RFID-based medical waste recycling dispensing system

Abstract

The application discloses an RFID-based medical waste recycling and distribution system and relates to the technical field of intelligent management of medical waste.The system comprises a container identification module, which stores a unique identification in a waste container and a clean container by means of an RFID tag; a data acquisition module, which acquires tag information through a read-write device and an operation terminal; a correlation binding module, which establishes a corresponding binding between waste recycling and clean distribution information; a trajectory tracking module, which tracks the position and state of the container in real time; an intelligent analysis module, which adopts an improved K nearest neighbor algorithm, introduces trajectory segment similarity measurement and dynamic time warping distance, and fuses the characteristics of the length of stay and the type of operation to analyze the trajectory; and a deployment optimization module, which generates deployment instructions according to the algorithm analysis results and the correlation relationship.The system realizes accurate analysis of the trajectory of the container, identification of abnormal flow and linkage deployment of the waste and clean containers, and optimizes the flow and configuration of the medical containers.

Description

Technical Field

[0001] This invention belongs to the field of intelligent medical waste management technology, specifically an RFID-based medical waste recycling and distribution system. Background Technology

[0002] Current management of medical waste and clean supplies largely relies on independent manual registration. Basic identification is achieved by reading container tags using RFID devices. Waste collection containers and clean supply dispensing containers are managed separately, and container trajectories are recorded solely through fixed-point RFID readings. Trajectory analysis often employs the conventional K-nearest neighbor algorithm, using only spatial coordinates as features and relying on Euclidean distance for similarity calculations. Container allocation depends on manual experience, lacking any data-driven basis.

[0003] Conventional RFID management systems suffer from limited trajectory feature dimensions, and their distance measurement methods cannot adapt to trajectory time shifts and local feature differences. This results in low accuracy in trajectory similarity determination, difficulty in accurately predicting container turnover efficiency, and inability to promptly identify turnover anomalies. Furthermore, the processes for waste collection and clean material distribution are independent, with no corresponding correlation between the two types of container information, and a lack of linked data support for allocation operations.

[0004] The trajectory analysis algorithm needs improvement by introducing trajectory segment similarity metrics and dynamic time warping distance, and integrating container dwell time and operation type into the feature dimension. A corresponding binding relationship between waste collection and clean material distribution needs to be established, and container allocation instructions should be generated based on trajectory analysis results to optimize container flow configuration. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art;

[0006] Therefore, this invention proposes an RFID-based medical waste collection and distribution system, comprising:

[0007] The container identification module fixes RFID tags to the surfaces of medical waste recycling containers and medical clean supply containers respectively, and the container RFID tags store a unique container identifier.

[0008] The data acquisition module deploys RFID reading and writing devices at medical waste collection points and medical clean product distribution points. The RFID reading and writing devices read information from the RFID tags on the containers. A recycling operation terminal is configured at the medical waste collection point, and a distribution operation terminal is configured at the medical clean product distribution point. The recycling operation terminal and the distribution operation terminal acquire the data read by the RFID reading and writing devices.

[0009] The association and binding module establishes a correspondence between medical waste collection and clean material distribution, and associates and binds the information of collected medical waste with the information of clean materials to be distributed.

[0010] The trajectory tracking module continuously tracks the location and status of the medical waste recycling container and the medical hygiene product dispensing container through a deployed RFID reader / writer.

[0011] The intelligent analysis module uses an improved K-nearest neighbor algorithm to analyze the movement trajectory of containers in order to predict container turnover efficiency and identify turnover anomalies. The improved K-nearest neighbor algorithm introduces trajectory segment similarity measurement and dynamic time warping distance, and integrates container dwell time and operation type as feature dimensions.

[0012] The allocation and optimization module generates container allocation instructions based on the analysis results of the improved K-nearest neighbor algorithm and the correspondence between medical waste recycling and clean material distribution, in order to optimize the flow and configuration of containers.

[0013] Furthermore, establishing the correspondence between medical waste collection and clean material distribution includes:

[0014] When the RFID reader reads the unique identifier of a medical waste recycling container at a medical waste recycling point, it marks the medical waste recycling container as "recycled" and records the recycling time and recycling point location information.

[0015] Within a preset association time window, when the RFID reader reads the unique container identifier of any medical waste dispensing container at any medical waste dispensing point, it selects the most recently recycled container from the list of "recycled" medical waste recycling containers to be associated.

[0016] In the background database, a relationship record is created to permanently bind the unique identifier of the selected medical waste recycling container, the recycling time, and the recycling point location information with the unique identifier of the currently read medical clean item distribution container, the distribution time, and the distribution point location information, forming a one-to-one correspondence between medical waste recycling and clean item distribution.

[0017] Furthermore, the continuous tracking of the location and status of the medical waste collection container and the medical hygiene product dispensing container via the deployed RFID reader / writer includes:

[0018] Multiple fixed RFID readers and writers are deployed at key path nodes in medical institutions, including corridor intersections, elevator lobbies, and floor entrances, forming an RFID monitoring network covering the main circulation paths.

[0019] When a medical waste recycling container or medical hygiene product dispensing container with a container RFID tag enters the reading range of the fixed RFID reader, the fixed RFID reader automatically reads its unique container identifier and records the reading time and its own device location code.

[0020] Each reading event is encapsulated as a container trajectory point data, which includes at least a unique container identifier, reading time, and device location code.

[0021] All container trajectory point data belonging to the same unique container identifier are arranged in chronological order of recognition time to generate a continuous movement trajectory sequence of the container. The movement trajectory sequence reflects the real-time position and historical path of the container.

[0022] Furthermore, the improved K-nearest neighbor algorithm analyzes the container movement trajectory to predict container turnover efficiency and identify turnover anomalies, including:

[0023] Extract the movement trajectory sequence of containers that have completed a full circulation cycle from historical data to form a historical trajectory sample library. Each historical trajectory sample carries its final circulation efficiency label or anomaly type label.

[0024] For the target container that is currently in circulation, extract its latest partial movement trajectory sequence starting from the recycling point as the query trajectory to be analyzed.

[0025] The comprehensive similarity between the queried trajectory and each historical trajectory sample in the historical trajectory sample library is calculated. The comprehensive similarity is obtained by weighted fusion of trajectory segment similarity measure, shape similarity based on dynamic time warping distance, and cosine similarity of dwell time and operation type feature vector.

