An artificial intelligence-based pharmaceutical warehouse management method, device and medium

Abstract

This application discloses an artificial intelligence-based drug storage management method, equipment, and medium, relating to the field of drug management technology. The method includes: collecting drug images and environmental perception data in a designated drug warehouse, analyzing the drug images to determine the target storage conditions for the designated drug to be identified; inputting the environmental perception data into a pre-trained multimodal fusion model to extract cross-modal correlation features, and combining the target storage conditions to generate a drug status anomaly probability score for the designated drug to be identified; if the drug status anomaly probability score exceeds a preset threshold, marking the location of the abnormal drug, and generating corresponding adjustment instructions based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instructions, thereby realizing the storage management of the designated drug.

Description

Technical Field

[0001] This application relates to the field of pharmaceutical management technology, and in particular to a pharmaceutical warehouse management method, equipment and medium based on artificial intelligence. Background Technology

[0002] Designated drug warehousing management is a core link in the security of the medical supply chain, involving categories such as temperature-sensitive drugs and high-value biological agents that require strict environmental control. Traditional warehousing systems mostly rely on manual inspections and single-sensor threshold alarm mechanisms, making it difficult to dynamically perceive the coupled impact of changes in drug packaging status and complex environmental factors. Especially in scenarios involving the mixed storage of multiple batches and multiple categories, existing systems cannot effectively correlate the spatiotemporal relationship between temperature and humidity fluctuation trends and drug shelf-life decay, resulting in delayed detection of abnormal conditions and increasing the risk of reduced drug potency or even spoilage.

[0003] Furthermore, traditional methods often rely on manual intervention after anomalies are detected, failing to generate migration paths that meet chemical compatibility requirements in real time. This can exacerbate drug deterioration due to secondary contamination during handling or sudden environmental changes. Current intelligent warehousing systems mostly employ preset fixed-path algorithms for path planning, neglecting crucial factors such as the dynamic constraints of robotic arms in handling equipment and chemical compatibility between shelves. This can easily lead to excessive transport vibration or the risk of mixing hazardous materials. Summary of the Invention

[0004] This application provides an artificial intelligence-based drug storage management method, equipment, and medium to solve the aforementioned technical problems.

[0005] On the one hand, embodiments of this application provide an artificial intelligence-based drug warehouse management method, including:

[0006] Collect images of drugs and environmental sensing data in a designated drug warehouse, and analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental sensing data includes temperature and humidity time series and drug batch metadata;

[0007] The environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, and combined with the target storage conditions, a drug status anomaly probability score for the specified drug to be identified is generated.

[0008] If the probability score of the abnormal drug status exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction and realize the designated drug warehouse management.

[0009] In one implementation of this application, the environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, specifically including:

[0010] The environmental perception data is input into a pre-trained multimodal fusion model, and a sliding window Fourier transform is performed on the temperature and humidity time series to detect periodic fluctuation patterns.

[0011] Based on the aforementioned periodic fluctuation pattern, the degree of deviation between the current fluctuation amplitude and the historical normal fluctuation range is calculated, and the degree of deviation is used as the first weighting factor.

[0012] The production date and expiration date in the drug batch metadata are parsed, and the remaining expiration date percentage of the specified drug to be identified is calculated in combination with the current timestamp, so that the remaining expiration date percentage is used as the second weighting factor;

[0013] By fusing the first weighting factor and the second weighting factor, a modal dynamic weighting matrix is ​​generated, and based on the modal dynamic weighting matrix, the confidence level of different data modalities in the time dimension is calculated;

[0014] Based on the confidence differences of different data modalities, modality weights are dynamically allocated, and a spatiotemporal convolutional network is used to extract cross-modal association features.

[0015] In one implementation of this application, if the probability score of the abnormal drug status exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction, thereby realizing the designated drug warehouse management, specifically including:

[0016] The abnormal drug status probability score is compared with a preset threshold. If the abnormal drug status probability score exceeds the preset threshold, the shelf number of the abnormal drug is located based on the three-dimensional coordinate map of the designated drug warehouse, and the location of the corresponding abnormal drug is marked.

[0017] Based on the drug distribution information in the designated drug warehouse, determine the adjacent shelf drugs of the abnormal drug, and generate a location relocation path planning instruction for the abnormal drug by combining the chemical characteristics of the adjacent shelf drugs.

[0018] The instructions for planning the relocation path of the cargo location are sent to the intelligent transfer equipment to drive the intelligent transfer equipment to transfer the abnormal medicines, and to rearrange the physical locations of the abnormal medicines to achieve designated medicine storage management.

