Lighting system alarm linkage response method and system based on regional intrusion detection

Abstract

The application relates to the technical field of intrusion detection, and discloses a lighting system alarm linkage response method and system based on regional intrusion detection, which comprises the following steps: positioning an intruder according to detection signals of a plurality of sensors in a monitoring area to obtain fusion position data; creating predefined hierarchical boundary data according to the monitoring area, and determining layered area parameters according to the fusion position data and the predefined hierarchical boundary data; performing trajectory prediction on a current motion state of the intruder according to the fusion position data to obtain trajectory prediction data, and performing bidirectional lag determination across hierarchical boundaries in combination with the layered area parameters to obtain hierarchical conversion control instructions; and generating a target lighting control strategy based on the fusion position data and the hierarchical conversion control instructions. The application can find an optimal balance point between threat prevention effect and energy consumption, realizes intelligent energy-saving control, guarantees security and protection effect, and reduces operation cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intrusion detection, and in particular to a lighting system alarm linkage response method and system based on regional intrusion detection. BACKGROUND

[0002] The traditional intrusion detection lighting system adopts a single sensor triggering mechanism and can only realize a simple "all-on or all-off" binary response mode, which cannot realize fine-grained hierarchical control according to the actual threat level and spatial position of the intruder. There are technical problems of response lag and frequent boundary switching when determining the position of the intruder. Due to the lack of effective motion trajectory prediction and boundary processing mechanism, when the intruder moves near the boundary of the monitoring area, the ping-pong effect is easily generated, which causes the lighting device to be frequently turned on and off, not only affecting the service life of the device, but also generating misleading alarm signals, thereby causing the accuracy of the prior art to be low. SUMMARY

[0003] The present application provides a lighting system alarm linkage response method and system based on regional intrusion detection, which can find the optimal balance point between threat prevention effect and energy consumption, realize intelligent energy-saving control, and not only ensure the security effect but also reduce the operating cost.

[0004] In a first aspect, the present application provides a lighting system alarm linkage response method based on regional intrusion detection, which comprises:

[0005] locating the intruder according to the detection signals of the plurality of sensors in the monitoring area to obtain fusion position data;

[0006] creating predefined hierarchical boundary data according to the monitoring area, and determining hierarchical region parameters according to the fusion position data and the predefined hierarchical boundary data;

[0007] performing trajectory prediction on the current motion state of the intruder according to the fusion position data to obtain trajectory prediction data, and performing bidirectional lag determination across the hierarchical boundary in combination with the hierarchical region parameters to obtain hierarchical conversion control instructions;

[0008] generating a target lighting control strategy based on the fusion position data and the hierarchical conversion control instructions.

[0009] In combination with the first aspect, in a first implementation manner of the first aspect of the present application, the fusion position data is obtained by locating the intruder according to the detection signals of the plurality of sensors in the monitoring area, and the fusion position data comprises:

[0010] acquire a heat source detection signal of an infrared sensor, a motion recognition signal of a microwave sensor, a target classification signal of an image recognition sensor and a distance measurement signal of a laser ranging sensor in a monitoring area;

[0011] add a time stamp mark to the heat source detection signal, the motion recognition signal, the target classification signal and the distance measurement signal respectively and perform time stamp alignment to obtain a time synchronization detection signal;

[0012] determine an azimuth angle and a pitch angle of an intruder relative to a protection target based on the time synchronization detection signal, and perform a trigonometric function operation based on the azimuth angle and the pitch angle to obtain a three-dimensional space position including an abscissa, an ordinate and a height coordinate;

[0013] perform a difference calculation on the three-dimensional space position of consecutive frames to obtain an inter-frame displacement, and divide the inter-frame displacement by a first time interval to obtain fusion position data.

[0014] In combination with the first aspect, in a second implementation manner of the first aspect of the application, the method further comprises:

[0015] pre-establishing a peripheral warning layer, a medium-distance early warning layer, a short-distance alarm layer, a dangerous approach layer and a core protection layer with the protection target as the center and determining pre-defined hierarchical boundary data;

[0016] extracting the abscissa and the ordinate of the intruder from the fusion position data, and calculating a radial distance of the intruder to the protection target based on the abscissa and the ordinate;

[0017] comparing the radial distance with the pre-defined hierarchical boundary data in a numerical manner to determine a current hierarchical identifier of the intruder, and searching for corresponding hierarchical region parameters according to the current hierarchical identifier.

[0018] In combination with the first aspect, in a third implementation manner of the first aspect of the application, the method further comprises:

[0019] determining a coordinate difference value between adjacent frames according to the fusion position data, and calculating a speed sequence data by dividing the coordinate difference value between adjacent frames by a second time interval;

[0020] calculating an acceleration vector based on a difference value of adjacent velocity vectors in the speed sequence data, and judging a trajectory curvature degree according to the acceleration vector;

[0021] According to the trajectory bending degree, the intruder movement behavior is identified as a straight line approach, a curve around, a jump movement or a wandering observation mode, and a trajectory curve fitting is performed to obtain a trajectory fitting result;

[0022] According to the trajectory fitting result, the current position coordinates, the current approach speed and the acceleration trend are substituted into a second-order kinematics prediction equation to calculate the spatial position of the intruder at the next time and the estimated time of reaching each level boundary, and trajectory prediction data is obtained;

[0023] The trajectory prediction data and the hierarchical region parameters are subjected to a two-way hysteresis determination across the level boundary to obtain a level conversion control instruction.

[0024] In combination with the first aspect, in a fourth implementation manner of the first aspect of the application, the two-way hysteresis determination of the trajectory prediction data and the hierarchical region parameters across the level boundary to obtain the level conversion control instruction comprises:

[0025] According to the trajectory prediction data, the radial distance of the spatial position at the next time and the current position is compared to obtain a movement direction determination result;

[0026] Based on the level boundary distance in the hierarchical region parameters, two-way hysteresis boundary parameters of each level are calculated respectively;

[0027] The radial distance of the intruder is compared with the two-way hysteresis boundary parameters to obtain a level crossing confirmation state;

[0028] When the level crossing confirmation state is a confirmation crossing upgrade, the output level conversion control instruction is a control instruction for increasing the lighting intensity of one level; when the level crossing confirmation state is a confirmation crossing downgrade, the output level conversion control instruction is a control instruction for reducing the lighting intensity of one level; and when the level crossing confirmation state is to maintain the current level, the output level conversion control instruction is a control instruction for maintaining the existing lighting parameters.

[0029] In combination with the first aspect, in a fifth implementation manner of the first aspect of the application, the two-way hysteresis boundary parameters of each level are calculated based on the level boundary distance in the hierarchical region parameters, comprising:

[0030] The level boundary distances of the outer warning level, the medium-distance early warning level, the near-distance warning level and the dangerous approach level are extracted from the hierarchical region parameters;

[0031] According to the level boundary distance and a preset target value, an inner hysteresis boundary distance is obtained by subtraction operation, and an outer hysteresis boundary distance is obtained by addition operation;

[0032] The inner lag boundary distance and the outer lag boundary distance are subjected to boundary checking to obtain a bidirectional lag boundary parameter.

