Intelligent lamp control method and system based on Internet of Things, electronic equipment and medium

Abstract

The invention discloses an intelligent lamp control method and system based on the Internet of Things, electronic equipment and a medium, and particularly relates to the field of Internet of Things control, and the method comprises the steps: collecting a response state change event of a lamp, generating a complete track sequence which is continuous according to time and space, a physically closed response subnet is extracted through co-occurrence frequency analysis and connectivity verification; in the subnet, behavior label labeling is carried out based on the time period to which the track belongs and the area span, a classifier is trained to identify different types of traffic behaviors, and high-frequency nodes in the typical track are extracted to construct a behavior template; and combining a classifier structure with a template to form a subnet-level behavior model for behavior matching and offset judgment of a real-time track. A sensitivity regulation and control instruction is gradually issued to the lamp nodes identified as the abnormal candidates; the sound control lamp abnormal response identification and sensitivity self-adjustment method based on Internet of Things cooperative perception, dynamic behavior pattern learning and intelligent regulation and control linkage is realized.

Description

Technical Field

[0001] The present invention relates to the field of Internet of Things control technology. More specifically, the present invention relates to an Internet of Things-based intelligent lighting control method, system, electronic device, and medium. Background Art

[0002] In enclosed spaces such as multi-story buildings, school dormitories, and office building corridors, voice-activated lights are usually distributed on the walls or ceilings of each section of the passage to respond to the sounds generated by passing people and turn on. Traditional voice-activated lights mostly use local sensors and simple amplification circuits to judge sound signals, and do not have the ability of system collaborative perception and self-performance regulation during operation. With the large-scale deployment of voice-activated lights and the increase in the service life, the sensitivity of the lights gradually declines without obvious faults, manifested as: some lights need a louder sound to trigger, and some lights remain off even when other lights have responded, forming "lighting holes" or "response blind spots". Since these problems do not trigger communication failures or alarms, it is difficult for operation and maintenance personnel to identify abnormalities through the platform and can only rely on user feedback or manual inspections to intervene.

[0003] Therefore, how to construct a method for identifying abnormal responses and self-adjusting the sensitivity of voice-activated lights based on Internet of Things collaborative perception, dynamic behavior pattern learning, and intelligent control linkage to achieve self-fault recognition and repair of Internet of Things intelligent lights is an urgent problem to be solved.

[0004] To solve the above problems, a technical solution is provided as follows. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an Internet of Things-based intelligent lighting control method to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: S1: Define the response behavior state change event of the voice-activated light, collect the response time series data including the response event timestamp and the node identifier, sort the response sequence within the set monitoring period, retrieve the set of response node pairs with connectable relationships, and splice them into a complete response trajectory; S2: Calculate the co-occurrence frequency of the trajectory nodes of the response trajectory and divide the response aggregation clusters, and organize the response aggregation clusters with physically connected closures into a response subnet structure; S3: Label the behavior tags for the trajectory sequences in the response subnet, and establish a classifier structure for outputting the behavior tag classification; S4: Extract the stable core paragraphs of the response nodes in the trajectory sequences corresponding to the behavior tags as the behavior template units, and combine the classifier structure and the behavior template units into a subnet-level behavior model; S5: Submit the real-time collected response timing data to the subnet-level behavior model for behavior matching and determination, locate the abnormal segments that fail to match successfully in the template structure, and construct a candidate list of abnormal nodes; S6: Send a sound pressure induction sensitivity parameter regulation instruction to each abnormal node in the candidate list of abnormal nodes, and update the status of the abnormal nodes based on the behavior matching determination result of the regulated abnormal nodes.

[0007] In a preferred embodiment, in S1, define the response behavior state change event of the voice-controlled lamp, collect the response timing data including the response event timestamp and the node identifier, perform response sequence sorting within the set monitoring period, retrieve the set of response node pairs with connectable relationships, and splice them into a complete response trajectory, specifically including: Divide the lighting area of the area where the intelligent voice-controlled lamps are arranged, establish a lighting area identifier, and define the lighting area attribution of the voice-controlled lamps; Connect all the voice-controlled lamps to the unified Internet of Things platform through a low-power wireless protocol, and configure a unique node identifier for each lamp; Define the response behavior state change event of the voice-controlled lamp, open the reporting interface of the response event at the platform end, and bind the reported data to the lighting area to form a basic data stream; The Internet of Things platform performs periodic polling on the reporting interfaces of all lamp nodes, and collects the response timing data including the response event timestamp and the node identifier; Perform ascending time sorting on the response timing data to generate a node response sequence that is continuous in time within the set monitoring period; After completing the response sequence re-sorting, based on the spatial coordinate distance between the lighting areas corresponding to the response nodes and the set time interval threshold condition, retrieve the set of node pairs with connectable relationships; Perform a connection operation based on a sliding window on the set of node pairs, continuously traverse the set of node pairs, and construct a response sequence segment with time and space continuity as the linking condition; Perform trajectory splicing processing on the response sequence segment, and merge the response sequence segments that meet the set splicing conditions into a complete response trajectory through the front-back relationship of the node pair identifiers and store them in the trajectory set.