[0026] In the historical trajectory sample database, select the K historical trajectory samples with the highest comprehensive similarity to the queried trajectory;

[0027] Based on the flow efficiency labels carried by these K historical trajectory samples, the expected flow efficiency of the target container is predicted by weighted averaging; or based on the anomaly type labels carried by these K historical trajectory samples, a voting mechanism is used to determine whether the target container is currently experiencing a flow anomaly and its anomaly type.

[0028] Furthermore, the improved K-nearest neighbor algorithm introduces a trajectory segment similarity metric and a dynamic time warping distance, and its working principle includes:

[0029] The movement trajectory sequence is divided into several trajectory segments according to the locations of the fixed RFID reading and writing devices it passes through. Each trajectory segment is identified by its start and end points.

[0030] For a query trajectory and a historical trajectory sample, compare the sequence of continuous trajectory segments starting from the starting point of the query trajectory and the historical trajectory sample, and count the number of trajectory segments with the same starting and ending positions. The ratio of the number of trajectory segments to the total number of segments is defined as the trajectory segment similarity.

[0031] The dynamic time warping method is applied to align the time series of the query trajectory with those of historical trajectory samples. The time series are composed of the time intervals between adjacent trajectory points. The dynamic time warping method allows the sequences to be non-linearly scaled and aligned on the time axis to find the minimum cumulative distance between the two time series. The minimum cumulative distance is the dynamic time warping distance, which is used to measure the similarity of trajectory shapes.

[0032] When calculating the overall similarity, weight coefficients are assigned to the trajectory segment similarity, the inverse transformation value of the dynamic time warping distance, and the cosine similarity between the dwell time and the operation type feature vector, and the weighted sum is used to obtain the final overall similarity.

[0033] Furthermore, based on the analysis results of the improved K-nearest neighbor algorithm and combined with the correspondence between medical waste recycling and clean material distribution, a container allocation instruction is generated, including:

[0034] Receive the predicted flow efficiency or flow anomaly judgment result of the target container output by the improved K nearest neighbor algorithm;

[0035] If the analysis result indicates that the predicted circulation efficiency is lower than the historical average, the efficiency optimization and allocation process is triggered to query the correspondence between the medical waste recycling and the clean material distribution, and to find the location of the clean material container that has been distributed and corresponds to the target container.

[0036] Generate a preset instruction to deliver spare cleaning containers in advance to the area where the already issued cleaning containers are located, in order to prevent a shortage of cleaning supplies due to delays in the circulation of the target containers;

[0037] If the analysis result indicates that the target container has experienced a flow abnormality, the abnormality handling and allocation process is triggered to determine the last known location of the target container and generate a scheduling instruction to assign the nearest recycling personnel to the last known location for inspection and handling, and to allocate a temporary medical waste recycling container to the clinical department corresponding to the medical waste recycling container that has experienced the flow abnormality.

[0038] Furthermore, within a preset association time window, when the RFID reader / writer reads the unique identifier of any medical waste dispensing container at any medical waste dispensing point, it selects the most recently recycled container from the list of "recycled" medical waste recycling containers to be associated, including:

[0039] A preset duration threshold for the associated time window is established; a timer is started to record the elapsed time since a medical waste recycling container is marked as "recycled" at a recycling point; before the elapsed time reaches the duration threshold, the medical waste recycling container is continuously retained in the list of "recycled" medical waste recycling containers to be associated; when the unique container identifier of any medical waste dispensing container is read at any medical waste dispensing point, the list of "recycled" medical waste recycling containers to be associated is immediately queried, and all medical waste recycling containers that have not yet established a binding relationship with other medical waste dispensing containers are selected from the list; among the selected containers, the recycling times of each container are compared, and the medical waste recycling container with the latest recycling time, i.e., the closest to the current time, is selected as the most recently recycled container.

[0040] Further, in calculating the comprehensive similarity, weight coefficients are assigned to the trajectory segment similarity, the inverse transformed value of the dynamic time warping distance, and the cosine similarity between the dwell time and the operation type feature vector, respectively, and the final comprehensive similarity is obtained by weighted summation, including:

[0041] A first weight coefficient is assigned to the trajectory segment similarity, a second weight coefficient is assigned to the inverse transformation value of the dynamic time warp distance, and a third weight coefficient is assigned to the cosine similarity between the dwell time and the operation type feature vector, wherein the sum of the first weight coefficient, the second weight coefficient and the third weight coefficient is a fixed value;

[0042] The shape similarity is obtained by performing a reciprocal transformation on the dynamic time warp distance. The transformation method is to divide 1 by the sum of the dynamic time warp distance and a preset constant, so that the shape similarity value is normalized to the range of 0 to 1.

[0043] The weighted sum of the trajectory segment similarity, the normalized shape similarity, and the cosine similarity is calculated. The weighted sum is calculated by multiplying the trajectory segment similarity by the first weight coefficient to obtain a first product, multiplying the normalized shape similarity by the second weight coefficient to obtain a second product, multiplying the cosine similarity by the third weight coefficient to obtain a third product, and then adding the first product, the second product, and the third product to obtain the comprehensive similarity.

[0044] Furthermore, based on the analysis results of the improved K-nearest neighbor algorithm and combined with the correspondence between medical waste recycling and clean material distribution, a container allocation instruction is generated, including an efficiency-optimized allocation process, specifically as follows:

[0045] When the analysis result of the improved K-nearest neighbor algorithm shows that the predicted turnover efficiency of the target container is lower than the historical average, the system automatically triggers the efficiency optimization and allocation process.

[0046] After the efficiency optimization and allocation process is initiated, the system retrieves the unique container identifier of the medical clean item distribution container that is permanently bound to the unique container identifier of the target medical waste recycling container with low circulation efficiency in the background database, based on the established correspondence between medical waste recycling and clean item distribution.

[0047] The system queries the real-time location of the medical supplies dispensing container based on the unique container identifier of the bound medical supplies dispensing container. The real-time location is determined by the latest reading record of the RFID reader / writer deployed at the medical supplies dispensing point or the fixed RFID reader / writer on the critical path node.