[0033] In combination with the first aspect, in a sixth implementation manner of the first aspect of the application, the numerical comparison of the radial distance of the intruder with the bidirectional lag boundary parameter to obtain a level crossing confirmation state comprises:

[0034] Finding a corresponding current level boundary numerical combination in the bidirectional lag boundary parameter;

[0035] Comparing the radial distance of the intruder with the inner lag boundary distance and the outer lag boundary distance in the current level boundary numerical combination respectively, recording a boundary crossing determination result as an inner boundary crossing when the radial distance is less than the inner lag boundary distance, recording a boundary crossing determination result as an outer boundary crossing when the radial distance is greater than the outer lag boundary distance, and recording a boundary crossing determination result as a boundary stay when the radial distance is between the two boundaries;

[0036] Combining the boundary crossing determination result with the motion direction determination result to determine a level crossing confirmation state, the level crossing confirmation state comprising confirming crossing upgrade, confirming crossing downgrade and maintaining the current level.

[0037] In combination with the first aspect, in a seventh implementation manner of the first aspect of the application, the generating a target lighting control strategy based on the fused position data and the level transition control instruction comprises:

[0038] Constructing a reinforcement learning state vector of the intruder according to the fused position data, and defining the lighting brightness adjustment, the color temperature change and the strobe setting in the level transition control instruction as a reinforcement learning action space;

[0039] Constructing a reinforcement learning environment based on the reinforcement learning state vector and the reinforcement learning action space;

[0040] In the reinforcement learning environment, calculating a reward function value based on the threat prevention success rate and the corresponding energy consumption data after executing the level transition control instruction, and taking the reward function value as a feedback signal of a reinforcement learning algorithm to obtain an action execution reward evaluation index;

[0041] According to the action execution reward evaluation index, configuring corresponding brightness percentage, color temperature Kelvin value and strobe frequency parameters for the intruder at each level respectively to obtain a target lighting control strategy.

[0042] In combination with the first aspect, in an eighth implementation form of the first aspect of the application, in the reinforcement learning environment, a reward function value is calculated based on the threat blocking success rate after the execution of the hierarchical transition control instruction and corresponding energy consumption data, and the reward function value is taken as a feedback signal of the reinforcement learning algorithm to obtain an action execution reward evaluation index, including:

[0043] In the reinforcement learning environment, the threat blocking success rate after the execution of the hierarchical transition control instruction is counted, and the current and voltage values of the lighting device during the execution of the hierarchical transition control instruction are monitored to obtain corresponding energy consumption data;

[0044] The reward function value is taken as a reward function value, which is the difference between the threat blocking success rate and the product of the energy consumption weight coefficient and the energy consumption data;

[0045] According to the positive and negative nature and the value size of the reward function value, a feedback signal of the reinforcement learning algorithm is generated, and an action execution reward evaluation index is calculated according to the feedback signal.

[0046] In a second aspect, the application provides a lighting system alarm linkage response system based on regional intrusion detection, which comprises:

[0047] A positioning module is configured to position an intruder according to detection signals of a plurality of sensors in a monitoring area to obtain fusion position data;

[0048] A creation module is configured to create predefined hierarchical boundary data according to the monitoring area, and determine layered region parameters according to the fusion position data and the predefined hierarchical boundary data;

[0049] A prediction module is configured to perform trajectory prediction on a current motion state of the intruder according to the fusion position data to obtain trajectory prediction data, and perform bidirectional lag determination across the hierarchical boundary in combination with the layered region parameters to obtain a hierarchical transition control instruction;

[0050] A generation module is configured to generate a target lighting control strategy based on the fusion position data and the hierarchical transition control instruction.

[0051] The technical scheme provided by the application realizes a technical leap from single triggering to stereoscopic perception through multi-sensor cooperative positioning technology, and significantly improves the invader positioning accuracy and reliability. The five-level depth area division mechanism is established to cooperate with the S-shaped weight gradual function, to realize the smooth and continuous change of the lighting intensity with the invasion depth, and completely solve the technical defects of the traditional "all bright or all dark" mutation control. The second-order kinematic prediction algorithm and the multiple motion mode recognition technology realize a breakthrough from passive response to active prediction, and preset the lighting parameters 2-3 seconds in advance to ensure the timeliness of the response. The bidirectional hysteresis boundary processing mechanism effectively avoids the problem of frequent switching of the level boundaries, and significantly improves the system stability. The lighting strategy optimization engine based on reinforcement learning can automatically identify threat patterns and continuously optimize the response strategy, realizing a technical leap from fixed mode to adaptive learning. The application accurately identifies four threat types and configures differentiated lighting strategies, finds the optimal balance point between threat prevention effect and energy consumption through the energy consumption-effect evaluation matrix, and realizes intelligent energy-saving control. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creating labor based on these drawings.

[0053] Figure 1 The step schematic diagram of the lighting system alarm linkage response method based on regional intrusion detection in the embodiment of the application;

[0054] Figure 2 The structure schematic diagram of the lighting system alarm linkage response system based on regional intrusion detection in the embodiment of the application. DETAILED DESCRIPTION

[0055] The embodiment of the application provides a lighting system alarm linkage response method and system based on regional intrusion detection. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0056] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 One embodiment of the alarm linkage response method of the lighting system based on regional intrusion detection in the embodiments of the present application includes:

[0057] Step S1, positioning the intruder according to the detection signals of multiple sensors in the monitoring area to obtain fused position data;

[0058] Specifically, a plurality of heterogeneous sensors are deployed in the monitoring area, including an infrared sensor for capturing heat source features, a microwave sensor for identifying small motion changes, an image recognition sensor for performing image classification and contour recognition, and a laser ranging sensor for providing high-precision ranging information, and heat detection signals, motion recognition signals, target classification signals and distance measurement signals output by these sensors are collected respectively. Add a uniform format timestamp to all signals to mark the collection time of each signal, and perform timestamp alignment processing on these signals through a time synchronization mechanism to form a synchronized data set under the same observation window. The spatial orientation information of the intruder relative to the protected target is calculated using the time-aligned detection signals, including two key parameters, azimuth and elevation. By analyzing the spatial geometric distribution characteristics in the infrared and laser ranging signals, combined with the shape direction judgment result of the image recognition sensor on the intruder target, the orientation relationship of the intruder is derived. Based on the obtained azimuth and elevation, using trigonometric function operation means, combined with the distance value provided in the laser ranging signal, the three-dimensional spatial position of the intruder in the current frame is analyzed, including horizontal coordinate, vertical coordinate and height coordinate. The three-dimensional spatial position of the continuous frame is calculated by difference, and the inter-frame displacement is extracted. Divide the inter-frame displacement by the corresponding time interval to obtain the speed vector per second, and form the dynamic fused position data.