[0008] In a preferred embodiment, in S2, calculate the co-occurrence frequency of trajectory nodes in the response trajectory and divide the response aggregation clusters, and organize the response aggregation clusters with physical connectivity closure into a response subnet structure, specifically including: Perform response statistical processing on the response trajectory nodes in the trajectory set, and form a response co-occurrence relationship matrix between nodes by calculating the co-occurrence frequency of each node response in the trajectory set within the monitoring period; Perform a distance tensor conversion operation on the co-occurrence relationship matrix, and generate a normalized interconnection strength tensor based on the ratio of the response co-occurrence times to the total number of trajectories, which is used to express the response collaboration density between any two nodes; Perform density clustering on the interconnection strength tensor, and use the local density maximum positioning method to extract the response center node group in the response concentration area, and divide the response aggregation clusters with this node group as the center; Filter out the response aggregation clusters with spatial interruptions, and retain the clustering results with physical connectivity closed; Organize the response aggregation clusters with physical connectivity closed into response subnets, and the response subnet structure consists of response node identifiers and trajectory sequences carrying response time spans.

[0009] In a preferred embodiment, in S3, label the behavior labels for the trajectory sequences in the response subnet, and the classifier structure for establishing the output behavior label classification specifically includes: Divide the monitoring period into several monitoring time periods, and label each trajectory based on the monitoring time period to which the time span of the trajectory sequence in the response subnet belongs, and the lighting area span corresponding to the trajectory; Input the labeled trajectory sequences of n monitoring periods into the classifier structure, load the training samples with the selected random forest model, and use the lighting area span and the distribution of the monitoring time periods to which the trajectory nodes respond as training features, and output the behavior label classifier parameter group of the response subnet; Use the same labeled trajectory sequences as the validation set, and evaluate the behavior consistency of the labels in a ten-fold cross-validation manner for the trained classifier structure, and eliminate the classifiers whose accuracy does not reach the set index.

[0010] In a preferred embodiment, in S4, extract the stable core paragraphs of the response nodes in the trajectory sequences corresponding to the behavior labels as behavior template units, and combine the classifier structure and the behavior template units into a subnet-level behavior model, which specifically includes: Perform pattern summary extraction on the set of typical trajectory sequences corresponding to each type of behavior label, and extract the stable core paragraphs of the response nodes in the trajectory sequence and record them as behavior template units by calculating the node response frequency distribution and the trajectory segment repetition rate; Combine the classifier structure and the behavior template units into a subnet-level behavior model, and map it to the response nodes with the behavior label number as the index, as the standard reference for the abnormal behavior of the response nodes.

[0011] In a preferred embodiment, in S5, submit the real-time collected response time series data to the subnet-level behavior model for behavior matching determination, locate the abnormal segments that do not match successfully in the template structure, and construct a list of abnormal node candidates, which specifically includes: The Internet of Things platform performs behavior matching determination on the response time-series data collected in real time, and uses the behavior template unit output by the subnet-level behavior model as the trajectory sequence matching reference corresponding to the current response time-series data; Apply a sliding window of fixed length on the trajectory sequence to compare it segment by segment with the node sequence of the behavior template unit, calculate the node matching offset using the edit distance, and generate a matching deviation curve; Based on the matching deviation curve, perform peak recognition for local response discontinuity to locate the abnormal segments in the trajectory sequence that do not match successfully in the template structure; Map the node indices of the located abnormal segments to specific lamp nodes to construct a list of abnormal node candidates under the current behavior trajectory.

[0012] In a preferred embodiment, in S6, send a sound pressure induction sensitivity parameter regulation instruction to each abnormal node in the list of abnormal node candidates, and update the abnormal node status based on the behavior matching determination result of the regulated abnormal node, which specifically includes: In the list of abnormal node candidates, bind a regulation status record container and an adjustment times counter to each abnormal node; In each monitoring period, the Internet of Things platform sends a sound pressure induction sensitivity parameter regulation instruction with a set regulation step size to the abnormal nodes that have not reached the upper limit of the adjustable parameters; Perform behavior matching determination on the real-time response time-series data after each parameter regulation. If the matching offset of the trajectory sequence corresponding to the current abnormal node is less than the acceptable threshold of the matching offset in consecutive m time periods, update the regulation status to regulation completed and stop the subsequent regulation task for this abnormal node; If the adjustment times count of the abnormal node reaches the set maximum allowable number, or the regulation parameter reaches the upper limit of the adjustable parameters, then set the status of this abnormal node to response degradation not recovered and write it into the fault log.