[0048] The system retrieves a spare cleaning container from the cleaning container inventory that is currently in "awaiting" status and whose physical location is closest to the area where the queried issued cleaning container is located.

[0049] The system generates a preset instruction, which includes the unique identifier of the selected backup cleaning container, the location information of the target delivery area (i.e., the area where the cleaning container has been distributed), and the preset execution time limit;

[0050] The system sends the preset instructions to the mobile operating terminal of the cleaning supplies dispatch personnel to drive them to perform the advance delivery of spare cleaning supplies containers, thereby preventing a shortage of clinical cleaning supplies due to delays in the circulation of the target medical waste recycling containers.

[0051] Furthermore, the system also includes a data cleaning module for cleaning and completing the RFID monitoring network's read data, specifically:

[0052] Receive raw read event streams from all fixed RFID reader / writer devices, the raw read event streams containing container unique identifiers, read times, and device location codes;

[0053] The original reading event stream is deduplicated within a time window. When the same container unique identifier is on the same device location code and multiple reading events are generated within a very short time threshold, only the first reading event is retained as a valid event.

[0054] The trajectory rationality test is performed on the effective event sequence after deduplication. Based on the position and distance of adjacent fixed RFID reading and writing devices, combined with the preset reasonable moving speed of the container, it is determined whether the continuity of two adjacent effective events in time and space meets the rationality threshold.

[0055] For trajectory discontinuities that do not meet the reasonableness threshold, one or more inferred virtual trajectory points are inserted in the background logic based on the reasonable movement speed and path topology of the container. The virtual trajectory points include the inferred time and location range to generate a smoother and more continuous movement trajectory sequence.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] Trajectory segment similarity measurement can segment and decompose the complete container movement trajectory, enabling fine-grained comparison of local trajectory features and improving the detail of trajectory feature matching. Dynamic time warping distance can adapt to the nonlinear offset of trajectory time series, reduce the interference of time axis misalignment on trajectory similarity calculation, and improve the adaptability of trajectory comparison across different time series. Integrating container dwell time and operation type as feature dimensions into the algorithm calculation enriches the constituent elements of trajectory features, allowing trajectory representation to simultaneously include multiple parameters such as spatial location, temporal attributes, and operational behavior. The trajectory similarity discrimination criteria are more comprehensive, significantly improving the accuracy of trajectory clustering and comparison. It can accurately distinguish between normal and abnormal flow trajectories, truly reflecting the differences in container flow efficiency, and accurately identifying abnormal states such as stagnation and deviation from the path during the flow process. The trajectory analysis results maintain a high degree of consistency with the actual container flow status.

[0058] A corresponding binding relationship is established between medical waste collection information and clean material distribution information, forming a linkage between the two types of container circulation data. Based on the trajectory analysis results of the improved K-nearest neighbor algorithm, and combined with the association data between waste and clean materials, container allocation instructions are generated. Allocation decisions are based on both trajectory analysis data and process-related data. Container allocation is expanded from single-location information to multi-dimensional circulation data, improving the matching degree between allocation operations and on-site collection and distribution needs, and enhancing the continuity of container circulation. Container configuration and distribution can fit the actual usage needs of each collection and distribution point, alleviating the problems of uneven container distribution and delayed circulation connections. The overall container circulation process is stable and orderly, the rational utilization of allocation resources is improved, and the container circulation and configuration status remains stable. Attached Figure Description

[0059] Figure 1 This is a timing diagram of the RFID-based medical waste collection and distribution system described in this invention.

[0060] Figure 2 A flowchart for acquiring sensor data frames from a laser profilometer device;

[0061] Figure 3 This is a flowchart for performing a cleaning and filtering operation on a sequence of state points. Detailed Implementation

[0062] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] See Figure 1 This invention provides an RFID-based medical waste collection and distribution system, specifically comprising:

[0064] The system comprises a container identification module, a data acquisition module, an association and binding module, a trajectory tracking module, an intelligent analysis module, and a dispatch optimization module. RFID tags are fixedly affixed to the surfaces of medical waste collection containers and medical clean item dispensing containers, each tag storing a unique identifier for the container. RFID readers are deployed at both medical waste collection and medical clean item dispensing points, and the collection and dispensing terminals acquire data read by these RFID readers. The system establishes a correspondence between medical waste collection and clean item dispensing, associating information on collected medical waste with information on clean items to be dispensed. The deployed RFID readers continuously track the location and status of the medical waste collection and medical clean item dispensing containers. An improved K-nearest neighbor algorithm is used to analyze the container movement trajectory. This algorithm incorporates trajectory segment similarity measurement and dynamic time warping distance, and integrates container dwell time and operation type as feature dimensions to predict container turnover efficiency and identify turnover anomalies. Based on the analysis results of the improved K-nearest neighbor algorithm and combined with the correspondence between medical waste collection and clean item dispensing, container dispatch instructions are generated to optimize container turnover and configuration.

[0065] In one embodiment of the present invention, see [reference] Figure 2 When an RFID reader reads the unique identifier of a medical waste collection container at a medical waste collection point, it marks the container as "recycled" and records the collection time and collection point location information. Within a preset association time window, when the RFID reader reads the unique identifier of any medical waste dispensing container at any medical waste dispensing point, it selects the most recently recycled container from the list of "recycled" medical waste collection containers to be associated. In the background database, a relationship record is created, permanently binding the unique identifier, collection time, and collection point location information of the selected medical waste collection container with the unique identifier, dispensing time, and dispensing point location information of the currently read medical waste dispensing container, forming a one-to-one correspondence between medical waste recycling and medical waste dispensing.

[0066] Multiple fixed RFID readers are deployed at key path nodes in medical institutions, such as corridor intersections, elevator lobbies, and floor entrances, forming an RFID monitoring network covering the main circulation paths. When a medical waste recycling container or a medical hygiene product dispensing container with an RFID tag enters the reading range of a fixed RFID reader, the reader automatically reads its unique container identifier and records the reading time and its own device location code. Each reading event is encapsulated as a container trajectory point data point, which includes at least the container's unique identifier, reading time, and device location code. All container trajectory point data belonging to the same container's unique identifier are arranged in chronological order of reading time to generate a continuous movement trajectory sequence for the container. This movement trajectory sequence reflects the container's real-time location and historical path.