[0019] In one implementation of this application, based on the drug distribution information in the designated drug warehouse, the adjacent shelf drugs of the abnormal drug are determined, and a location relocation path planning instruction for the abnormal drug is generated by combining the chemical characteristics of the adjacent shelf drugs. Specifically, this includes:

[0020] Based on the drug distribution information in the designated drug warehouse, determine the shelf to be moved, the target shelf, and the adjacent shelves corresponding to the abnormal drug, and determine the chemical characteristics of the drugs on the adjacent shelves; the target shelf is used to indicate the shelf when the abnormal drug meets the target storage conditions.

[0021] Retrieve the three-dimensional coordinate data corresponding to the shelf to be moved and the target shelf, and construct the passable path space of the intelligent transfer equipment in the designated drug warehouse based on the chemical characteristics of the medicines on the adjacent shelves;

[0022] The A* search algorithm is used, combined with the available path space, to calculate the smooth path corresponding to the intelligent transfer equipment, and deceleration buffer trajectories are inserted at turning nodes to generate the location relocation path planning instruction for the abnormal medicines; the smooth path is used to represent the path with the lowest energy consumption among all available paths.

[0023] In one implementation of this application, after collecting images of drugs and environmental perception data from a designated drug warehouse, the method further includes:

[0024] A specified drug shelf life decay model is constructed, and real-time environmental temperature and humidity fluctuation data and the initial shelf life of the drug are input into the specified drug shelf life decay model to calculate the remaining effective duration of the corresponding specified drug.

[0025] Based on the remaining valid time, the shelf position of the specified medicines whose remaining valid time is less than the preset remaining time threshold is adjusted to the priority outbound area, and an inventory warning is triggered simultaneously.

[0026] In one implementation of this application, it further includes:

[0027] Obtain historical consumption data, seasonal demand trends, and prescription volume changes for each specified drug in a designated drug warehouse, and train a time series prediction model based on the historical consumption data, the seasonal demand trends, and the prescription volume changes.

[0028] Based on the trained time-series prediction model, the estimated drug demand for future time periods is generated. Combined with the estimated drug demand and current inventory information, replenishment suggestions and shelf allocation strategies are generated.

[0029] In one implementation of this application, if the probability score of the abnormal drug status exceeds a preset threshold, after marking the location of the abnormal drug, the method further includes:

[0030] The verification results of abnormal drugs are obtained, and the verification results are input as labeled data into the multimodal fusion model to update the model parameters of the multimodal fusion model;

[0031] The anomaly detection accuracy of the updated multimodal fusion model is periodically verified at preset time intervals, and model parameter optimization is triggered when the anomaly detection accuracy is lower than a preset detection threshold.

[0032] In one implementation of this application, collecting drug images and environmental perception data in a designated drug warehouse specifically includes:

[0033] By periodically collecting temperature and humidity data from the drug storage area using pre-deployed embedded temperature and humidity sensors in a designated drug warehouse, a time series of temperature and humidity data can be obtained.

[0034] Real-time images of the drugs are acquired using optical sensors pre-deployed in the designated drug warehouse; the optical sensors consist of a high-resolution camera and a near-infrared spectrometer.

[0035] Based on the label on the drug packaging, obtain the drug batch metadata; the drug batch metadata includes the drug batch, production date, and expiration date of the specified drug to be identified.

[0036] On the other hand, embodiments of this application also provide an artificial intelligence-based pharmaceutical warehouse management device, the device comprising:

[0037] At least one processor;

[0038] And, a memory communicatively connected to the at least one processor;

[0039] The memory stores instructions that can be executed by the at least one processor, which are then executed by the at least one processor to enable the at least one processor to perform an artificial intelligence-based pharmaceutical warehouse management method as described above.

[0040] On the other hand, this application embodiment also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, implement the above-described artificial intelligence-based drug storage management method.

[0041] This application provides a method, equipment, and medium for pharmaceutical warehouse management based on artificial intelligence, which has at least the following beneficial effects:

[0042] By employing a multimodal data fusion and dynamic weight allocation mechanism, the drawbacks of isolated analysis of environmental parameters and drug attributes in traditional warehousing systems are effectively addressed. The periodic fluctuation characteristics of temperature and humidity time series are extracted using sliding window Fourier transform, and combined with the expiration date decay law in drug batch metadata, a dynamic modal weight matrix is ​​dynamically generated. This allows the anomaly probability score to accurately reflect the spatiotemporal correlation between environmental fluctuations and drug stability. In the anomaly handling stage, migration path planning instructions based on chemical compatibility constraints and equipment kinematic models solve the problems of secondary pollution and transportation vibration caused by manual intervention in traditional solutions. Through the generated smooth path combined with deceleration and buffer trajectories, along with real-time gyroscope attitude monitoring and air cushion suspension transportation mode, the vibration amplitude during drug handling is reduced to within a safe threshold, avoiding the risk of mixed storage of hazardous materials and ensuring the physicochemical stability of designated drugs. Attached Figure Description