[0013] On the other hand, the present invention provides an Internet of Things intelligent lamp control method system, including a data access module, a trajectory splicing module, a subnet clustering module, a behavior matching and recognition module, and a parameter regulation module: Data access module: used to divide the building lighting area, establish the mapping relationship between lamps and areas, and complete the Internet of Things node access and response event data reporting of voice-controlled lamps; Trajectory splicing module: sort the response time-series data, perform node pair screening and sliding window splicing processing, and generate a set of complete behavior trajectories with continuous and clear time sequence for subsequent recognition; Subnet clustering module: construct a response subnet through the co-occurrence relationship of trajectory nodes, and generate a behavior recognition model by training a classifier for the trajectory sequence based on the lighting area and time period characteristics; Behavior matching and identification module: performs sliding window matching between real-time trajectories and behavior templates, extracts abnormal response fragments based on edit distance, and constructs a candidate list of abnormal nodes for control and positioning; Parameter control module: sends sensitivity control instructions to abnormal nodes on a periodic basis, and determines the control completion status or marks it as a response degraded node based on behavior matching feedback.

[0014] The technical effects and advantages of the method, system, electronic device and medium for controlling intelligent lighting based on the Internet of Things of the present invention are as follows: Based on the IoT platform, the time series data collection and trajectory structure construction of the response behavior of voice-controlled lamps are combined with spatial connectivity and response co-occurrence rules to realize the intelligent identification and regulation of the sensitivity degradation problem of voice-controlled lamps. The system continuously splices the response paths of nodes in the lighting area, extracts stable behavior trajectories and constructs clustered response subnets, so that the nodes are no longer isolated sensors, but integrated into the behavior pattern analysis process. The trajectory classifier and behavior template structure are introduced to establish a standard response reference system under multiple behavior scenarios, which significantly improves the context judgment ability of abnormal identification. Combined with the offset analysis of real-time response data, the abnormal nodes that fail to match the behavior template are accurately located. The system can issue parameter control instructions in a targeted manner and dynamically adjust the strategy based on feedback to realize the hierarchical recovery mechanism of the sound pressure sensing threshold.

[0015] The solution achieves low-interference, low-cost, sustainable self-repairing control of large-scale voice-controlled lamps without relying on additional acoustic sensors or manual calibration, enhancing the system's environmental adaptability and autonomous maintenance capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of a method for controlling an intelligent lamp based on the Internet of Things according to the present invention; Figure 2 The present invention is a structural schematic diagram of an Internet of Things-based intelligent lighting control system. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0018] Figure 1 The present invention provides a method for controlling an intelligent lamp based on the Internet of Things, which comprises the following steps: S1: Define the response behavior state change event of the voice-controlled lamp, collect the response timing data including the response event timestamp and node identifier, perform response sequence sorting within the set monitoring period, retrieve the set of response node pairs with connectable relationships, and splice them into a complete response trajectory. S2: Calculate the co-occurrence frequency of trajectory nodes for the response trajectory and divide the response aggregation clusters, and organize the response aggregation clusters with physical connectivity closure into a response subnet structure. S3: Label the behavior tags for the trajectory sequences in the response subnet, and establish a classifier structure for outputting the behavior tag classification. S4: Extract the stable core paragraphs of the response nodes in the trajectory sequences corresponding to the behavior tags as behavior template units, and combine the classifier structure and the behavior template units into a subnet-level behavior model. S5: Submit the real-time collected response timing data to the subnet-level behavior model for behavior matching determination, locate the abnormal segments that fail to match successfully in the template structure, and construct a list of abnormal node candidates. S6: Send a sound pressure induction sensitivity parameter regulation instruction to each abnormal node in the list of abnormal node candidates, and update the abnormal node status based on the behavior matching determination result of the regulated abnormal node.

[0019] In S1, define the response behavior state change event of the voice-controlled lamp, collect the response timing data including the response event timestamp and node identifier, perform response sequence sorting within the set monitoring period, retrieve the set of response node pairs with connectable relationships, and splice them into a complete response trajectory.

[0020] According to the building structure diagram and the voice-controlled lamp layout plan, use continuous corridor segments, corner nodes, and spatial physical partitions as the boundaries of the lighting areas, and divide the entire building into several logical lighting areas. Each area is assigned a unique identification number, and structural parameters such as the spatial coordinate range, the floor it belongs to, and the adjacent area relationship are recorded. Establish an attribution relationship between the voice-controlled lamp nodes installed in each area and the corresponding lighting area, complete the mapping through the location information during platform registration, and maintain a two-way index table in the data management layer in the form of "lamp number - lighting area number".

[0021] All voice-controlled lamps are connected to a unified Internet of Things platform, use a low-power wireless communication protocol for node networking, and are uniformly networked and managed through a gateway. The system assigns a unique communication node identifier to each lamp, and binds its physical label, the lighting area number it belongs to, and the deployment location attribute. The node registration process includes handshake authentication, area number confirmation, and node status synchronization to ensure that all response events can be traced back to a clear device entity.

[0022] Define the response behavior of the intelligent voice-controlled lamp on the platform side as the "lighting-on event". When a certain lamp changes from the off state to the lit state, a status change event is automatically generated. The event content includes: node identifier, lighting area number, response timestamp, event type code (fixed as "lighting-on"), and node spatial position code. The platform sets up a multi-protocol event access channel, supporting two types of transmission methods: edge device reporting and gateway forwarding. All events are appended with their belonging area information before being stored in the database and written into the main table of the platform response data.