[0067] The data cleaning module receives raw read event streams from all fixed RFID readers, which include the container's unique identifier, read time, and device location code. The module performs deduplication within a time window. If multiple read events occur within a very short time threshold for the same container's unique identifier and the same device location code, only the first read event is retained as a valid event. The module then performs a trajectory rationality check on the deduplicated valid event sequence. Based on the positions and distances of adjacent fixed RFID readers, combined with a preset reasonable container movement speed, it determines whether the temporal and spatial continuity of two adjacent valid events meets the rationality threshold. For trajectory discontinuities that do not meet the rationality threshold, the background logic inserts one or more inferred virtual trajectory points based on the container's reasonable movement speed and path topology. These virtual trajectory points contain inferred time and location ranges to generate a smoother, more continuous movement trajectory sequence.

[0068] In practical implementation, the association and binding module of the RFID-based medical waste recycling and distribution system performs the following operations: When the RFID reader / writer set up at the medical waste recycling point reads the RFID tag on the surface of a medical waste recycling container, the system obtains the container's unique identifier and marks the status of the medical waste recycling container as "recycled," while recording the recycling time and recycling point location information; the system presets a duration threshold for an association time window, for example, set to four hours, and starts a timer to record the elapsed time since the medical waste recycling container was marked as "recycled"; during the elapsed time before exceeding the duration threshold, the medical waste recycling container is retained in the list of "recycled" medical waste recycling containers to be associated; when any medical waste distribution point When the RFID reader reads the unique identifier of any medical waste dispensing container, the system immediately queries the list of "recycled" medical waste collection containers to be associated, filtering out all medical waste collection containers that have not yet established a binding relationship with other medical waste dispensing containers. Among the filtered medical waste collection containers, the system compares the recycling times of each container and selects the medical waste collection container with the latest recycling time, i.e., the closest to the current time, as the associated object. A relationship record is created in the background database, permanently binding the unique identifier, recycling time, and recycling point location information of the selected medical waste collection container with the unique identifier, dispensing time, and dispensing point location information of the current medical waste dispensing container, forming a one-to-one correspondence between medical waste recycling and medical waste dispensing.

[0069] In some embodiments, the trajectory tracking module achieves container tracking in the following way: multiple fixed RFID readers are deployed at key path nodes such as corridor intersections, elevator lobbies, and floor entrances in medical institutions to form an RFID monitoring network covering the main circulation paths; when a medical waste recycling container or medical hygiene product dispensing container with a container RFID tag enters the reading range of any fixed RFID reader, the fixed RFID reader automatically reads the container's unique identifier and records the reading time and its own device location code; the system encapsulates each reading event into a container trajectory point data, which at least includes the container's unique identifier, reading time, and device location code; the system categorizes the container trajectory point data according to the container's unique identifier and arranges all container trajectory point data belonging to the same container unique identifier in chronological order of reading time to generate a continuous movement trajectory sequence of the container, which can reflect the container's real-time location and historical path.

[0070] Optionally, the data cleaning module processes the RFID monitoring network's read data: The module receives raw read event streams from all fixed RFID readers, which include the container's unique identifier, read time, and device location code. It performs deduplication within a time window; if multiple read events occur within a very short time threshold for the same container's unique identifier and the same device location code, only the first read event is retained as a valid event. The module then performs trajectory rationality checks on the deduplicated valid event sequence, determining whether the time difference and displacement of two adjacent valid events meet the rationality threshold based on the distance between adjacent fixed RFID readers and the preset reasonable container movement speed. For trajectory discontinuities that do not meet the rationality threshold, the module inserts one or more inferred virtual trajectory points in the background logic based on the container's reasonable movement speed and path topology. These virtual trajectory points include inferred timestamps and location ranges, thus making the movement trajectory sequence smoother and more continuous. It can be understood that the data cleaning module's processing helps improve the accuracy of subsequent trajectory analysis.

[0071] In one embodiment of the present invention, see [reference] Figure 3 The system extracts movement trajectory sequences of containers that have completed a full circulation cycle from historical data, forming a historical trajectory sample library. Each historical trajectory sample carries either its final circulation efficiency label or anomaly type label. For the target container currently in circulation, the latest partial movement trajectory sequence from the collection point is extracted as the query trajectory to be analyzed. The comprehensive similarity between the query trajectory and each historical trajectory sample in the historical trajectory sample library is calculated. The comprehensive similarity is obtained by weighted fusion of trajectory segment similarity measure, shape similarity based on dynamic time warping distance, and cosine similarity of dwell time and operation type feature vector. From the historical trajectory sample library, the K historical trajectory samples with the highest comprehensive similarity to the query trajectory are selected. Based on the circulation efficiency labels carried by these K historical trajectory samples, the expected circulation efficiency of the target container is predicted by weighted averaging, or based on the anomaly type labels carried by these K historical trajectory samples, a voting mechanism is used to determine whether the target container is currently experiencing a circulation anomaly and its type.

[0072] The movement trajectory sequence is divided into several trajectory segments according to the locations of fixed RFID readers / writers it passes through. Each trajectory segment is identified by its start and end positions. For a query trajectory and a historical trajectory sample, the sequences of continuous trajectory segments starting from the start point are compared. The number of trajectory segments with the same start and end positions is counted, and the ratio of this number to the total number of segments is defined as the trajectory segment similarity. A dynamic time warping method is applied to align the time series of the query trajectory and the historical trajectory sample. This time series is composed of the time intervals between adjacent trajectory points. The dynamic time warping method allows for non-linear scaling and alignment of the sequences on the time axis to find the minimum cumulative distance between the two time series. This minimum cumulative distance is the dynamic time warping distance, used to measure the similarity of trajectory shapes. When calculating the overall similarity, weight coefficients are assigned to the trajectory segment similarity, the inverse transformation value of the dynamic time warping distance, and the cosine similarity between the dwell time and the operation type feature vector. The weighted sum is then used to obtain the final overall similarity.