[0043] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0044] Figure 1 A flowchart illustrating an artificial intelligence-based drug storage management method provided in this application embodiment;

[0045] Figure 2 A schematic diagram of the overall process of an artificial intelligence-based drug storage management method provided for embodiments of this application;

[0046] Figure 3 A schematic diagram of an artificial intelligence-based pharmaceutical warehouse management system provided for an embodiment of this application;

[0047] Figure 4 This is a schematic diagram of the internal structure of an artificial intelligence-based pharmaceutical storage management device provided in an embodiment of this application. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0050] Figure 1This is a flowchart illustrating an artificial intelligence-based drug storage management method provided in an embodiment of this application.

[0051] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0052] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0053] like Figure 1 As shown in the embodiment of this application, a drug warehousing management method based on artificial intelligence is provided, including:

[0054] Step 101: Collect images of drugs and environmental perception data in the designated drug warehouse, and analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental perception data includes temperature and humidity time series and drug batch metadata.

[0055] In this embodiment, the acquisition of drug images and environmental perception data is accomplished collaboratively through pre-deployed multi-source heterogeneous sensing devices. For example, drug image acquisition employs a combination of a high-resolution optical camera and a near-infrared spectrometer. The camera is deployed on a sliding pan-tilt unit at the top of the warehouse, allowing for multi-angle scanning of the shelves along a preset path. It should be noted that the near-infrared spectrometer penetrates the outer packaging of the drug to detect the integrity characteristics of the inner sealing film, for example, detecting whether the inner sealing film has wrinkles or breaks. It is understood that the temperature and humidity time series in the environmental perception data is generated by a three-dimensional, layered network of temperature and humidity sensors, with sensor nodes distributed according to shelf levels, forming a three-dimensional environmental field monitoring system.

[0056] Specifically, the analysis of drug images includes multi-scale feature extraction and rule matching. First, a convolutional neural network is used to separate the barcode area, label area, and damaged area on the outer surface of the drug packaging box. An OCR engine is then used to identify the drug name, batch number, and expiration date text information on the label. For stacking morphology analysis, an edge detection algorithm is used to calculate the tilt angle of the contact surface between adjacent packaging boxes, and a gravity simulation model is combined to predict stacking stability. It should be noted that the determination of the target storage conditions in this embodiment relies on a preset storage rule base. This rule base is associated with the chemical properties of the drug and storage specifications. Chemical properties include photosensitivity and thermal stability, while storage specifications include maximum stacking height and temperature and humidity thresholds. For example, for drugs requiring light-protected storage, the system detects the light-shielding performance of the packaging based on near-infrared spectral data and dynamically adjusts the lighting intensity limits of the permissible storage area.

[0057] It should be noted that the batch metadata of drugs is acquired in batches through RFID readers and verified in real time with the publicly available data on the National Drug Regulatory Code Platform to ensure the authenticity and consistency of data such as production date and expiration date.

[0058] In this embodiment, drug information in a drug image is identified to determine the drug name and type. Then, based on the drug name and type, target storage conditions such as the storage temperature range, humidity limit, and maximum stacking layer number corresponding to the drug are searched in a preset drug constraint database. It should be noted that this application achieves accurate analysis of drug label readability and stacking stability through the above method, solving the problems of label misidentification and frequent tipping accidents in the prior art.

[0059] Step 102: Input the environmental perception data into the pre-trained multimodal fusion model to extract cross-modal correlation features, and combine them with the target storage conditions to generate a drug status anomaly probability score for the specified drug to be identified.

[0060] In this embodiment, the multimodal fusion model employs a spatiotemporal attention mechanism to achieve dynamic weight allocation of data modalities. For example, in processing temperature and humidity time series, the time segments are first divided using a sliding window, and Fourier transform is used to extract the frequency domain features of each segment to identify abnormal fluctuation patterns, such as sudden changes in diurnal temperature range or periodic oscillations caused by equipment failure. It should be noted that the expiration date information in the drug batch metadata is combined with the current timestamp, and the drug degradation rate is calculated using the Arrhenius equation as a fundamental parameter for chemical stability assessment.