[0023] In addition to the passive reporting mechanism of the response behavior, the platform performs periodic polling operations on all connected lamps. At the start of each round of the cycle, the system sends a status synchronization request to each node and actively pulls the current status and the most recent response event of the nodes that have not reported data. The timestamp comparison and record completion processing are performed on the event data obtained during the polling to ensure that the system can still completely restore the response history sequence of the nodes in case of communication interruption or partial data loss.

[0024] The platform calls the lamp response behavior data within the set monitoring period in the event time series database, re-arranges all data records in ascending order according to the timestamp field, and forms a time series structure with the response time - node identifier as the basic unit. Each record has the unique number of the responding node, the response time, the number of the belonging lighting area, and the spatial position coordinates. The sorting result is temporarily stored in the trajectory initialization buffer table in the form of a time linear structure.

[0025] Among them, the default monitoring period length is selected as 24 hours as the basic period. After the time series is constructed, for any two adjacent response events in time, based on the spatial coordinate distance between their belonging lighting areas (not exceeding the maximum sound pressure induction distance set by the lamp at the factory) and the set time interval threshold condition (specifically set based on the spatial distance of the lamp, default set to 5 seconds), it is judged whether they meet the connectable condition. The node pairs that meet the set maximum time interval and maximum spatial distance constraint conditions are included in the set of connectable node pairs, and the sequence pair number and the front-back order identifier are appended.

[0026] Apply a sliding window mechanism with a fixed width to the set of connectable node pairs. Each time the window covers several consecutive node pairs. In each window, the response sequence segments that meet the trajectory segment construction conditions are selected according to time monotonicity and spatial continuity. The sequence segment structure consists of the starting node, the ending node, the response time series, and the lighting area number series.

[0027] For example, within a certain monitoring period, the platform continuously receives four response events from nodes L1, L2, L3, and L4. The response intervals are no more than 5 seconds, and the spatial coordinates are all linearly distributed in one dimension. After sorting, the system identifies that L1→L2→L3→L4 constitutes a unidirectional continuous response path, and then generates a trajectory number T-001, and stores the trajectory structure in the trajectory set for subsequent behavior pattern modeling.

[0028] Perform path splicing processing on all the constructed response sequence segments according to time continuity and the front-back relationship of nodes. For the segments where there is a connection relationship between the tail node and the head node of the next sequence segment, the system combines them into a single trajectory structure and gradually constructs a complete response path. Each trajectory is finally indexed by the trajectory number, recording all node identifiers, response time series, spatial region path, and node response duration in the trajectory.

[0029] In S2, calculate the co-occurrence frequency of trajectory nodes for the response trajectories and divide the response aggregation clusters, and organize the response aggregation clusters with physical connectivity closure into a response subnet structure.

[0030] Traverse all the trajectory structures in the trajectory set, count the number of times each node appears in all the trajectories, and at the same time accumulate the co-occurrence times between any two nodes. The condition for a node pair to be regarded as a co-occurrence relationship is that they appear together in the same trajectory regardless of the order. Fill the co-occurrence frequencies of all node pairs into a two-dimensional relationship matrix. Each unit of the matrix represents the co-occurrence times of the corresponding node pair. If they do not co-occur, it is set to zero, forming a response co-occurrence relationship matrix between nodes.

[0031] To eliminate the influence of the difference in the number of trajectories on the co-occurrence relationship, perform normalization processing on each unit in the co-occurrence relationship matrix. The normalization method is: divide the co-occurrence times by the total number of trajectories in the current monitoring period to generate an interconnection strength value ranging from 0 to 1, representing the collaborative density level between two nodes in the trajectory behavior. All the normalization results form an interconnection strength tensor, which is used as the input for similarity determination in the subsequent clustering processing stage.

[0032] Taking the interconnection strength tensor as the input, the local density maximum clustering method is used to identify several response concentration regions within the entire node set. The clustering algorithm calculates the local density score based on the average interconnection strength of each node, and uses the density extreme point as the response center, further classifying the high-strength connection nodes into the response aggregation cluster centered on this center. Each aggregation cluster structurally represents a set of nodes that are often linked and lit up in multiple trajectories. For example, in a set of trajectories, nodes L2, L3, and L4 appear simultaneously in most paths, and the interconnection strength exceeds 0.8. The system constructs a clustering with L3 as the local density center and initially forms the aggregation cluster C-01. Through spatial topology inspection, it is found that these three are continuously arranged in the same corridor section and meet the connectivity requirements. Therefore, the aggregation cluster C-01 is confirmed as the effective response subnet S-01, and the node list and corresponding trajectory number list are recorded in the structure.

[0033] Perform a spatial structure connectivity test on all the completed response aggregation clusters. The test method is to judge whether the spatial positions of the nodes within each aggregation cluster form a connectable path in the lighting area structure diagram. If there are node pairs separated by more than two physical lighting areas or with obvious partitions in the spatial layout, then the aggregation cluster is marked as spatially interrupted and does not enter the response subnet set. Only the clustering structures with continuous physical node distribution and traversable spatial paths are retained.