[0073] In practical implementation, the intelligent analysis module uses an improved K-nearest neighbor algorithm to analyze the container movement trajectory to predict circulation efficiency and identify circulation anomalies: the module pre-extracts the movement trajectory sequences of medical waste collection containers and medical clean material dispensing containers that have completed a full circulation cycle from the system's historical data, forming a historical trajectory sample library. Each historical trajectory sample is accompanied by a labeled circulation efficiency tag or anomaly type tag; for the target container currently in circulation, the module extracts the latest movement trajectory sequence of the target container starting from the collection point as the query trajectory to be analyzed; the module calculates the comprehensive correlation between the query trajectory and each historical trajectory sample in the historical trajectory sample library. Similarity is achieved through a weighted fusion of trajectory segment similarity, shape similarity based on dynamic time warping distance, and cosine similarity between dwell time and operation type feature vectors. In the historical trajectory sample library, the module selects the K historical trajectory samples with the highest overall similarity to the query trajectory. If the task is efficiency prediction, the module calculates the expected flow efficiency of the target container using a weighted average method based on the flow efficiency labels carried by the K historical trajectory samples. If the task is anomaly detection, the module determines whether the target container has flow anomalies and their specific types based on the anomaly type labels carried by the K historical trajectory samples through a majority voting mechanism.

[0074] In some embodiments, the improved K-nearest neighbor algorithm needs to construct a trajectory segment similarity metric and a dynamic time warping distance before calculating the overall similarity: The algorithm divides the moving trajectory sequence into several trajectory segments based on the locations of the fixed RFID readers it passes through, and each trajectory segment is jointly identified by the start position and the end position; For the query trajectory and a single historical trajectory sample, the algorithm compares the continuous trajectory segment sequences of the two starting from the start point, and counts the number of trajectory segments with the same start position and end position. The trajectory segment similarity metric is defined as the ratio of the number of identical trajectory segments to the total number of segments; The algorithm extracts the time interval between adjacent trajectory points in the query trajectory and the historical trajectory sample to form a time series, and applies a dynamic time warping method to align the two time series. The dynamic time warping method allows the time series to scale non-linearly on the time axis. The dynamic time warping distance is obtained by finding the minimum cumulative distance between the two series. The dynamic time warping distance is used to quantify the difference in trajectory shape.

[0075] Optionally, the formula for calculating the overall similarity is expressed as follows:

[0076]

[0077] in: Indicates the overall similarity. This represents a measure of similarity between trajectory segments. Indicates the dynamic time-warped distance. Indicates to The shape similarity is obtained after performing a reciprocal transformation and normalization. This represents the cosine similarity between the dwell time and the feature vector of the operation type. , , These are the weighting coefficients assigned to the three entities.

[0078] In one embodiment of the present invention, the predicted turnover efficiency or turnover anomaly judgment result of the target container output by the improved K-nearest neighbor algorithm is received. If the analysis result indicates that the predicted turnover efficiency is lower than the historical average level, an efficiency optimization allocation process is triggered to query the correspondence between medical waste recycling and clean material distribution, and to find the location of the already distributed clean material container corresponding to the target container. A preset instruction is generated to pre-deliver a spare clean material container to the area where the already distributed clean material container is located to prevent a shortage of clean material supply due to the turnover delay of the target container. If the analysis result indicates that the target container has a turnover anomaly, an anomaly handling allocation process is triggered to determine the last known location of the target container and generate a scheduling instruction to assign the nearest recycling personnel to the last known location for inspection and handling, and to allocate a temporary medical waste recycling container to the clinical department corresponding to the medical waste recycling container with the turnover anomaly.

[0079] A timer is started to record the elapsed time since a medical waste disposal container was marked as "recycled" at a collection point, with a preset duration threshold for the associated time window. Before the elapsed time reaches the duration threshold, the medical waste disposal container remains in the list of "recycled" medical waste disposal containers to be associated. When the unique identifier of any medical waste dispensing container is read at any medical waste dispensing point, the list of "recycled" medical waste disposal containers to be associated is immediately queried, and all medical waste disposal containers that have not yet established a binding relationship with other medical waste dispensing containers are selected. Among the selected containers, the recycling times of each container are compared, and the medical waste disposal container with the latest recycling time (i.e., the one closest to the current time) is selected as the most recently recycled container.

[0080] In practical implementation, the allocation and optimization module performs container allocation based on the analysis results output by the improved K-nearest neighbor algorithm: the module receives the predicted circulation efficiency value or the judgment result of the circulation anomaly type of the target container output by the intelligent analysis module; if the predicted circulation efficiency value is lower than the historical average circulation efficiency value, the module triggers the efficiency optimization allocation process, queries the correspondence database of medical waste recycling and clean material distribution, and locates the area where the medical clean material distribution container is permanently bound to the unique identifier of the target medical waste recycling container; the module generates a preset instruction to deliver the spare medical clean material container to the area where the distributed medical clean material container is located in advance to offset the impact of the circulation delay of the target medical waste recycling container; if the judgment result is a circulation anomaly, the module triggers the anomaly handling allocation process, obtains the last known location coordinates of the target medical waste recycling container through the trajectory tracking module, generates a scheduling instruction to assign the nearest recycling personnel to go for verification, and temporarily allocates and supplements medical waste recycling containers for relevant clinical departments.

[0081] In some embodiments, the association binding module has a clear association time window logic when processing the association between recycling and distribution: the module presets a duration threshold for the association time window, for example, set to T hours, and starts a timer to record the length of time elapsed since the initial moment when the medical waste recycling container is marked as "recycled" at the recycling point; during the period when the elapsed time is less than or equal to T hours, the medical waste recycling container is always kept in the list of "recycled" medical waste recycling containers to be associated; when the RFID reader / writer at any medical waste distribution point reads the unique identifier of the medical waste distribution container, the system immediately scans the list to be associated and filters out all medical waste recycling container records that have not established a binding relationship; the system compares the recycling timestamps of these records and selects the medical waste recycling container corresponding to the record with the latest recycling time for association binding.