[0061] Specifically, the dynamic allocation of modal weights is achieved through a gating mechanism. First, based on the periodic fluctuation pattern, the deviation of the current fluctuation amplitude from the historical normal fluctuation range is calculated, and the production date and expiration date in the drug batch metadata are parsed to calculate the remaining expiration percentage of the specified drug to be identified, combined with the current timestamp. Then, the spectral deviation of temperature and humidity fluctuations is used as the first weight factor, and the remaining expiration percentage of the drug is used as the second weight factor. Both are normalized by the Sigmoid function to generate a weight matrix. Understandably, the spatiotemporal convolutional network extracts cross-modal correlation features under the guidance of this matrix, such as the coupling effect between high-frequency temperature and humidity fluctuations and drugs nearing their expiration date. Finally, the feature vector is compared with the target storage conditions for similarity, and an anomaly probability score in the 0-1 range is output. For example, when it is detected that the temperature and humidity fluctuation spectrum of a batch of drugs on the shelf differs significantly from the historical normal pattern, and its remaining expiration date is less than a preset percentage, the model will generate a high anomaly score.

[0062] It should be noted that the model training uses labeled historical warehouse data, including normal storage samples and various abnormal scenarios, such as damaged packaging and uncontrolled temperature and humidity. The network parameters are optimized through the cross-entropy loss function to ensure sensitivity to different risk levels.

[0063] In this embodiment, when calculating the anomaly probability score of the specified drug to be identified, the extracted cross-modal association feature vector is compared with the target storage conditions corresponding to the specified drug to be identified, and the cosine similarity between the two is obtained. This application also needs to determine the drug category to which the specified drug to be identified belongs. Drug categories include high-value drugs and ordinary drugs. It should be noted that in this embodiment, high-value drugs are used to represent specified drugs whose temperature and humidity deviations must be less than a first preset deviation threshold, and ordinary drugs are used to represent specified drugs whose temperature and humidity deviations can be greater than the first preset deviation threshold, but must be less than a second preset deviation threshold. The first preset deviation threshold is less than the second preset deviation threshold. High-value drugs are more sensitive to temperature and humidity deviations, and their medicinal properties are only unaffected within a small deviation range. Ordinary drugs are less sensitive to temperature and humidity deviations than high-value drugs; controlling the temperature and humidity deviations within the second preset deviation threshold will not affect the medicinal properties of ordinary drugs. Then, the anomaly probability score of the specified drug to be identified is calculated using the Sigmoid function, combined with the preset scaling factor corresponding to each drug category.

[0064] Through the above method, the embodiments of this application can automatically identify drug anomalies based on spectrum analysis, realize the accurate quantification of drug status risk, and solve the problem of neglecting drug sensitivity in the prior art.

[0065] In this embodiment, the model self-optimization mechanism continuously improves anomaly detection capabilities through manual review and feedback, as well as periodic verification. For example, the review results originate from on-site inspection reports submitted by warehouse managers via mobile terminals, including high-resolution re-images, measured environmental parameters, and a final status label (normal / abnormal). It should be noted that the system encrypts and de-identifies the review data during transmission, ensuring that drug privacy information (such as batch numbers and manufacturers) participates in calculations using hash encoding during training.

[0066] The drug images and environmental perception data collected during the review process are spatiotemporally matched with the original detection records to ensure consistency in input feature dimensions. An online learning algorithm is used to update the fully connected layer parameters of the multimodal fusion model, with a focus on optimizing the weight allocation logic in false alarm / missed alarm cases. For example, if manual review confirms that a high-temperature alarm was a false alarm caused by a sensor malfunction, the model will reduce the confidence weight of the temperature modality in that scenario and increase the focus of the image modality on device status (such as sensor damage marks). Then, each updated model is generated with an independent version number and stored in the model repository along with the training dataset and hyperparameter configurations, supporting rollback to historical stable versions within a time range.

[0067] Step 103: If the probability score of abnormal drug status exceeds the preset threshold, mark the location of the abnormal drug and generate a corresponding adjustment instruction based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction and realize the designated drug warehouse management.

[0068] In this embodiment, when the probability score of an abnormal drug status exceeds a preset threshold, the system triggers a tiered emergency response mechanism. For example, the preset threshold is a dynamically adjusted value, with different triggering criteria set according to the drug's hazard level, such as highly toxic drugs, anesthetics, or cold chain drugs. It should be noted that the threshold dynamic adjustment module calculates the threshold by querying the risk coefficient in the Drug Safety Data Sheet (SDS) and combining it with the overall environmental status of the current warehouse, such as whether it is the rainy season.