[0034] For the response aggregation clusters that pass the connectivity verification, extract the list of node identifiers they contain, and collect all the trajectory numbers and response time ranges they belong to to construct the response subnet structure unit. Each subnet structure is indexed by the subnet number and includes a node set, a set of participating trajectories, a time coverage range, and a set of lighting area numbers it belongs to. All the subnet structures are written into the response subnet library for subsequent behavior classifier training and template extraction.

[0035] In S3, label the trajectory sequences in the response subnet with behavior labels and establish a classifier structure for outputting behavior label classification.

[0036] Perform a time dimension slicing operation on the monitoring period defined in the platform, dividing each monitoring period into several monitoring time periods of fixed length, such as morning rush hour, noon break, evening departure, etc. Then, traverse all the trajectory sequences in the response subnet, extract the time span interval of each trajectory, and map it to the label of the monitoring time period it belongs to. At the same time, calculate the list of lighting area numbers involved in the trajectory sequence and set the behavior space scale according to its area span value. Combining the two-dimensional information of time and space, formulate a trajectory label naming rule, such as "high-frequency short-area crossing", "low-frequency multi-area patrol", etc., and assign labels to all the trajectory sequences to form a labeled trajectory set with behavior semantics.

[0037] The set of labeled trajectory sequences for n (n ≥ 3) monitoring periods is used as the classifier training sample data and input into the supervised learning process. The selected training model is a random forest structure, and the sample features consist of two parts: one is the distribution density and the number of crossings of trajectory nodes on the lighting area; the other is the position of the monitoring time period that appears in the trajectory response event set. After feature extraction, all samples are loaded to execute the training process for the model, and the output structure is a set of classifier parameter groups for the behavior labels of the response subnet, which is used for the behavior type recognition of real-time trajectories.

[0038] Execute the ten-fold cross-validation process on the trained classifier structure, and use the same labeled trajectory samples as the training set in the validation set. In each round of cross-validation, the system records the matching accuracy between the label prediction result and the true label, and generates a consistency scoring metric for the overall model. Classifier structures with an accuracy lower than the set threshold (specifically set based on the complexity of the subnet trajectory mode, with a default setting of 90%) are excluded and not included in the behavior model structure library, and the classifiers that pass the validation are retained as the behavior recognition benchmarks for subsequent anomaly recognition and template extraction. For example, a certain trajectory appears between 7:50 and 8:05 am every day, and the path crosses from floor L1 to L3, which is named the "going upstairs" behavior label. This label appears in a total of 45 trajectory samples within a week. After participating in the classifier training, it together with behaviors such as "cleaning inspection" and "staying in place in the corridor" constitutes the classification model. After training, the ten-fold cross-validation results show that the average prediction accuracy of the "going upstairs" label reaches 93%, so this model is retained for real-time behavior path discrimination.

[0039] In S4, extract the stable core paragraph of the response nodes in the trajectory sequence corresponding to the behavior label as the behavior template unit, and combine the classifier structure and the behavior template unit into a subnet-level behavior model.

[0040] In the trajectory set with labeled behavior labels, divide it by label type, and extract the trajectory subset under each label type as the typical trajectory sequence set. Perform node frequency statistics on all trajectory structures in this set, record the frequency value of each node appearing in the set, and count the repeated trajectory paragraphs in the trajectory and their coverage ratio. Dynamically set the frequency threshold (the frequency value corresponding to the first 40% of the boundary nodes) and the paragraph coincidence rate threshold (specifically set according to the required recognition accuracy, with a default setting of the average of the paragraph coincidence rates of all repeated trajectory paragraphs), and filter out the nodes and path segments that stably appear in the vast majority of trajectories, and combine them to form a response node sequence structure with behavior representativeness, which is called the behavior template unit. This structure is used to depict the standard response path under this behavior label.

[0041] Among them, the proportion of repeated nodes in a single trajectory does not exceed 30% of the total number of nodes. The total number of trajectory nodes should be within a preset length range (such as between 3 and 8 nodes), and the trajectory path structure ranks in the top 40% in terms of the occurrence frequency in the set of labeled trajectories. Trajectories meeting the above conditions are identified as typical trajectories.

[0042] After the generation of the behavior template unit, it is structurally integrated with the previously verified behavior classifier structure to construct a complete behavior model unit. Each model unit uses the behavior label number as the index key and maps to two types of data structures: one is the core response node sequence in the behavior template unit, and the other is the classifier model parameter group. This combined structure is called the subnet-level behavior model and is uniformly stored in the behavior model structure set. This model can quickly retrieve the standard response structure corresponding to the behavior in subsequent real-time trajectory judgment and serve as the direct reference basis for behavior deviation evaluation.

[0043] For example, under the behavior label "faculty and staff go downstairs and leave", a total of 35 trajectory sequences are aggregated. The system analyzes and finds that nodes L5, L4, L3, and L2 appear continuously in the same order in more than 90% of the trajectories, and the average coverage ratio of this section of the path reaches 82%. The system extracts this sequence as the "go downstairs and leave" template unit and combines it with the trained classifier structure to generate the behavior model M-04. Any subsequent real-time trajectory that conforms to this behavior label uses M-04 as the structural reference for response node integrity and deviation analysis.