[0059] Step S2, creating pre-defined hierarchical boundary data according to the monitoring area, and determining hierarchical region parameters according to the fused position data and the pre-defined hierarchical boundary data;

[0060] Specifically, the fixed position of the protection target in the monitoring area is taken as the center, and a plurality of annular regions in the form of concentric circles are established outwardly around the center in sequence, respectively forming a peripheral warning layer, a medium-distance early warning layer, a close-distance warning layer, a dangerous approach layer and a core protection layer. Each layer has a clear radial boundary range, and is solidified as pre-defined layer boundary data in numerical form. The pre-defined layer boundary data is set as several continuous intervals, wherein the core protection layer is 0-5 meters, the dangerous approach layer is 5-15 meters, the close-distance warning layer is 15-30 meters, the medium-distance early warning layer is 30-50 meters, and the peripheral warning layer is 50-100 meters. According to the fusion position data obtained in real time, the horizontal coordinate and the vertical coordinate of the intruder in the three-dimensional space are extracted, and the center coordinate of the protection target is taken as the reference point. The radial distance of the current intruder relative to the protection target is calculated by the Euclidean distance formula, and the three-dimensional position data is simplified to a single distance scale to match the layered boundary structure. The radial distance is used as the basis for judgment, and is compared with the preset boundary of each layer in numerical value. It is judged which boundary interval the radial distance falls into, and the layer identification of the intruder is confirmed. Through the layer identification, the lighting strategy parameters bound to the corresponding layer are extracted from the established layered region parameter library, such as lighting brightness ratio, color temperature setting, response mode and linkage subsystem control value, etc.

[0061] Step S3, according to the fusion position data, the current motion state of the intruder is predicted to obtain trajectory prediction data, and the layered region parameters are combined to make a two-way lag judgment across the layer boundary to obtain a layer conversion control instruction;

[0062] Specifically, according to the spatial coordinates of the continuous sampling frames in the fused position data, the coordinate difference value between adjacent frames is calculated to reflect the actual displacement of the intruder in unit time, and by dividing the coordinate difference value by the time interval between two frames, a velocity vector sequence representing the motion speed and direction is obtained. The difference operation is performed on the continuous velocity vectors in the velocity sequence to calculate the acceleration vector reflecting the change of the motion state, and the curvature characteristics of the intrusion path are judged based on the direction change amplitude and frequency change trend of the acceleration. If the acceleration change direction is stable and the amplitude is small, it is determined as a straight-line approach mode; if the direction changes continuously and the path deviates obviously, it is determined as a curve around; if the speed changes suddenly and is accompanied by instantaneous high displacement, it is classified as a jumping movement; if the motion direction and speed repeatedly fluctuate in a small range, it is identified as a wandering observation mode. The trajectory fitting model matched with the motion mode is selected to perform fitting calculation on the collected position data, to establish the mathematical expression of the current trajectory of the intruder and generate the trajectory fitting result; for example, linear fitting is used for straight-line approach, quadratic or cubic curve fitting is used for curve around, and sudden point analysis and probability distribution modeling are introduced for jumping and wandering behaviors. The current spatial position, approach speed and acceleration trend are substituted into the second-order kinematics prediction equation to obtain the predicted position of the intruder in the future time window through simulation calculation, and the estimated time for the intruder to reach each level boundary from the current position is calculated, and the trajectory prediction data is output. In order to avoid the system from producing excessive response or frequent switching of lighting due to the frequent back-and-forth movement of the intruder near the boundary, the upper and lower floating ranges are established at each level boundary according to the pre-defined buffer width, and the bidirectional hysteresis judgment logic is introduced: when the trajectory prediction result shows that the intruder will continuously enter the inside of a level boundary beyond a specified depth threshold (e.g. 2.5 meters) and maintain for a sufficient time, the lighting strategy switching to the inner level is triggered; on the contrary, when the predicted trajectory shows that it continuously exits the level boundary and reaches the corresponding outer threshold position, the system only performs the degradation operation. Through the combination of delay judgment and displacement confirmation, the level conversion control instruction is output.

[0063] Step S4, generating a target lighting control strategy based on the fused position data and the level conversion control instruction.

[0064] Specifically, the reinforcement learning state vector of the intruder is constructed according to the fusion position data. It includes the horizontal coordinate, vertical coordinate and height coordinate information of the intruder at the current moment, and integrates dynamic variables such as radial distance, moving speed, trajectory curvature, current level identifier and historical path characteristics, forming a reinforcement learning state vector with spatial position and motion characteristics. At the same time, the lighting brightness adjustment, color temperature change and stroboscopic mode setting contained in the level conversion control instruction are defined as the action space of the reinforcement learning model, wherein the brightness adjustment range covers the continuous value domain from 10% to 100%, the color temperature change is controlled between 2700K and 6500K, and the stroboscopic setting includes three response modes of static lighting, low frequency flicker and high frequency flicker. Based on the reinforcement learning state vector and the reinforcement learning action space, the reinforcement learning environment is constructed. The reinforcement learning environment takes the intervention effect of the current lighting response behavior on the intruder behavior as the basis, takes the behavior result obtained after the actual execution of the control action as the learning feedback, and dynamically generates reinforcement learning samples in the environment. Each system response process is accompanied by energy consumption of lighting resources and intervention results of intruder situation, and through real-time evaluation of whether the intruder stops, retreats or continues to approach after the execution of the control action, the threat prevention success rate of this response is determined; at the same time, the energy consumption data is formed by recording the consumed energy and the total amount of resources in the response period. According to the threat prevention effect and the energy consumption cost, the reward function is constructed, and the "effect score minus weighted energy consumption cost" is used for quantitative evaluation, so that the response strategy with high efficiency and low energy consumption obtains higher reward value, and vice versa, which is punished, ensuring that the system gradually converges to the optimal lighting control strategy in the learning process. The reward function value is used as the feedback signal in the reinforcement learning algorithm to update the Q value, the policy function or the neural network parameters, forming a lighting response optimization mechanism with sustainable learning ability. After completing multiple rounds of training and sample accumulation, according to the action execution reward evaluation index learned, the optimal response strategy is configured for the intrusion situation at different levels. The lighting system is set to 10% brightness, 3000K warm white color temperature and static lighting mode at the outer warning level to realize low interference and low energy consumption warning; 25% brightness, 4000K neutral white color temperature and low frequency flicker are set at the medium distance warning level to improve visibility and attract the attention of the intruder; 50% brightness, 5000K cold white color temperature and high frequency flicker are configured at the close distance warning level to form a strong visual deterrent effect; and 75% and 100% brightness are respectively increased at the dangerous approach layer and the core protection layer, and the red and blue alternating flashing and strong sound linkage are matched to achieve the defense purpose of synchronous triggering of suppression, driving and alarm.