[0069] A 3D coordinate map corresponding to a specified drug warehouse is constructed using laser scanning point cloud data. The shelf numbers associated with anomaly probability scores are spatially mapped to this 3D coordinate map, and the abnormal shelf areas are highlighted in red on the visualization interface. It's understood that the drug distribution information database stores the molecular structure formula, chemical stability parameters, and incompatible drug lists for each shelf unit. For example, for a detected abnormal biological agent shelf, the system automatically searches all shelves within a specified radius for drugs and identifies adjacent chemicals that may cause protein denaturation through comparison.

[0070] When generating storage location relocation path planning instructions, high-risk target shelves are first excluded based on a chemical compatibility rule engine, and then path search is performed by combining the kinematic parameters of the intelligent transfer equipment. For example, the improved A* algorithm introduces equipment-specific constraints on top of traditional path costs (such as distance and number of turns), converts the maximum extension radius of the handling robot arm into the effective passage width of the path corridor, sets path smoothness weights according to the drug's vibration sensitivity level, inserts Bézier curves at curvature abrupt change points, and applies temperature field constraints to the cold chain drug relocation path, allowing passage only through areas where cryogenic maintenance equipment has been deployed.

[0071] It should be noted that the route planning module receives real-time environmental monitoring data. When a sudden change in temperature or humidity is detected on the target migration path, such as in an area with air conditioning failure, the system automatically triggers route replanning. For example, if a batch of vaccines needs to be migrated to a low-temperature zone, and the original planned route passes through a temporarily heated passage, the system will dynamically adjust the route to a backup low-temperature corridor.

[0072] After the migration command is sent to the intelligent transfer device, the system uses a distributed communication protocol, such as ROS2, to achieve collaborative control of multiple devices.

[0073] In this embodiment, if the anomaly probability score exceeds a preset threshold, it indicates an anomaly in the storage of the specified drug. At this point, adjustments can be made to the target storage conditions or the storage conditions at the current storage location for the specified drug, based on the actual situation. For example, if, based on the batch metadata of the specified drug, it is determined that the humidity at the storage location does not meet the humidity requirements of the target storage conditions, and considering the drug's characteristics, if humidity has no impact on the drug's properties or quality, then the humidity requirements in the current drug storage conditions can be adjusted using machine processing to restore normalcy after the anomaly is resolved. If humidity does affect the drug's properties or quality, manual processing can be used, for example, by adding humidity-enhancing equipment, such as a large humidifier, to adjust the humidity data at the storage location of the specified drug, thus restoring normalcy after the anomaly is resolved.

[0074] In this embodiment, the construction of a drug shelf-life decay model and the early warning mechanism constitute the core module of drug shelf-life management. For example, the shelf-life decay model is developed based on the framework of chemical kinetics theory and establishes a quantitative relationship between temperature and humidity accelerating factors and the drug degradation rate. It should be noted that the initial shelf life of the drug is extracted from batch metadata and used together with real-time environmental temperature and humidity fluctuation data as model input parameters.

[0075] When the remaining validity period of a batch of medicines is calculated to be less than a preset remaining validity period threshold, the warehouse layout optimization engine is invoked to move the target medicine shelf to the priority outbound area, typically near the sorting conveyor entrance. It should be noted that the shelves in the priority outbound area are equipped with RFID enhanced reader / writer devices to ensure improved outbound scanning efficiency.

[0076] In one embodiment, the warning level is set according to the difference between the remaining time and the threshold. For example, a level 1 warning pushes a pop-up alarm to the warehouse manager's mobile terminal and automatically generates a priority outbound task order. A level 2 warning highlights the relevant shelves in the visual dashboard of the warehouse management system and simultaneously triggers the automatic replenishment protocol of the supply chain system.

[0077] For example, the setting of the preset remaining time threshold follows the principle of dynamic adjustment: for cold chain pharmaceuticals, the threshold is adaptively shortened based on the historical failure rate of refrigeration equipment during transportation; for high-value biological agents, the threshold is dynamically extended in conjunction with the hospital prescription volume prediction model to optimize inventory turnover.

[0078] In this embodiment, historical consumption data is obtained by integrating multiple sources, including warehouse management system outbound records, hospital prescription system interfaces, and medical insurance settlement data. It should be noted that the historical consumption data includes structured fields such as drug specifications, distribution quantity, and receiving department. Seasonal demand trends are extracted using a time series decomposition algorithm, and periodic fluctuation components are separated using the STL method, such as the surge in demand for antiviral drugs during flu season.

[0079] Specifically, missing values ​​are imputed and outliers are corrected in historical consumption data, and continuous time segment samples are generated using the sliding window method. Understandably, prescription volume change data is analyzed using natural language processing technology to extract diagnostic keywords from electronic medical records, such as "malignant tumor" and "postoperative anticoagulation," to establish a correlation map between disease spectrum and drug consumption.