[0044] In S5, the real-time collected response time series data is submitted to the subnet-level behavior model for behavior matching determination to locate abnormal segments that fail to match successfully in the template structure and construct a list of abnormal node candidates.

[0045] After the IoT platform receives the real-time trajectory data, it calls the behavior classifier structure to determine the behavior label to which it belongs and indexes the subnet-level behavior model under the corresponding label, and extracts the internally stored behavior template unit as the standard comparison structure. The current trajectory node sequence is compared with the template node sequence in terms of structure to form a set of trajectories to be matched.

[0046] The system enables a fixed-length sliding window processing mechanism on the current trajectory, divides the trajectory into several continuous node sub-segments, and compares them with paragraphs of the same length in the behavior template path one by one. Between each pair of node segments, the node matching offset is calculated through the edit distance to determine whether there are node omissions, order misalignments, or replacements in the trajectory segment. The matching results are converted into a set of offset values and arranged in the order of trajectory time to form a complete matching deviation curve.

[0047] Perform a fluctuation analysis on the matching deviation curve, and the system locates the peak positions therein to identify segments in the trajectory where the offset is significantly greater than the normal fluctuation threshold (the setting method is the ratio of the allowed number of node mismatches to the window size). These segments are regarded as local response discontinuities, that is, the actual trajectory fails to match the template path in this section, and it is preliminarily determined that there is a response loss or node failure.

[0048] According to the trajectory position index of the local abnormal segment, reverse-lookup the node number that should appear at this position and check whether it exists at the corresponding position in the actual trajectory. For the nodes that should appear but are missing, the system extracts their lamp numbers, spatial coordinates, and trajectory timestamps in the lighting area and writes them into the abnormal node candidate list. This list is used for subsequent parameter regulation instruction generation and status update, as the basis for locating potential failure nodes.

[0049] In S6, send a sound pressure induction sensitivity parameter regulation instruction to each abnormal node in the abnormal node candidate list, and update the abnormal node status based on the behavior matching judgment result of the regulated abnormal node.

[0050] After generating the abnormal node candidate list, the system creates a regulation status record container for each node to record the execution process status, response trajectory matching result, and current regulation parameter value during the regulation stage. At the same time, configure an independent adjustment times counter for each node to monitor the number of regulation rounds that the node has executed. The regulation status is initially set to be regulated, and the initial value of the counter is set to 0.

[0051] The platform enters the periodic regulation execution stage, traversing all abnormal nodes that are in the "to be regulated" state and have not reached the parameter adjustment upper limit. For the qualified nodes, generate a sound pressure induction sensitivity parameter regulation instruction according to their current status. The regulation instruction contains information such as the set parameter step size (for example, reducing the threshold by 1.5 dB), the target parameter value range, the execution cycle identifier, and the status receipt request. The platform synchronously sends this instruction to the target node through the Internet of Things communication interface.

[0052] After each round of regulation is completed, the system enters the real-time response monitoring process, captures the subsequent response timing data of the regulated node, and generates a trajectory sequence structure. The system calls the behavior template unit in the behavior model to perform a matching offset analysis on whether there is still a response loss in the trajectory. If the matching offset of this node is less than the set tolerance threshold in m (m is greater than or equal to 5) consecutive monitoring periods, it is considered that the current regulation has restored the response ability of the target node. The system updates the regulation status of this node to regulated completed and terminates its subsequent regulation plan.

[0053] If a certain abnormal node has reached the set maximum number of regulation rounds during continuous regulation, or its sensitivity parameter has been adjusted to the minimum allowable response threshold, but still fails to meet the offset tolerance standard, the system considers that the node has not recovered within the adjustable range. The platform marks the status of this node as "unrecovered response degradation" and writes it into the fault log structure as the basis for subsequent manual maintenance or node replacement tasks. The log content includes the number of regulation rounds, the final parameter value, the trajectory offset situation, and the determination time point. Embodiment

[0054] The difference between Embodiment 2 and Embodiment 1 of the present invention is that this embodiment introduces an Internet of Things intelligent lighting control system.

[0055] Figure 2 The structural schematic diagram of an Internet of Things intelligent lighting control system according to the present invention is given. An Internet of Things intelligent lighting control system includes a data access module, a trajectory splicing module, a subnet clustering module, a behavior matching and recognition module, and a parameter regulation module: Data access module: used to divide the building lighting area, establish the mapping relationship between lamps and areas, and complete the access of Internet of Things nodes of voice-controlled lamps and the reporting of response event data; Trajectory splicing module: sorts the response time series data, screens node pairs, and performs sliding window splicing processing to generate a set of complete behavior trajectories with continuity and clear time sequence for subsequent recognition; Subnet clustering module: constructs a response subnet through the co-occurrence relationship of trajectory nodes, and generates a behavior recognition model by training a classifier for the trajectory sequence based on the lighting area and time period characteristics; Behavior matching and recognition module: performs sliding window matching between the real-time trajectory and the behavior template, extracts abnormal response segments based on the edit distance, and constructs a candidate list of abnormal nodes for regulation positioning; Parameter regulation module: sends sensitivity regulation instructions to abnormal nodes periodically, and judges the regulation completion status based on the behavior matching feedback or marks it as a response degradation node.