[0065] In a specific embodiment, the process of performing step S1 can specifically include the following steps:

[0066] acquire a heat source detection signal of an infrared sensor, a motion recognition signal of a microwave sensor, a target classification signal of an image recognition sensor and a distance measurement signal of a laser ranging sensor in a monitoring area;

[0067] add a time stamp mark to the heat source detection signal, the motion recognition signal, the target classification signal and the distance measurement signal respectively and perform time stamp alignment to obtain a time synchronization detection signal;

[0068] determine an azimuth angle and a pitch angle of an intruder relative to a protected target based on the time synchronization detection signal, and perform a trigonometric function operation based on the azimuth angle and the pitch angle to obtain a three-dimensional space position including an abscissa, an ordinate and a height coordinate;

[0069] perform a difference calculation on the three-dimensional space positions of consecutive frames to obtain an inter-frame displacement, and divide the inter-frame displacement by a first time interval to obtain fused position data.

[0070] Specifically, a multi-type sensor network covering the entire monitoring area is deployed, and signals including a heat source detection signal provided by an infrared sensor, a motion recognition signal output by a microwave sensor, a target classification signal captured by an image recognition sensor, and a distance measurement signal returned by a laser ranging sensor are obtained. The infrared signal is used to identify active heat sources and can stably detect typical heat targets such as humans and animals in a complex lighting environment; the microwave signal realizes displacement change perception of dynamic targets based on the Doppler effect and has high penetration ability and anti-interference characteristics; the image recognition signal obtains the shape, posture and category characteristics of the intruding target through video stream analysis; and the laser ranging signal provides extremely high precision target distance information based on the time-of-flight principle. By simultaneously obtaining the above signals and taking them as preliminary perception data of the system, multi-source original information about target existence, motion state, category characteristics and spatial distance in multiple dimensions is obtained. A uniform time stamp is added to the data stream from different sensors to ensure that all signals are comparable in the time domain. Due to the different sampling frequencies of various sensors and the physical delay of some sensors, a unified clock source is used for time synchronization processing of various signals, all sensor outputs are re-calibrated to a unified time reference through a clock synchronization module or a network time protocol, and linear interpolation, sample retention or nearest neighbor algorithm is used to realize time stamp alignment, so that each sensor signal has effective observation values at the same time. After time synchronization, a time-synchronized detection signal set is formed. Based on the time-synchronized detection signal set, the radial distance between the target and the protection center is determined using laser ranging data, and the image perspective and pixel coordinates obtained by the image recognition module, the motion direction output by the microwave sensor, and the spatial response distribution of the heat source distribution by the infrared sensor are combined to derive the pitch angle and azimuth angle information of the target in the monitoring scene. By setting the protected target as a spatial reference center, the position of the target in the image and the distance provided by the laser ranging are input, the horizontal position and vertical position of the intruding target in the monitoring coordinate system are calculated by inverse calculation through trigonometric functions, and the vertical coordinate of the target is derived in combination with the height setting of the sensor array, the three-dimensional spatial position of the target in the current frame is calculated. The three-dimensional spatial coordinates collected in continuous multiple frames are processed by frame-by-frame difference, and the displacement vector between adjacent frames is calculated. The size and direction of the displacement between each frame reflect the motion characteristics of the intruder in unit time, and by dividing the inter-frame displacement vector by the time interval between adjacent frames, the average speed vector in this time period is obtained, and the fusion position data formed is a multi-dimensional data set including horizontal coordinate, vertical coordinate, height, speed size and speed direction.

[0071] In a specific embodiment, the process of performing step S2 can specifically include the following steps:

[0072] A peripheral warning layer, a medium-distance early warning layer, a close-distance alarm layer, a dangerous approach layer and a core protection layer are pre-established with the protected target as the center, and pre-defined layer boundary data is determined.

[0073] extracting the horizontal coordinate and the vertical coordinate of the intruder from the fused position data, and calculating a radial distance of the intruder to the protected target based on the horizontal coordinate and the vertical coordinate;

[0074] numerically comparing the radial distance with predefined hierarchical boundary data, judging a current belonging hierarchical identifier of the intruder, and searching for corresponding hierarchical region parameters according to the current belonging hierarchical identifier.

[0075] Specifically, a spatial hierarchical structure is established according to the radial distance with the physical coordinate of the protected target as the center. The spatial hierarchical structure divides the entire monitoring area into a plurality of concentric regions, from outside to inside, as a peripheral warning layer, a middle-distance early warning layer, a close-distance alarm layer, a dangerous approach layer, and a core protection layer. The boundary range of each layer is preset and stored as static predefined hierarchical boundary data in meters. The peripheral warning layer is defined as a region between 50 and 100 meters, which is used for preliminary detection and early warning at a long distance. The middle-distance early warning layer is defined as 30 to 50 meters, which is used to strengthen the monitoring coverage of the target approaching direction. The close-distance alarm layer is 15 to 30 meters, which is used for high sensitivity response before entering the core warning area. The dangerous approach layer is 5 to 15 meters, which indicates that the intruder is very close to the core target and needs to enter a high response state. The core protection layer is 0 to 5 meters, which is the highest defense circle layer with the highest defense strength and the fastest response. The horizontal coordinate and the vertical coordinate information of the intruder in the current frame are extracted from the fused position data, and the position of the protected target is taken as the origin of the coordinate system. The radial distance of the intruder relative to the protected target is calculated using the Euclidean distance formula. The square values of the horizontal coordinate and the vertical coordinate are added and the square root is taken to obtain a scalar distance value reflecting the spatial proximity. The radial distance is compared with the predefined hierarchical boundary data layer by layer, and the current distance value is determined to be in which hierarchical interval in the order from the core layer to the outside. If the current distance is less than 5 meters, it belongs to the core protection layer; if it is between 5 and 15 meters, it is divided into the dangerous approach layer; if it is between 15 and 30 meters, it is identified as the close-distance alarm layer; if it is between 30 and 50 meters, it belongs to the middle-distance early warning layer; if it is between 50 and 100 meters, it enters the peripheral warning layer. If the distance exceeds 100 meters, it is determined that the target has not entered the effective response range of the monitoring area, and no response strategy is triggered. According to the current belonging hierarchical identifier of the intruder, the corresponding parameter set is called from the preset hierarchical region parameter library, including lighting response brightness percentage, color temperature setting value, flicker mode configuration, linkage alarm activation state, and monitoring video trigger condition. Different levels correspond to different response intensities. For example, the lighting brightness of the peripheral warning layer is set to 10%, the color temperature is set to 3000K warm white light, the stroboscopic mode is closed, and only the intrusion trajectory is recorded. The core protection layer corresponds to 100% brightness, cold white to blue white mixed color temperature, red and blue alternating high frequency flicker, high intensity sound and light alarm, communication notification linkage, and other parameter configurations to maximize the intrusion deterrence and security protection.