[0080] Before constructing the association graph, this application first extracts diagnostic text from the hospital's electronic medical record (EMR) system and obtains the corresponding patient's medication dispensing record from the prescription system, as well as the medication consumption and timestamp from the associated medical insurance settlement data. It should be noted that the diagnostic text is, for example, an acute exacerbation of bronchial asthma, and the medication dispensing record is, for example, salbutamol inhaler, twice daily. Then, the diagnostic text is parsed using a natural language processing model, keyword recognition is performed, and the disease name, symptoms, and severity are extracted. The disease name is, for example, asthma; the symptoms are, for example, dyspnea; and the severity is, for example, acute. Furthermore, the medication names are standardized, mapping brand names to corresponding generic names, for example, mapping Ventolin to salbutamol. In addition, this application also associates the prescription medication consumption of the same patient within the next 7 days based on the diagnosis date, generating a time-series sequence corresponding to diagnostic events and medication consumption.

[0081] When constructing the association graph, disease nodes and drug nodes are generated. Disease nodes include the disease code and corresponding disease name, while drug nodes include the generic name of the drug and its corresponding dosage form, such as aerosol or spray. Historical cases are obtained, and the support of diseases A and B is calculated based on the number of cases of disease A, the number of cases of disease B, and the total number of cases in the historical cases. The support of disease A and the support of disease B are calculated using the same method. Based on the support of disease A and disease B, and the support of disease A and disease B, the lift of disease A and disease B can be calculated. Combined with a preset seasonality factor and lift, the final edge weight between disease A and disease B is generated. Using the above method, the edge weight corresponding to each edge in the association graph is calculated, generating an association graph between disease and drug consumption. This association graph is then stored in the database in the form of an attribute graph. It should be noted that node attributes include disease epidemic season indicators and drug temperature sensitivity, etc.

[0082] This application constructs an association graph that can update the edge weights of each edge in real time based on real-time data, thereby linking temperature and humidity data with the remaining effective duration of the medicine, promptly identifying problems and issuing early warnings, thus solving the problem of prediction lag in the prior art.

[0083] Then, heterogeneous data such as historical consumption, seasonality index, prescription growth rate, and weather forecast data are normalized into vector sequences with unified timestamps. For example, for cold chain pharmaceuticals, regional logistics disruption early warning data is additionally introduced as a negative influencing factor. An LSTM network improved with a multi-head attention mechanism is used to dynamically adjust the importance weights of different features.

[0084] The system automatically adjusts the safety stock threshold based on the predicted demand volatility. For medicines with high demand variance (such as emergency medicines), Brown's double exponential smoothing method is used to dynamically expand the buffer inventory capacity. The predicted demand is compared with the current inventory and the number of orders in transit in real time. When the available inventory is lower than (predicted demand × safety factor), a tiered replenishment instruction is triggered.

[0085] Figure 2 This is a flowchart illustrating another artificial intelligence-based drug storage management method provided in an embodiment of this application. Figure 2 As shown, step 201 involves collecting images of medicines and environmental perception data from the warehouse.

[0086] Step 202: Using the collected drug images as input data, artificial intelligence technology is used to analyze the drug images, identify the drug information in the images, and thus determine what drug the drug image corresponds to, and what the target storage conditions are for the specified drug, such as certain specified drugs needing to be stored under specified temperature, humidity or conditions.

[0087] Step 203: Input the environmental perception data as input data into the pre-trained multimodal fusion model to extract cross-modal correlation features.

[0088] Step 204: Based on the environmental conditions of the drug in the current environmental perception data in the associated features, and combined with the storage conditions required for the currently specified drug in the determined target storage conditions, determine the anomaly probability score of the currently specified drug having an abnormal storage location.

[0089] Step 205: Determine whether the anomaly probability score exceeds the preset threshold.

[0090] Step 206: If the detected anomaly probability score does not exceed the preset threshold, it means that the storage environment of the current specified drug is within the target storage condition range and will not affect the drug's properties or quality. At this time, it indicates that the storage location of the current specified drug meets the requirements and there is no need to move the storage location of the current specified drug.

[0091] Step 207: If the detected anomaly probability score exceeds a preset threshold, it indicates that the storage environment of the currently specified drug does not meet the corresponding target storage condition range, which may affect the drug's properties or quality. Therefore, the location of the currently specified drug needs to be adjusted. At this point, the current abnormal location of the drug is marked. Based on the drug storage conditions on the surrounding shelves and the target storage location in the target storage conditions, a corresponding location adjustment instruction is generated for the currently specified drug.

