classification model, the system matches the preset disposal area database to generate specific disposal location information, which is then simultaneously transmitted to the classification warning module. After parsing this information, the classification warning module matches the corresponding category's light parameters in the warning parameter database and sends an activation command to the LED warning lights in the corresponding disposal area, causing them to light up in the specified color to guide staff to dispose of waste accurately.

[0036] As the central hub of the system, the control module integrates four major units: data processing, instruction generation, module collaboration, and data storage. It coordinates the collaborative operation of each module. The data processing unit summarizes and analyzes the data transmitted by each module, and through anomaly data tracing and verification algorithms, it locates and identifies errors, warning malfunctions, and issues, and generates investigation results. The instruction generation unit automatically generates various management instructions based on the data processing results.

[0037] The module collaboration unit schedules each module according to management instructions. When a single module fails, it automatically switches to backup mode, calls historical data to ensure that core functions are not interrupted, and sends fault warnings and maintenance prompts to management personnel. The data storage unit uses encryption to store the data transmitted by each module, analysis results and operation logs, providing complete data support for system optimization and supervision and traceability, and realizing a closed-loop operation of medical waste classification, warning and control.

[0038] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the present invention is defined by the appended technical solutions and their equivalents.

Claims

1. A smart classification and closed-loop management system for medical waste in public health scenarios, characterized in that, include: The data training module acquires image data of various types of medical waste and trains and optimizes the intelligent classification model. The intelligent classification model uses an improved deep learning model as its base model; In the classification module, staff place medical waste in the camera recognition area of ​​the classification module. Based on the intelligent classification model, combined with the camera and spectrometer, the current medical waste is collaboratively identified, the type and composition characteristics of the current medical waste are determined, and the corresponding disposal location information is provided. The classification module integrates a supplementary lighting unit to adaptively adjust the brightness in the current environment, increasing the camera's recognition efficiency; The corresponding classification warning module generates corresponding light warnings in the corresponding disposal area based on the disposal location information, so as to remind staff of the current disposal area for medical waste; The control module receives data transmitted from each module in real time, summarizes, analyzes, and verifies it; and generates management commands. Manage the coordinated operation of each module.

2. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 1, characterized in that: Medical waste image data includes infectious, pathological, traumatic, pharmaceutical, and chemical medical waste. Various types of medical waste image data are collected from multiple dimensions using cameras, macro lenses, and 3D image acquisition equipment from different departments. The acquired image data undergoes standardization, image enhancement, and normalization processing.

3. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 1, characterized in that: The intelligent classification model was trained and optimized by selecting the improved MobileNetV2 lightweight convolutional neural network model and introducing a lightweight attention mechanism to strengthen the attention weight of key feature regions of medical waste. The model framework includes an input layer, a feature extraction layer, an attention fusion layer, and a classification layer; The intelligent classification model inputs the preprocessed training dataset into the improved deep learning model and uses the backpropagation algorithm to adjust the model's weight parameters, bias parameters, and network layer parameters through multiple rounds of iterative training. The training effect of the model is verified in real time by validating the dataset. When the model performance reaches a recognition accuracy of ≥95%, training is stopped and the initial training optimization of the model is completed.

4. The intelligent classification and closed-loop management system for medical waste in a public health setting according to claim 1, characterized in that: The classification module includes a supplementary lighting unit, which adaptively adjusts the brightness in the current environment. It collects the ambient lighting parameters of the current identification area in real time through a built-in light sensor, including the brightness value and the uniformity of the light. It compares these parameters with the set optimal light parameter threshold to determine whether the light is sufficient in the current environment. If the light is insufficient, the supplementary lighting unit automatically adjusts the supplementary lighting power and angle. After the supplementary lighting adjustment is completed, the camera starts to collect images of the medical waste from multiple angles and transmits them to the intelligent classification model in real time.

5. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 1, characterized in that, The following steps are used for the collaborative identification of medical waste based on an intelligent classification model combined with cameras and spectrometers: The camera captures images, and the spectrometer scans them, covering the surface and shallow interior of the medical waste. The wavelength range is set to 400-1000nm to obtain the spectral characteristic curves of the medical waste. The spectral curves are analyzed to extract the core component characteristics of the medical waste. After the extracted component characteristic data is standardized, it is synchronously transmitted to the intelligent classification model. The intelligent classification model receives image feature data and component feature data, processes the image data through a feature extraction layer to extract the visual features of medical waste, and performs bidirectional fusion of image visual features and spectral component features, eliminating the recognition bias between the two features based on a feature matching algorithm. The intelligent classification model, based on the fusion features after collaborative recognition and combined with the classification logic optimized by training, determines the category of medical waste and divides it into corresponding categories. Combining spectral composition features, it further determines the core components and potential risks of medical waste, generates category determination results and component feature descriptions, and transmits them to the control module for data verification.

6. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 1, characterized in that: After receiving the verification pass signal from the control module, the classification module matches the preset delivery area database based on the category determination result and generates delivery location information; the delivery location information is then synchronously output to the classification warning module for light warning.

7. The intelligent classification and closed-loop management system for medical waste in a public health setting according to claim 1, characterized in that: After receiving the placement location information, the corresponding category warning module parses the information. Based on the parsing results, it matches the corresponding category of light warning parameters in the warning parameter database and sends a start command to the LED warning lights in the corresponding placement area to activate the lights and illuminate them according to the matched colors.

8. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 1, characterized in that, The control module includes: The data processing unit is used to summarize, analyze, and verify the data transmitted from each module. The instruction generation unit is used to generate management instructions based on the data processing results; The module coordination unit coordinates and schedules various modules based on management commands; The data storage unit is used to store the data transmitted and analyzed between the various modules.

9. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 8, characterized in that: The data processing unit uses an abnormal data tracing and verification algorithm to locate identification errors, warning failures, and delivery anomalies, and generates investigation results.

10. The intelligent classification and closed-loop management system for medical waste in a public health scenario according to claim 8, characterized in that: When a single module fails, the module collaboration unit automatically switches to backup mode, retrieves historical data for troubleshooting, and sends fault warnings and maintenance prompts to management personnel.

Images