[0082] Optionally, the priority for dispatching recovery personnel in the anomaly response and allocation process can be calculated using a distance-weighted scoring formula:

[0083]

[0084] in: The score represents the priority dispatch score for the recovery personnel. This represents a baseline distance constant set within the medical institution. This represents the real-time calculated distance between the current location of the retrieval personnel and the last known location of the aberration container. This represents the availability weighting coefficient of the current workload status of the retrieval personnel. It can be understood that this formula is used to quantitatively select the optimal dispatch target from multiple candidates. Refer to Table 1, which shows the changes in the proportion of successfully associated medical waste retrieval containers and medical clean supply containers under different association time window threshold settings:

[0085] Table 1: Successful Association Rate Between Medical Waste Recycling Containers and Medical Supplies Dispensing Containers

[0086] 2 hours 150 120 80.0% 4 hours 150 138 92.0% 6 hours 150 142 94.7%

[0087] In one embodiment of the present invention, a first weighting coefficient is assigned to the trajectory segment similarity, a second weighting coefficient is assigned to the inverse transformation value of the dynamic time warp distance, and a third weighting coefficient is assigned to the cosine similarity between the dwell time and the operation type feature vector, wherein the sum of the first, second, and third weighting coefficients is a fixed value. The shape similarity is obtained by performing an inverse transformation on the dynamic time warp distance, which is achieved by dividing 1 by the sum of the dynamic time warp distance and a preset constant, thus normalizing the shape similarity value to the range of 0 to 1. A weighted sum of the trajectory segment similarity, the normalized shape similarity, and the cosine similarity is calculated. The weighted sum is calculated by multiplying the trajectory segment similarity by the first weighting coefficient to obtain a first product, multiplying the normalized shape similarity by the second weighting coefficient to obtain a second product, multiplying the cosine similarity by the third weighting coefficient to obtain a third product, and then adding the first, second, and third products to obtain the comprehensive similarity.

[0088] In its implementation, the improved K-nearest neighbor algorithm performs explicit weighting and normalization operations when calculating the overall similarity: a first weight coefficient is assigned to the trajectory segment similarity, a second weight coefficient is assigned to the inverse transformation value of the dynamic time warp distance, and a third weight coefficient is assigned to the cosine similarity of the dwell time and operation type feature vector. The sum of the first, second, and third weight coefficients is set to a fixed constant. The dynamic time warp distance is mathematically transformed by dividing the number 1 by the sum of the dynamic time warp distance and a preset minimum constant to obtain the shape similarity, which ranges from 0 to 1. When calculating the overall similarity, the trajectory segment similarity is multiplied by the first weight coefficient to obtain the first product, the normalized shape similarity is multiplied by the second weight coefficient to obtain the second product, and the cosine similarity is multiplied by the third weight coefficient to obtain the third product. The first, second, and third products are then added together to obtain the final overall similarity value, which is directly used for subsequent neighbor sample retrieval.

[0089] In some embodiments, the specific values ​​of the weight coefficients can be adjusted according to the actual business scenario. See Table 2, which shows a set of weight configuration combinations after debugging in a specific hospital environment and their corresponding trajectory matching focuses:

[0090] Table 2: Weight Configuration Combinations and Corresponding Trajectory Matching Emphasis Table

[0091] Balanced 0.4 0.3 0.3 Balanced consideration of path and behavior Path priority 0.6 0.2 0.2 Strong dependence on physical path coincidence Behavior sensitive 0.25 0.25 0.5 Focus on stay and operation features

[0092] Optionally, the shape similarity conversion process can be precisely expressed by the following formula:

[0093]

[0094] in: This represents the calculated shape similarity. This represents the dynamic time-warped distance between the query trajectory and historical samples. This represents a tiny positive constant set to prevent the denominator from being zero. It can be understood that this calculation ensures that when the trajectory shapes are highly consistent, the shape similarity approaches its maximum value.

[0095] In one embodiment of the present invention, when the analysis result of the improved K-nearest neighbor algorithm indicates that the predicted turnover efficiency of the target container is lower than the historical average level, the system automatically triggers an efficiency optimization and allocation process. After the efficiency optimization and allocation process is initiated, the system retrieves the unique identifier of the medical waste distribution container permanently bound to the unique identifier of the target medical waste recycling container with low turnover efficiency in the background database, based on the established correspondence between medical waste recycling and clean waste distribution. The system queries the real-time location of the bound medical waste distribution container based on its unique identifier. The real-time location is determined by the latest reading record of the RFID reader / writer deployed at the medical waste distribution point or a fixed RFID reader / writer on a critical path node. The system retrieves a spare clean waste container from the clean waste container inventory that is currently in "pending" status and whose physical location is closest to the area where the queried already distributed clean waste container is located. The system generates a preset instruction, which includes the unique identifier of the selected spare clean waste container, the location information of the target delivery area (i.e., the area where the already distributed clean waste container is located), and a preset execution time limit. The system sends preset instructions to the mobile operating terminal of the cleaning supplies dispatch personnel to drive them to carry out the advance delivery of spare cleaning supplies containers, thereby preventing the shortage of clinical cleaning supplies due to the delay in the circulation of the target medical waste recycling containers.

[0096] In practical implementation, the operation steps of the allocation optimization module for the efficiency optimization allocation process are as follows: When the analysis results of the improved K-nearest neighbor algorithm show that the predicted circulation efficiency of a target medical waste recycling container is lower than the historical average level, the system automatically triggers the efficiency optimization allocation process; after the process is started, the system accesses the background database and, based on the persistently stored correspondence records between medical waste recycling and clean item distribution, retrieves the unique identifier of the medical clean item distribution container permanently bound to the unique identifier of the target medical waste recycling container with the currently low circulation efficiency; based on the retrieved unique identifier of the medical clean item distribution container, the system queries the latest location of the medical clean item distribution container, which is derived from the RFID deployed at the medical clean item distribution point. The system retrieves the latest reading record uploaded by the reading and writing device or the fixed RFID reading and writing device at the critical path node; in the medical supplies container inventory management system, the system searches for the spare medical supplies container whose current status is marked as "pending" and whose physical storage location is closest to the area where the issued medical supplies container is located; the system generates a preset instruction, which includes the unique identifier of the selected spare medical supplies container, the location code of the target delivery area (i.e., the area where the issued medical supplies container is located), and the preset execution time limit; the system sends the preset instruction to the mobile operation terminal held by the medical supplies dispatching personnel, driving the dispatching personnel to perform the advance delivery operation of the spare medical supplies container to make up for the clinical supplies supply gap that may be caused by the delay in the circulation of the target medical waste recycling container.