[0076] In the embodiment, the corresponding hierarchical region parameter is found according to the current belonging level identifier, including: extracting the corresponding reference lighting intensity threshold from the preset level lighting mapping table based on the current belonging level identifier, wherein the peripheral warning layer corresponds to 10% reference lighting intensity, the medium-distance early warning layer corresponds to 25% reference lighting intensity, the near-distance warning layer corresponds to 50% reference lighting intensity, the dangerous approaching layer corresponds to 75% reference lighting intensity, and the core protection layer corresponds to 100% reference lighting intensity, to obtain the level reference lighting parameter; the radial distance is substituted into the S-type weight progressive function to calculate the depth weight coefficient, the S-type weight progressive function is a sigmoid function of the intrusion depth distance, the function curve shape is adjusted by controlling the steepness parameter, the maximum detection distance parameter and the level step parameter, to obtain the depth weight coefficient corresponding to the intrusion depth; the depth weight coefficient is used for weight modulation calculation on the level reference lighting parameter, the depth weight coefficient is multiplied by the reference lighting intensity threshold to obtain the accurate lighting intensity value corresponding to the current position, to ensure that the lighting intensity changes smoothly and progressively with the intrusion depth instead of suddenly switching, to obtain the depth modulation lighting intensity; an overlapping buffer area with a preset width is set at the boundary of adjacent levels, when the intruder is located in the buffer area, the transition lighting intensity in the buffer area is calculated by a linear interpolation algorithm, to realize soft boundary smooth transition of the lighting intensity between adjacent levels, to obtain the boundary transition lighting parameter; the depth modulation lighting intensity and the boundary transition lighting parameter are combined to form a complete parameter combination including the current level identifier, the accurate lighting intensity value, the color temperature adjustment range and the boundary buffer parameter, to obtain the final hierarchical region parameter.

[0077] In a specific embodiment, the process of performing step S3 can specifically include the following steps:

[0078] According to the fusion position data, the coordinate difference between adjacent frames is determined, and the speed sequence data is calculated by dividing the coordinate difference between adjacent frames by the second time interval;

[0079] Based on the difference between adjacent velocity vectors in the speed sequence data, the acceleration vector is calculated, and the trajectory curvature is judged according to the acceleration vector;

[0080] According to the trajectory curvature, the intruder motion behavior is identified as a straight-line approach, a curve around, a jump movement or a wandering observation mode, and a trajectory curve fitting is performed to obtain a trajectory fitting result;

[0081] According to the trajectory fitting result, the current position coordinates, the current approach speed and the acceleration trend are substituted into the second-order kinematics prediction equation to calculate the spatial position of the intruder at the next moment and the estimated time of reaching each level boundary, to obtain trajectory prediction data;

[0082] The trajectory prediction data and the hierarchical region parameters are bidirectionally lagged to determine a cross-level boundary, and a level conversion control instruction is obtained.

[0083] Specifically, based on the fusion position data, time series analysis is performed, and the spatial coordinates between consecutive frames are subjected to difference operation to obtain the displacement vector of the target in unit time. The position state of the intruder at different time points is represented in three-dimensional coordinate form, the coordinate points of adjacent frames are subjected to component-by-component subtraction, the difference amounts in the horizontal, vertical and vertical directions are calculated, and then the three-dimensional coordinate difference vector is divided by the time interval between adjacent frames to obtain the velocity vector, i.e. the velocity sequence data. The velocity data at multiple time points are continuously accumulated to form a velocity trajectory set reflecting the motion change trend of the target. Based on the velocity sequence data, difference calculation is performed on two consecutive velocity vectors to obtain the velocity change rate, i.e. the acceleration vector. By analyzing the direction change and amplitude change trend of the acceleration vector, the spatial change degree of the motion path, i.e. the curvature of the trajectory, is derived. When the acceleration change is minimal or the direction consistency is strong, it indicates that the intruder presents a straight-line approaching feature; if the acceleration direction gradually deviates and the path changes continuously, it is a curve circling behavior; when the acceleration suddenly changes and is interspersed with obvious speed discontinuity, it is classified as a jumping movement; if the acceleration direction frequently reverses, the speed fluctuation is significant, and the motion range is limited within a certain area, it is identified as a wandering observation mode. According to the above judgment results, the motion behavior of the intruder is classified. Based on the current behavior mode, the corresponding trajectory fitting algorithm is selected and the fusion position data is subjected to curve fitting. For the straight-line approaching mode, linear least squares fitting is adopted, for the curve circling, quadratic or cubic polynomial fitting is adopted, for the jumping movement, piecewise fitting and mutation detection mechanism are introduced, and for the wandering observation mode, probability model or path distribution analysis based on clustering is more suitable. The core of the fitting process is to extract the spatial structure features of the path completed by the intruder and construct a predictable motion trend model. After fitting, a set of function parameters expressing the path trend are obtained as the structural description of the motion trend. Combined with the trajectory fitting result, the three-dimensional spatial coordinates at the current time, the current approaching speed and the trend value extracted from the foregoing acceleration sequence are substituted into the standard second-order kinematics prediction equation to realize the prospective calculation of the spatial position of the target in the next prediction time window. The prediction time window is set to 1-2 seconds, based on which the position of the target at multiple future time points is calculated, and trajectory prediction data are obtained, including the predicted path, the predicted endpoint, the speed evolution trend and the estimated time point of reaching each hierarchical boundary under the current path. The predicted path is compared and analyzed with the pre-defined hierarchical region parameters to judge whether the intruder may cross the current hierarchical boundary in the future.To avoid frequent response switching caused by path oscillation or target critical wandering, a two-way hysteresis decision mechanism is adopted, i.e. two buffers are introduced at each level boundary, which are set to 2.5 meters inward and outward respectively. When the predicted path enters the buffer of the next level and meets the requirements of position stability and motion trend consistency, the inward level switching control is triggered. Conversely, when the predicted path shows that the target will exit the current level and stably stay outside the buffer of the previous level, the system triggers the downgrade operation, thereby avoiding the "ping-pong effect" caused by short-term disturbance or measurement noise. The output level transition control instruction determines the change of lighting response level, and synchronously mobilizes the associated systems such as sound-light alarm, monitoring frame rate, communication notification mode, etc., forming a proactive linkage strategy based on prediction driving.