[0056] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0057] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0058] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0059] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0060] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0061] The module described as a separation component may or may not be physically separated. The component shown as a module may or may not be a physical module, and it may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0062] In addition, in each embodiment of this application, each functional module can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0063] If the above-mentioned function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such understanding, the technical solution of this application, in essence, or the part that contributes to the prior art or part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0064] As described above, this is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0065] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An Internet of Things-based intelligent lamp control method, characterized in that, It includes the following steps: S1: Define the response behavior state change event of the voice-controlled lamp, collect the response time series data including the response event timestamp and node identifier, perform response sequence sorting within the set monitoring period, retrieve the set of response node pairs with connectable relationships, and splice them into a complete response trajectory; S2: Calculate the co-occurrence frequency of trajectory nodes for the response trajectory and divide the response aggregation clusters, and organize the response aggregation clusters with physically connected closures into a response subnet structure; S3: Label the behavior tags for the trajectory sequences in the response subnet, and establish a classifier structure for outputting the classification of behavior tags; S4: Extract the stable core paragraphs of the response nodes in the trajectory sequence corresponding to the behavior tags as behavior template units, and combine the classifier structure and the behavior template units into a subnet-level behavior model; S5: Submit the real-time collected response time series data to the subnet-level behavior model for behavior matching determination, locate the abnormal segments that do not match successfully in the template structure, and construct a list of abnormal node candidates; S6: Send a sound pressure induction sensitivity parameter adjustment instruction to each abnormal node in the list of abnormal node candidates, and update the abnormal node status based on the behavior matching determination result of the adjusted abnormal node.

2. The method for controlling an intelligent lamp based on the Internet of Things according to claim 1, wherein In S1, defining the response behavior state change event of the voice-controlled lamp, collecting the response time series data including the response event timestamp and node identifier, performing response sequence sorting within the set monitoring period, retrieving the set of response node pairs with connectable relationships, and splicing them into a complete response trajectory specifically includes: Divide the lighting area of the area where the intelligent voice-controlled lamp is installed, establish a lighting area identifier and define the lighting area attribution of the voice-controlled lamp; Connect all voice-controlled lamps to a unified Internet of Things platform through a low-power wireless protocol, and configure a unique node identifier for each lamp; Define the response behavior state change event of the voice-controlled lamp, open the reporting interface for the response event at the platform end, and bind the reported data to the lighting area to form a basic data stream; The Internet of Things platform performs periodic polling on the reporting interfaces of all lamp nodes, and collects the response time series data including the response event timestamp and node identifier; Perform ascending time sorting on the response time series data to generate a node response sequence that is continuous in time within the set monitoring period; After completing the response sequence re-sorting, based on the spatial coordinate distance between the lighting areas corresponding to the response nodes and the set time interval threshold condition, retrieve the set of node pairs with connectable relationships; Perform a connection operation based on a sliding window on the set of node pairs, continuously traverse the set of node pairs to construct a response sequence segment with time and space continuity as the link condition; Perform trajectory splicing processing on the response sequence segment, and merge the response sequence segments that meet the set splicing conditions into a complete response trajectory and store it in the trajectory set through the front-back relationship of the node pair identifiers.

3. The method for controlling an intelligent lamp based on the Internet of Things according to claim 2, wherein, In S2, calculating the co-occurrence frequency of trajectory nodes for the response trajectory and dividing the response aggregation clusters, and organizing the response aggregation clusters with physically connected closures into a response subnet structure specifically includes: Perform response statistical processing on the response trajectory nodes in the trajectory set, and form a response co-occurrence relationship matrix between nodes by calculating the co-occurrence frequency of each node response in the trajectory set within the monitoring period; The distance tensor conversion operation is performed on the co-occurrence relationship matrix, and a normalized interconnection strength tensor is generated based on the ratio of the response co-occurrence times to the total number of trajectories, which is used to express the response coordination density between any two nodes; Density clustering is performed on the interconnection strength tensor, and the local density maximum location method is used to extract the response center node group in the response concentration area, and the response clustering cluster is divided with the node group as the center; Eliminate the response clusters with spatial interruptions and retain the cluster results with closed physical connectivity; The response clusters with closed physical connectivity are organized into response subnets, and the response subnet structure consists of response node identifiers and trajectory sequences carrying response time spans.