[0092] Step 208: Based on the execution of the adjustment instruction, complete the transfer of the currently specified drug from the abnormal location to the target storage location, thereby realizing the storage management of the specified drug.

[0093] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide an artificial intelligence-based pharmaceutical warehouse management system, the structure of which is as follows: Figure 3 As shown.

[0094] Figure 3 This is a schematic diagram of an artificial intelligence-based pharmaceutical warehouse management system provided as an embodiment of this application. Figure 3 As shown, system 300 includes: data acquisition module 301, score generation module 302, and drug adjustment module 303.

[0095] The data acquisition module 301 is used to collect images of drugs and environmental perception data in a designated drug warehouse, and to analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental perception data includes temperature and humidity time series and drug batch metadata.

[0096] The scoring generation module 302 is used to input environmental perception data into a pre-trained multimodal fusion model to extract cross-modal correlation features and, in combination with the target storage conditions, generate a drug status anomaly probability score for the specified drug to be identified.

[0097] The drug adjustment module 303 is used to mark the location of abnormal drugs if the probability score of abnormal drug status exceeds a preset threshold, and generate corresponding adjustment instructions based on the drug distribution information in the designated drug warehouse, so as to control the intelligent transfer equipment to execute the adjustment instructions and realize the designated drug storage management.

[0098] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide an artificial intelligence-based pharmaceutical storage management device, the structure of which is as follows: Figure 4 As shown.

[0099] Figure 4 This is a schematic diagram of the internal structure of an artificial intelligence-based pharmaceutical storage management device provided as an embodiment of this application. Figure 4 As shown, the device includes:

[0100] At least one processor;

[0101] And, a memory that is communicatively connected to at least one processor;

[0102] The memory stores instructions that can be executed by at least one processor, and the instructions, when executed by at least one processor, enable at least one processor to:

[0103] Collect images of drugs and environmental sensing data in a designated drug warehouse, analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental sensing data includes temperature and humidity time series and drug batch metadata;

[0104] Environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, and combined with the target storage conditions, a drug status anomaly probability score is generated for the specified drug to be identified.

[0105] If the probability score of abnormal drug status exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction and realize the warehouse management of the designated drug.

[0106] This application also provides a non-volatile computer storage medium storing computer-executable instructions, which, when executed, can:

[0107] Collect images of drugs and environmental sensing data in a designated drug warehouse, analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental sensing data includes temperature and humidity time series and drug batch metadata;

[0108] Environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, and combined with the target storage conditions, a drug status anomaly probability score is generated for the specified drug to be identified.

[0109] If the probability score of abnormal drug status exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction and realize the warehouse management of the designated drug.

[0110] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0111] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0112] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0113] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0114] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0115] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0116] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0117] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0118] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0119] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A pharmaceutical warehouse management method based on artificial intelligence, characterized in that, The method includes: Collect images of drugs and environmental sensing data in a designated drug warehouse, and analyze the drug images to determine the target storage conditions for the designated drug to be identified; the environmental sensing data includes temperature and humidity time series and drug batch metadata; The environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, and combined with the target storage conditions, a drug status anomaly probability score for the specified drug to be identified is generated. If the probability score of the abnormal status of the drug exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse to control the intelligent transfer equipment to execute the adjustment instruction and realize the designated drug warehouse management. The environmental perception data is input into a pre-trained multimodal fusion model to extract cross-modal correlation features, specifically including: The environmental perception data is input into a pre-trained multimodal fusion model, and a sliding window Fourier transform is performed on the temperature and humidity time series to detect periodic fluctuation patterns. Based on the aforementioned periodic fluctuation pattern, the degree of deviation between the current fluctuation amplitude and the historical normal fluctuation range is calculated, and the degree of deviation is used as the first weighting factor. The production date and expiration date in the drug batch metadata are parsed, and the remaining expiration date percentage of the specified drug to be identified is calculated in combination with the current timestamp, so that the remaining expiration date percentage is used as the second weighting factor; By fusing the first weighting factor and the second weighting factor, a modal dynamic weighting matrix is ​​generated, and based on the modal dynamic weighting matrix, the confidence level of different data modalities in the time dimension is calculated; Based on the confidence differences of different data modalities, modality weights are dynamically allocated, and a spatiotemporal convolutional network is used to extract cross-modal association features.