[0084] In a specific embodiment, the process of performing step of two-way hysteresis decision on trajectory prediction data and hierarchical region parameters across level boundaries to obtain level transition control instruction can specifically include the following steps:

[0085] According to the trajectory prediction data, compare the radial distance between the spatial position at the next time and the current position to obtain the motion direction decision result;

[0086] Based on the level boundary distance in the hierarchical region parameters, calculate the two-way hysteresis boundary parameters of each level respectively;

[0087] Numerically compare the radial distance of the intruder with the two-way hysteresis boundary parameters to obtain the level crossing confirmation state;

[0088] When the level crossing confirmation state is confirmed to cross up, output the level transition control instruction as the control instruction of increasing the lighting intensity of one level; when the level crossing confirmation state is confirmed to cross down, output the level transition control instruction as the control instruction of reducing the lighting intensity of one level; when the level crossing confirmation state is to maintain the current level, output the level transition control instruction as the control instruction of maintaining the existing lighting parameters.

[0089] Specifically, the spatial position of the intruder at the next moment is obtained based on the trajectory prediction data, and the predicted radial distance is obtained by calculating the Euclidean distance between the three-dimensional coordinates of the predicted position and the spatial origin of the protected target. Meanwhile, the spatial coordinates at the current moment are extracted from the fused position data and the corresponding current radial distance is calculated. The predicted radial distance and the current radial distance are compared numerically. If the predicted distance is less than the current distance, it indicates that the intruder is approaching the protected target, i.e. the movement direction is "inward". If the predicted distance is greater than the current distance, it indicates that the intruder is moving away from the protected target, i.e. the movement direction is "outward". According to the pre-defined hierarchical region parameters, the boundary distance corresponding to each level is extracted, and a lag boundary is set on both sides of each level boundary to construct a bidirectional buffer judgment mechanism. The lag boundary parameter is defined as a shift of a certain width, for example 2.5 meters, inward or outward from each level boundary, forming an "inner lag boundary" and an "outer lag boundary". For the boundary between the adjacent two levels, a transition zone with a width of 5 meters is generated, where the boundary position close to the inner layer minus the lag distance is defined as the "upgrade judgment limit", and the boundary position close to the outer layer plus the lag distance is defined as the "downgrade judgment limit", thus forming a judgment buffer zone in space to avoid frequent control switching when the intruder wanders in the boundary area. The actual radial distance of the intruder at the current moment is compared with the lag boundary parameter corresponding to the current level. If the movement direction is "inward" and the predicted radial distance is less than the "upgrade judgment limit" between the current level and the next level, and the intruder stays in this interval continuously for more than a preset threshold (e.g. 500 milliseconds), it is determined as "confirmed crossing upgrade". Conversely, if the movement direction is "outward" and the predicted radial distance is greater than the "downgrade judgment limit" between the current level and the previous level, and the intruder stays stably in this area, it is determined as "confirmed crossing downgrade". If the predicted position is inside the current level or near the boundary but not in the lag interval, or the composite criteria of position stability and consistent trend are not met, the state of "maintaining the current level" is maintained to avoid false judgment caused by fluctuations near the boundary. After completing the level crossing confirmation state judgment, the corresponding level conversion control command is generated according to the result. When the crossing upgrade is confirmed, the output control command is "increase the lighting intensity by one level", which specifically means that the lighting brightness is increased from the current value to the brightness configuration corresponding to the higher level, for example from 25% to 50%, and the additional measures such as higher level sound and light alarm, stroboscopic mode adjustment and monitoring video intensity enhancement are triggered simultaneously. When the crossing downgrade is confirmed, the "lower the lighting intensity by one level" command is output, which lowers the lighting parameter to the previous level preset value, such as reducing the brightness from 75% to 50%, and closes some high-frequency response modules to save energy and reduce noise. If the current level is maintained, the "maintain the existing lighting parameter" control command is output, and the lighting device maintains the current state operation without triggering any change operation to realize stable state control.

[0090] In a specific embodiment, the process of calculating the bidirectional lag boundary parameter of each level based on the level boundary distance in the hierarchical region parameter can specifically include the following steps:

[0091] Extracting the level boundary distance of the outer warning layer, the medium distance early warning layer, the close distance warning layer and the dangerous approaching layer from the hierarchical region parameter;

[0092] According to the level boundary distance and the preset target value, the inner lag boundary distance is obtained by subtraction operation, and the outer lag boundary distance is obtained by addition operation;

[0093] The inner lag boundary distance and the outer lag boundary distance are subjected to boundary check to obtain the bidirectional lag boundary parameter.

[0094] Specifically, the boundary positions corresponding to each defense level are extracted from the hierarchical configuration data, and the boundary positions are taken as the spatial reference center of the protection target to divide a plurality of concentric circle hierarchical structures by fixed radial distances. Among them, the boundary distances of the outer warning layer, the middle distance warning layer, the near distance warning layer and the dangerous approach layer represent the dividing lines between them and the next inner layer, and are set to 100 meters, 50 meters, 30 meters and 15 meters respectively. By setting a uniform hysteresis compensation distance (i.e. a preset target value, set to 2.5 meters), an additional judgment buffer zone is constructed inside and outside each level boundary. According to the subtraction operation of the level boundary distance and the preset target value, the "inner hysteresis boundary distance" is obtained, which represents the actual judgment lower limit required for entering the next level of defense, i.e. the intruder needs to approach the target more than 2.5 meters beyond the normal boundary to be judged as "real crossing"; at the same time, according to the addition operation of the level boundary distance and the preset target value, the "outer hysteresis boundary distance" is obtained, which represents the case that the intruder needs to exit the current level and move away from the boundary more than 2.5 meters, and is judged as "real withdrawal". Boundary checking operations are performed on all inner and outer hysteresis boundary distances to ensure that the generated bidirectional hysteresis boundary parameters have logical consistency and numerical validity in space. The inner hysteresis boundary value of each level is checked for non-negativity, and if the boundary distance of a certain level is less than the hysteresis value, the subtraction result is negative, and the minimum value of the boundary is set to 0 meters to prevent illegal radius; the inner and outer hysteresis boundaries of each layer are cross-checked with the space overlap between the upper and lower levels to ensure that the hysteresis boundaries of adjacent levels do not overlap, thereby avoiding ambiguity in judgment, for example, if the outer hysteresis boundary of a certain layer is less than the inner hysteresis boundary of the next layer, it means that the judgment intervals of the two levels overlap, and the priority or the hysteresis width is redistributed to solve the conflict; all hysteresis boundaries are trimmed and calibrated according to the effective monitoring radius of the current scene, and if the outer hysteresis boundary of a certain layer exceeds the monitoring boundary limit (such as greater than 100 meters), it is automatically reduced to the upper limit to avoid the judgment range exceeding the perception boundary and having no practical significance. After the boundary checking process, a set of "inner hysteresis boundary distance" and "outer hysteresis boundary distance" corresponding to each level is formed to constitute the bidirectional hysteresis boundary parameter table.