[0097] In some embodiments, the selection strategy for backup medical containers can incorporate a distance scoring mechanism to quantitatively evaluate the suitability of each candidate backup medical container to the target area: the system obtains the location coordinates of the area where the distributed medical containers are located, and obtains the real-time storage location coordinates of each candidate backup medical container; the system calculates the physical straight-line distance or path planning distance between the two, and sorts each candidate backup medical container according to the distance, prioritizing the backup medical container with the shortest distance for delivery. Optionally, when generating a preset execution time limit, the system can calculate a time margin based on historical flow data, and the estimation formula is as follows:

[0098]

[0099] in: The suggested duration representing the preset execution time limit This represents the historical average transit time for similar medical waste recycling containers from the collection point to the corresponding distribution point. This represents the time elapsed since the completion of the recycling process for the target medical waste collection container. This formula can be understood as being used to estimate the remaining available time window.

[0100] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An RFID-based medical waste collection and distribution system, characterized in that, include: The container identification module fixes RFID tags to the surfaces of medical waste recycling containers and medical clean supply containers respectively, and the container RFID tags store a unique container identifier. The data acquisition module deploys RFID reading and writing devices at medical waste collection points and medical clean product distribution points. The RFID reading and writing devices read information from the RFID tags on the containers. A recycling operation terminal is configured at the medical waste collection point, and a distribution operation terminal is configured at the medical clean product distribution point. The recycling operation terminal and the distribution operation terminal acquire the data read by the RFID reading and writing devices. The association and binding module establishes a correspondence between medical waste collection and clean material distribution, and associates and binds the information of collected medical waste with the information of clean materials to be distributed. The trajectory tracking module continuously tracks the location and status of the medical waste recycling container and the medical hygiene product dispensing container through a deployed RFID reader / writer. The intelligent analysis module uses an improved K-nearest neighbor algorithm to analyze the movement trajectory of containers in order to predict container turnover efficiency and identify turnover anomalies. The improved K-nearest neighbor algorithm introduces trajectory segment similarity measurement and dynamic time warping distance, and integrates container dwell time and operation type as feature dimensions. The allocation and optimization module generates container allocation instructions based on the analysis results of the improved K-nearest neighbor algorithm and the correspondence between medical waste recycling and clean material distribution, in order to optimize the flow and configuration of containers.

2. The RFID-based medical waste collection and distribution system according to claim 1, characterized in that, The establishment of the correspondence between medical waste collection and clean material distribution includes: When the RFID reader reads the unique identifier of a medical waste recycling container at a medical waste recycling point, it marks the medical waste recycling container as "recycled" and records the recycling time and recycling point location information. Within a preset association time window, when the RFID reader reads the unique container identifier of any medical waste dispensing container at any medical waste dispensing point, it selects the most recently recycled container from the list of "recycled" medical waste recycling containers to be associated. In the background database, a relationship record is created to permanently bind the unique identifier of the selected medical waste recycling container, the recycling time, and the recycling point location information with the unique identifier of the currently read medical clean item distribution container, the distribution time, and the distribution point location information, forming a one-to-one correspondence between medical waste recycling and clean item distribution.

3. The RFID-based medical waste collection and distribution system according to claim 1, characterized in that, The continuous tracking of the location and status of the medical waste collection container and the medical hygiene product dispensing container via the deployed RFID reader / writer includes: Multiple fixed RFID readers and writers are deployed at key path nodes in medical institutions, including corridor intersections, elevator lobbies, and floor entrances, forming an RFID monitoring network covering the main circulation paths. When a medical waste recycling container or medical hygiene product dispensing container with a container RFID tag enters the reading range of the fixed RFID reader, the fixed RFID reader automatically reads its unique container identifier and records the reading time and its own device location code. Each reading event is encapsulated as a container trajectory point data, which includes at least a unique container identifier, reading time, and device location code. All container trajectory point data belonging to the same unique container identifier are arranged in chronological order of recognition time to generate a continuous movement trajectory sequence of the container. The movement trajectory sequence reflects the real-time position and historical path of the container.

4. The RFID-based medical waste collection and distribution system according to claim 3, characterized in that, The improved K-nearest neighbor algorithm analyzes the container movement trajectory to predict container turnover efficiency and identify turnover anomalies, including: Extract the movement trajectory sequence of containers that have completed a full circulation cycle from historical data to form a historical trajectory sample library. Each historical trajectory sample carries its final circulation efficiency label or anomaly type label. For the target container that is currently in circulation, extract its latest partial movement trajectory sequence starting from the recycling point as the query trajectory to be analyzed. The comprehensive similarity between the queried trajectory and each historical trajectory sample in the historical trajectory sample library is calculated. The comprehensive similarity is obtained by weighted fusion of trajectory segment similarity measure, shape similarity based on dynamic time warping distance, and cosine similarity of dwell time and operation type feature vector. In the historical trajectory sample database, select the K historical trajectory samples with the highest comprehensive similarity to the queried trajectory; Based on the flow efficiency labels carried by these K historical trajectory samples, the expected flow efficiency of the target container is predicted by weighted averaging; or based on the anomaly type labels carried by these K historical trajectory samples, a voting mechanism is used to determine whether the target container is currently experiencing a flow anomaly and its anomaly type.

5. The RFID-based medical waste collection and distribution system according to claim 4, characterized in that, The improved K-nearest neighbor algorithm introduces a trajectory segment similarity metric and a dynamic time warping distance. Its working principle includes: The movement trajectory sequence is divided into several trajectory segments according to the locations of the fixed RFID reading and writing devices it passes through. Each trajectory segment is identified by its start and end points. For a query trajectory and a historical trajectory sample, compare the sequence of continuous trajectory segments starting from the starting point of the query trajectory and the historical trajectory sample, and count the number of trajectory segments with the same starting and ending positions. The ratio of the number of trajectory segments to the total number of segments is defined as the trajectory segment similarity. The dynamic time warping method is applied to align the time series of the query trajectory with those of historical trajectory samples. The time series are composed of the time intervals between adjacent trajectory points. The dynamic time warping method allows the sequences to be non-linearly scaled and aligned on the time axis to find the minimum cumulative distance between the two time series. The minimum cumulative distance is the dynamic time warping distance, which is used to measure the similarity of trajectory shapes. When calculating the overall similarity, weight coefficients are assigned to the trajectory segment similarity, the inverse transformation value of the dynamic time warping distance, and the cosine similarity between the dwell time and the operation type feature vector, and the weighted sum is used to obtain the final overall similarity.