2. The method for pharmaceutical warehouse management based on artificial intelligence according to claim 1, characterized in that, If the probability score of the abnormal drug status exceeds a preset threshold, the location of the abnormal drug is marked, and a corresponding adjustment instruction is generated based on the drug distribution information in the designated drug warehouse. This adjustment instruction is then executed by the intelligent transfer equipment to achieve designated drug warehouse management, specifically including: The abnormal drug status probability score is compared with a preset threshold. If the abnormal drug status probability score exceeds the preset threshold, the shelf number of the abnormal drug is located based on the three-dimensional coordinate map of the designated drug warehouse, and the location of the corresponding abnormal drug is marked. Based on the drug distribution information in the designated drug warehouse, determine the adjacent shelf drugs of the abnormal drug, and generate a location relocation path planning instruction for the abnormal drug by combining the chemical characteristics of the adjacent shelf drugs. The instructions for planning the relocation path of the cargo location are sent to the intelligent transfer equipment to drive the intelligent transfer equipment to transfer the abnormal medicines, and to rearrange the physical locations of the abnormal medicines to achieve designated medicine storage management.

3. The artificial intelligence-based pharmaceutical warehouse management method according to claim 2, characterized in that, Based on the drug distribution information in the designated drug warehouse, the adjacent shelf drugs of the abnormal drug are determined. Then, combined with the chemical characteristics of the adjacent shelf drugs, a location relocation path planning instruction for the abnormal drug is generated, specifically including: Based on the drug distribution information in the designated drug warehouse, determine the shelf to be moved, the target shelf, and the adjacent shelves corresponding to the abnormal drug, and determine the chemical characteristics of the drugs on the adjacent shelves; the target shelf is used to indicate the shelf when the abnormal drug meets the target storage conditions. Retrieve the three-dimensional coordinate data corresponding to the shelf to be moved and the target shelf, and construct the passable path space of the intelligent transfer equipment in the designated drug warehouse based on the chemical characteristics of the medicines on the adjacent shelves; The A* search algorithm is used, combined with the available path space, to calculate the smooth path corresponding to the intelligent transfer equipment, and deceleration buffer trajectories are inserted at turning nodes to generate the location relocation path planning instruction for the abnormal medicines; the smooth path is used to represent the path with the lowest energy consumption among all available paths.

4. The artificial intelligence-based pharmaceutical warehouse management method according to claim 1, characterized in that, After collecting images of drugs and environmental perception data from a designated drug warehouse, the method further includes: A specified drug shelf life decay model is constructed, and real-time environmental temperature and humidity fluctuation data and the initial shelf life of the drug are input into the specified drug shelf life decay model to calculate the remaining effective duration of the corresponding specified drug. Based on the remaining valid time, the shelf position of the specified medicines whose remaining valid time is less than the preset remaining time threshold is adjusted to the priority outbound area, and an inventory warning is triggered simultaneously.

5. The drug warehousing management method based on artificial intelligence according to claim 1, characterized in that, The method further includes: Obtain historical consumption data, seasonal demand trends, and prescription volume changes for each specified drug in a designated drug warehouse, and train a time series prediction model based on the historical consumption data, the seasonal demand trends, and the prescription volume changes. Based on the trained time-series prediction model, the estimated drug demand for future time periods is generated. Combined with the estimated drug demand and current inventory information, replenishment suggestions and shelf allocation strategies are generated.

6. The pharmaceutical warehousing management method based on artificial intelligence according to claim 1, characterized in that, If the probability score of the abnormal drug status exceeds a preset threshold, after marking the location of the abnormal drug, the method further includes: The verification results of abnormal drugs are obtained, and the verification results are input as labeled data into the multimodal fusion model to update the model parameters of the multimodal fusion model; The anomaly detection accuracy of the updated multimodal fusion model is periodically verified at preset time intervals, and model parameter optimization is triggered when the anomaly detection accuracy is lower than a preset detection threshold.

7. The artificial intelligence-based pharmaceutical warehouse management method according to claim 1, characterized in that, Collect images of medicines and environmental perception data in a designated drug warehouse, specifically including: By periodically collecting temperature and humidity data from the drug storage area using pre-deployed embedded temperature and humidity sensors in a designated drug warehouse, a time series of temperature and humidity data can be obtained. Real-time images of the drugs are acquired using optical sensors pre-deployed in the designated drug warehouse; the optical sensors consist of a high-resolution camera and a near-infrared spectrometer. Based on the label on the drug packaging, obtain the drug batch metadata; the drug batch metadata includes the drug batch, production date, and expiration date of the specified drug to be identified.

8. A pharmaceutical warehouse management device based on artificial intelligence, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform an artificial intelligence-based pharmaceutical warehouse management method as described in any one of claims 1-7.

9. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement an artificial intelligence-based drug storage management method as described in any one of claims 1-7.

Images