[0095] In a specific embodiment, the process of performing numerical comparison between the radial distance of the intruder and the bidirectional hysteresis boundary parameters to obtain the level crossing confirmation state can specifically include the following steps:

[0096] Finding the corresponding current level boundary value combination in the bidirectional hysteresis boundary parameters;

[0097] The radial distance of the intruder is compared with the inner lag boundary distance and the outer lag boundary distance in the current level boundary value combination, when the radial distance is less than the inner lag boundary distance, the boundary crossing determination result is recorded as an inner boundary crossing, when the radial distance is greater than the outer lag boundary distance, the boundary crossing determination result is recorded as an outer boundary crossing, and when the radial distance is between the two boundaries, the boundary crossing determination result is recorded as a boundary stay.

[0098] The boundary crossing determination result and the motion direction determination result are combined to determine the level crossing confirmation state, and the level crossing confirmation state includes confirmation crossing upgrade, confirmation crossing downgrade and maintaining the current level.

[0099] Specifically, the boundary value combination corresponding to the level where the current intruder is located is located in the bidirectional lag boundary parameter, that is, the inner lag boundary distance and the outer lag boundary distance corresponding to the level are found, and the two boundary values form a judgment interval in the current level, which is used to detect whether the intruder has exceeded the established level boundary buffer zone. The radial distance of the intruder is compared with the inner lag boundary distance and the outer lag boundary distance in the current level boundary value combination, so as to execute boundary crossing state recognition. If the radial distance is less than the inner lag boundary distance of the current level, that is, the intruder has continuously approached the target and crossed the inner buffer zone of the level boundary, it indicates that the behavior has broken through the inward safety threshold, and the boundary crossing determination result is recorded as "inner boundary crossing"; if the radial distance is greater than the outer lag boundary distance of the current level, it indicates that the intruder has withdrawn from outside the level boundary to outside the buffer zone of the upper level, and this behavior is marked as "outer boundary crossing"; and when the radial distance of the intruder is between the inner lag boundary and the outer lag boundary, it indicates that it is still within the effective level range and no crossing behavior has occurred, so the determination result is "boundary stay". The above three determination states respectively reflect the static distribution of the current intrusion behavior on the space boundary. The boundary crossing determination result and the motion direction determination result are logically combined to form a joint judgment condition. Three combination rules are used for processing: when the boundary crossing determination result is "inner boundary crossing" and the motion direction determination result is "inward", it indicates that the intruder is continuously approaching the target center and has broken through the boundary determination line, and the level crossing confirmation state is set to "confirmation crossing upgrade"; when the boundary crossing determination result is "outer boundary crossing" and the motion direction is "outward", it indicates that the intruder is moving away from the core area and has exited the current level response range, so the behavior is confirmed as "confirmation crossing downgrade"; and when the boundary crossing state is "boundary stay" or the determination state and the motion direction do not match, such as "outer boundary crossing" but the direction is "inward", etc., it indicates that the target has not formed a clear crossing trend or is in a boundary fluctuation stage, and it is determined as "maintaining the current level" to avoid triggering the lighting intensity adjustment and redundant linkage control.

[0100] In a specific embodiment, the process of performing step S4 can specifically include the following steps:

[0101] According to the fusion position data, a reinforcement learning state vector of the intruder is constructed, and the illumination brightness adjustment, color temperature change and stroboscopic setting in the hierarchical transition control instruction are defined as a reinforcement learning action space;

[0102] Based on the reinforcement learning state vector and the reinforcement learning action space, a reinforcement learning environment is constructed;

[0103] In the reinforcement learning environment, a reward function value is calculated based on the threat prevention success rate and the corresponding energy consumption data after the hierarchical transition control instruction is executed, and the reward function value is taken as a feedback signal of the reinforcement learning algorithm to obtain an action execution reward evaluation index;

[0104] According to the action execution reward evaluation index, corresponding brightness percentage, color temperature Kelvin value and stroboscopic frequency parameters are respectively configured for the intruder at each level to obtain a target lighting control strategy.

[0105] Specifically, based on the fusion position data, the intruder's reinforcement learning state vector is constructed, which contains the intruder's current horizontal coordinate, vertical coordinate and height coordinate in the three-dimensional coordinate system, and integrates a series of comprehensive parameters reflecting spatial position and dynamic behavior such as radial distance between the intruder and the protected target, the identification of the hierarchical area it belongs to, the current movement speed, the speed direction vector, the acceleration trend and the trajectory prediction residual, forming a state representation with high-dimensional description capability. At the same time, three types of key lighting parameters defined in the hierarchical transition control instruction are extracted, including lighting brightness adjustment, color temperature change and stroboscopic setting, and after discretization or continuous, they are uniformly included in the action space. The brightness adjustment is set to a linear variation interval of 0% to 100%, the color temperature value is defined between 2700K and 6500K, and the stroboscopic setting includes three typical response modes of constant light, low frequency flicker and high frequency flicker. The reinforcement learning action space represents the operation set that the system can choose in each decision cycle. Based on the reinforcement learning state vector and the reinforcement learning action space, the reinforcement learning environment is constructed. The reinforcement learning environment simulates the real lighting linkage response scene, where the state change is driven by the intruder behavior, the action is triggered by the lighting parameter control strategy, and the environment feedback evaluates the actual effect after the lighting response. In actual operation, every time the system executes a lighting control instruction, such as increasing brightness, switching color temperature or starting stroboscopic, it monitors in real time whether the intruder has stopped approaching, changed trajectory or evacuated the scene due to the behavior, and judges whether the current control action is effective in combination with the threat blocking judgment result obtained by the monitoring platform. At the same time, the energy consumption data consumed by this control operation is recorded in real time, including electric energy consumption, power change and equipment response delay, etc. The threat blocking success rate and the corresponding energy consumption data are input into the preset reward function model for calculation. The reward function model is designed as a difference function with effect priority, that is, the threat blocking success rate is the positive target and the energy consumption is the negative cost, which quantifies the comprehensive value of this action. The reward function value is used as the feedback signal of the reinforcement learning algorithm to update the policy evaluation network or the value function in the learning agent, forming the reward evaluation index after action execution. According to the action execution reward evaluation index, the intruder learns the optimal response strategy in different hierarchical environments, that is, according to the hierarchical identification field of the state vector, the optimal lighting parameter configuration result corresponding to the current hierarchical level is called from the learning strategy set to form the target lighting control strategy, and it is clearly specified that the lighting brightness should be set to a specific percentage, the color temperature should be adjusted to a certain Kelvin value, whether the stroboscopic response is turned on and its frequency setting, etc.