6. The RFID-based medical waste collection and distribution system according to claim 1, characterized in that, Based on the analysis results of the improved K-nearest neighbor algorithm, and combined with the correspondence between medical waste recycling and clean material distribution, a container allocation instruction is generated, including: Receive the predicted flow efficiency or flow anomaly judgment result of the target container output by the improved K nearest neighbor algorithm; If the analysis result indicates that the predicted circulation efficiency is lower than the historical average, the efficiency optimization and allocation process is triggered to query the correspondence between the medical waste recycling and the clean material distribution, and to find the location of the clean material container that has been distributed and corresponds to the target container. Generate a preset instruction to deliver spare cleaning containers in advance to the area where the already issued cleaning containers are located, in order to prevent a shortage of cleaning supplies due to delays in the circulation of the target containers; If the analysis result indicates that the target container has experienced a flow abnormality, the abnormality handling and allocation process is triggered to determine the last known location of the target container and generate a scheduling instruction to assign the nearest recycling personnel to the last known location for inspection and handling, and to allocate a temporary medical waste recycling container to the clinical department corresponding to the medical waste recycling container that has experienced the flow abnormality.

7. The RFID-based medical waste collection and distribution system according to claim 2, characterized in that, Within a preset association time window, when the RFID reader / writer reads the unique identifier of any medical waste dispensing container at any medical waste dispensing point, it selects the most recently recycled container from the list of "recycled" medical waste recycling containers to be associated, including: A preset duration threshold for the associated time window is established; a timer is started to record the elapsed time since a medical waste recycling container is marked as "recycled" at a recycling point; before the elapsed time reaches the duration threshold, the medical waste recycling container is continuously retained in the list of "recycled" medical waste recycling containers to be associated; when the unique container identifier of any medical waste dispensing container is read at any medical waste dispensing point, the list of "recycled" medical waste recycling containers to be associated is immediately queried, and all medical waste recycling containers that have not yet established a binding relationship with other medical waste dispensing containers are selected from the list; among the selected containers, the recycling times of each container are compared, and the medical waste recycling container with the latest recycling time, i.e., the closest to the current time, is selected as the most recently recycled container.

8. The RFID-based medical waste collection and distribution system according to claim 5, characterized in that, In calculating the comprehensive similarity, weight coefficients are assigned to the trajectory segment similarity, the inverse transformed value of the dynamic time warping distance, and the cosine similarity between the dwell time and the operation type feature vector, respectively. These are then weighted and summed to obtain the final comprehensive similarity, including: A first weight coefficient is assigned to the trajectory segment similarity, a second weight coefficient is assigned to the inverse transformation value of the dynamic time warp distance, and a third weight coefficient is assigned to the cosine similarity between the dwell time and the operation type feature vector, wherein the sum of the first weight coefficient, the second weight coefficient and the third weight coefficient is a fixed value; The shape similarity is obtained by performing a reciprocal transformation on the dynamic time warp distance. The transformation method is to divide 1 by the sum of the dynamic time warp distance and a preset constant, so that the shape similarity value is normalized to the range of 0 to 1. The weighted sum of the trajectory segment similarity, the normalized shape similarity, and the cosine similarity is calculated. The weighted sum is calculated by multiplying the trajectory segment similarity by the first weight coefficient to obtain a first product, multiplying the normalized shape similarity by the second weight coefficient to obtain a second product, multiplying the cosine similarity by the third weight coefficient to obtain a third product, and then adding the first product, the second product, and the third product to obtain the comprehensive similarity.

9. The RFID-based medical waste collection and distribution system according to claim 1, characterized in that, Based on the analysis results of the improved K-nearest neighbor algorithm, and combined with the correspondence between medical waste recycling and clean material distribution, container allocation instructions are generated, including an efficiency-optimized allocation process, specifically as follows: When the analysis result of the improved K-nearest neighbor algorithm shows that the predicted turnover efficiency of the target container is lower than the historical average, the system automatically triggers the efficiency optimization and allocation process. After the efficiency optimization and allocation process is initiated, the system retrieves the unique container identifier of the medical clean item distribution container that is permanently bound to the unique container identifier of the target medical waste recycling container with low circulation efficiency in the background database, based on the established correspondence between medical waste recycling and clean item distribution. The system queries the real-time location of the medical supplies dispensing container based on the unique container identifier of the bound medical supplies dispensing container. The real-time location is determined by the latest reading record of the RFID reader / writer deployed at the medical supplies dispensing point or the fixed RFID reader / writer on the critical path node. The system retrieves a spare cleaning container from the cleaning container inventory that is currently in "awaiting" status and whose physical location is closest to the area where the queried issued cleaning container is located. The system generates a preset instruction, which includes the unique identifier of the selected backup cleaning container, the location information of the target delivery area (i.e., the area where the cleaning container has been distributed), and the preset execution time limit; The system sends the preset instructions to the mobile operating terminal of the cleaning supplies dispatch personnel to drive them to perform the advance delivery of spare cleaning supplies containers, thereby preventing a shortage of clinical cleaning supplies due to delays in the circulation of the target medical waste recycling containers.

10. The RFID-based medical waste collection and distribution system according to claim 3, characterized in that, The system also includes a data cleaning module, used to clean and complete the RFID monitoring network's read data, specifically: Receive raw read event streams from all fixed RFID reader / writer devices, the raw read event streams containing container unique identifiers, read times, and device location codes; The original reading event stream is deduplicated within a time window. When the same container unique identifier is on the same device location code and multiple reading events are generated within a very short time threshold, only the first reading event is retained as a valid event. The trajectory rationality test is performed on the effective event sequence after deduplication. Based on the position and distance of adjacent fixed RFID reading and writing devices, combined with the preset reasonable moving speed of the container, it is determined whether the continuity of two adjacent effective events in time and space meets the rationality threshold. For trajectory discontinuities that do not meet the reasonableness threshold, one or more inferred virtual trajectory points are inserted in the background logic based on the reasonable movement speed and path topology of the container. The virtual trajectory points include the inferred time and location range to generate a smoother and more continuous movement trajectory sequence.

Images