4. The method for controlling an intelligent lamp based on the Internet of Things according to claim 3, wherein, In S3, in order to label the behavior labels of the trajectory sequences in the response subnet, the classifier structure for outputting the behavior label classification is established, which specifically includes: The monitoring cycle is divided into several monitoring periods, and a behavior label is given to each trajectory based on the monitoring period to which the time span of the trajectory sequence in the response subnet belongs and the lighting area span corresponding to the trajectory; The labeled trajectory sequence of n monitoring cycles is input into the classifier structure, the selected random forest model is loaded with training samples, the lighting area span of the trajectory node response and the distribution of the monitoring period to which it belongs are used as training features, and the behavior label classifier parameter group of the response subnet is output; Using the same labeled trajectory sequence as the validation set, the behavioral consistency of the labels of the trained classifier structure is evaluated using a ten-fold crossover method, and classifiers whose accuracy does not meet the set indicators are eliminated.

5. The method for controlling an intelligent lamp based on the Internet of Things according to claim 4, wherein In S4, the stable core segments of the response nodes in the trajectory sequence corresponding to the behavior label are extracted as the behavior template unit, and the classifier structure and the behavior template unit are combined into a subnet-level behavior model, which specifically includes: Perform pattern summary extraction on the typical trajectory sequence set corresponding to each type of behavior label. By calculating the node response frequency distribution and trajectory segment repetition rate, extract the stable core segments of the response nodes in the trajectory sequence and record them as behavior template units. The classifier structure and the behavior template unit are combined into a subnet-level behavior model, which is mapped to the response node with the behavior label number as the index, and serves as a standard reference for abnormal behavior of the response node.

6. The method for controlling an intelligent lamp based on the Internet of Things according to claim 5, wherein, In S5, the response time series data collected in real time is submitted to the subnet-level behavior model behavior matching judgment, the abnormal fragments that are not successfully matched in the template structure are located, and the abnormal node candidate list is constructed, which specifically includes: The IoT platform performs behavioral matching on the response time series data collected in real time, and uses the behavior template unit output by the subnet-level behavior model as a reference for matching the trajectory sequence corresponding to the current response time series data; A fixed-length sliding window is applied to the trajectory sequence to compare it with the node sequence of the behavior template unit segment by segment, the node matching offset is calculated using the edit distance, and a matching deviation curve is generated; Based on the matching deviation curve, the peak recognition local response discontinuity is performed to locate the abnormal segments in the trajectory sequence that are not successfully matched in the template structure; The node index corresponding to the located abnormal fragment is mapped to the specific lamp node, and a candidate list of abnormal nodes under the current behavior trajectory is constructed.

7. The method for controlling an intelligent lamp based on the Internet of Things according to claim 6, wherein In S6, a sound pressure induction sensitivity parameter regulation instruction is sent to each abnormal node in the abnormal node candidate list, and the abnormal node status is updated based on the behavior matching determination result after regulation. Specifically, it includes: In the abnormal node candidate list, a regulation status record container and an adjustment times counter are bound to each abnormal node; In each monitoring period, the Internet of Things platform sends a sound pressure induction sensitivity parameter regulation instruction with a set regulation step size to the abnormal nodes that have not reached the upper limit of adjustable parameters; Perform behavior matching determination on the real-time response timing data after each parameter regulation. If the matching offset of the corresponding trajectory sequence of the current abnormal node in consecutive m time periods is less than the acceptable threshold of the matching offset, update the regulation status to regulation completed and stop the subsequent regulation tasks of this abnormal node; If the adjustment times count of the abnormal node reaches the set maximum allowed times, or the regulation parameter reaches the upper limit of adjustable parameters, set the status of this abnormal node to response degradation not recovered and write it into the fault log.

8. An Internet of Things-based intelligent lamp control system for implementing an Internet of Things-based intelligent lamp control method according to any one of claims 1-7, characterized in that, It includes a data access module, a trajectory splicing module, a subnet clustering module, a behavior matching and recognition module, and a parameter regulation module: Data access module: used to divide the building lighting area, establish the mapping relationship between lamps and areas, and complete the access of Internet of Things nodes of voice-controlled lamps and the reporting of response event data; Trajectory splicing module: sort, node pair screening and sliding window splicing processing are performed on the response timing data to generate a complete set of continuous and clearly timed behavior trajectories for subsequent identification; Subnet clustering module: construct a response subnet through the co-occurrence relationship of trajectory nodes, and generate a behavior recognition model by training a classifier for the trajectory sequence based on the lighting area and time period characteristics; Behavior matching and recognition module: perform sliding window matching between the real-time trajectory and the behavior template, extract abnormal response segments based on the edit distance, and construct an abnormal node candidate list for regulation and positioning; Parameter regulation module: send sensitivity regulation instructions to abnormal nodes periodically, and judge the regulation completion status or mark it as a response degradation node based on the behavior matching feedback.

9. A computer device, characterized in that, It includes a display screen, a memory, a processor, and a computer program. The computer program is stored in the memory. When the computer program is executed by the processor, the processor executes the steps of a method for controlling intelligent lamps based on the Internet of Things according to any one of claims 1-7.

10. A readable storage medium, characterized in that, The computer program is stored on the readable storage medium. When the computer program is executed by the processor, the processor executes the steps of a method for controlling intelligent lamps based on the Internet of Things according to any one of claims 1-7.

Images