[0106] In this embodiment, the reinforcement learning state vector of the intruder is constructed according to the fused position data, and the lighting brightness adjustment, color temperature change and stroboscopic setting in the hierarchical transition control instruction are defined as the reinforcement learning action space, including: extracting the current position coordinates, moving speed and belonging level identifier of the intruder from the fused position data, combining the horizontal coordinate, vertical coordinate and height coordinate of the position coordinates, the three components of the speed vector and the level identifier to form a seven-dimensional numerical vector as the basic state vector of reinforcement learning; the lighting brightness adjustment parameter in the hierarchical transition control instruction is discretized into 101 brightness levels of 0% to 100%, the color temperature change parameter is discretized into 39 color temperature positions in the range of 2700K to 6500K, and the stroboscopic setting parameter is defined as three modes of constant, slow flash and fast flash, and the combination of the three types of parameters forms a total of 12177 possible lighting control combinations; the basic state vector is extended based on the fuzziness feature of the intruder behavior, when the intruder wanders near the level boundary, the threat level has fuzziness, the threat fuzziness coefficient is obtained by calculating the distance ratio of the intruder and the level boundary, and the coefficient is added to the state vector as the eighth-dimensional feature; an adaptive switching mechanism of the lighting strategy is implemented according to the threat fuzziness coefficient, when the fuzziness coefficient is lower than 0.3, a deterministic lighting strategy is adopted, when the fuzziness coefficient is higher than 0.7, a conservative lighting strategy is adopted, and when the fuzziness coefficient is between 0.3 and 0.7, a progressive lighting strategy is adopted, and the strategy type identifier is added to the state vector as the ninth-dimensional feature; a transfer learning mechanism of the lighting strategy is established, the successful lighting control experience in the historical similar threat scene is transferred to the current scene, the highest similarity of the historical strategy is selected as the initial action selection basis by calculating the similarity of the current nine-dimensional state vector and the historical state vector, and a complete reinforcement learning framework definition including nine-dimensional state features and 12177-dimensional action space is obtained.

[0107] In a specific embodiment, the execution step in the reinforcement learning environment, based on the threat prevention success rate after executing the hierarchical transition control instruction and the corresponding energy consumption data, the reward function value is calculated, and the reward function value is taken as the feedback signal of the reinforcement learning algorithm, and the process of obtaining the action execution reward evaluation index can specifically include the following steps:

[0108] In the reinforcement learning environment, the threat prevention success rate after executing the hierarchical transition control instruction is counted, and the current and voltage values of the lighting device during the execution of the hierarchical transition control instruction are monitored to obtain the corresponding energy consumption data;

[0109] The difference between the threat prevention success rate and the product of the energy consumption weight coefficient and the energy consumption data is taken as the reward function value;

[0110] The feedback signal of the reinforcement learning algorithm is generated according to the positive and negative nature and the value size of the reward function value, and the action execution reward evaluation index is calculated according to the feedback signal.

[0111] Specifically, after executing each level conversion control instruction, the actual threat prevention effect brought by the control action is counted in real time, and the energy consumption data generated by the lighting device in the response process is collected synchronously. The threat prevention success rate is taken as a key indicator for measuring the effectiveness of control, which is judged by monitoring the evolution result of the motion trajectory of the intrusion target in the monitoring platform. If the target stops approaching, deviates from the protected target or completely evacuates the layered area after the execution of the control action, it is considered as a successful event of threat prevention behavior. The average success rate of the control instruction in the current state is calculated by the ratio of the number of successful events to the total number of triggered events. At the same time, the physical power characteristics during the execution of the lighting adjustment action are monitored by using the current sensor and the voltage sensor, and the continuous current value and the real-time voltage data in the control instruction execution process are recorded. The total energy consumption data is obtained by integrating the two. The time stamp when the instruction starts to execute and the confirmation time after the lighting parameters are stable are taken as the start and end points of the integral interval. The threat prevention success rate and the total energy consumption data are brought into the set reward function, and the function model with the control effect as the positive target and the energy consumption cost as the penalty factor is used to calculate the reward value. The specific form is: reward function value = threat prevention success rate - (energy consumption weight coefficient x energy consumption data), wherein the energy consumption weight coefficient is a balance parameter set according to the actual running scene, which is used to adjust the priority weight between lighting efficiency and energy saving target. The reward function value calculated represents the comprehensive value evaluation of the current control strategy in this state. According to the positive and negative nature and the value size of the reward function value, the feedback signal required by the reinforcement learning algorithm is generated. If the reward value is positive and the value is large, it means that the current control action has achieved a better balance between effect and energy consumption, and the learning system adjusts the strategy evaluation function in a positive reinforcement manner to improve the probability of adopting the current action in similar states. If the reward value is negative or close to zero, it means that the current action brings high energy consumption but limited effect or insufficient intervention, so the Q value or strategy weight is updated in the form of negative reinforcement or punishment to reduce the possibility of selecting the action in the future. The reinforcement learning algorithm calculates the reward evaluation index of action execution according to the feedback signal combined with the historical state, action and reward data sequence, and continuously iterates and optimizes the strategy function to realize the adaptive search and continuous update of the optimal lighting response scheme in the complex dynamic environment.

[0112] The above describes the alarm linkage response method of the lighting system based on regional intrusion detection in the embodiment of the application. The alarm linkage response system of the lighting system based on regional intrusion detection in the embodiment of the application is described below. Please refer to Figure 2 The alarm linkage response system of the lighting system based on regional intrusion detection in the embodiment of the application includes one embodiment:

[0113] The positioning module 211 is configured to position the intruder according to the detection signals of the plurality of sensors in the monitoring area, and obtain fusion position data.

[0114] The creating module 212 is configured to create predefined hierarchical boundary data according to the monitoring area, and determine hierarchical area parameters according to the fusion position data and the predefined hierarchical boundary data.

[0115] The predicting module 213 is configured to perform trajectory prediction on the current motion state of the intruder according to the fusion position data, obtain trajectory prediction data, and perform bidirectional lag determination across the hierarchical boundary in combination with the hierarchical area parameters, to obtain hierarchical conversion control instructions.

[0116] The generating module 214 is configured to generate a target lighting control strategy based on the fusion position data and the hierarchical conversion control instructions.

[0117] Through the cooperation of the above components, through the fusion of the detection signals of the infrared sensor, the microwave sensor, the image recognition sensor and the laser ranging sensor, a technical leap from single triggering to stereoscopic perception is realized, and the accuracy and reliability of the intruder position positioning are significantly improved, and the false alarm and missed alarm problems caused by the traditional system relying on a single sensor are completely solved. A five-level depth area division mechanism is innovatively established, and an S-shaped weight progressive function is used to realize the smooth and continuous change of the lighting intensity with the intrusion depth, avoiding the traditional "all bright or all dark" abrupt control mode, which reduces the energy consumption and reduces the visual impact on the user. Through the second-order kinematic prediction algorithm and various motion mode recognition technologies, a technical breakthrough from passive response to active prediction is realized, and the system can preset the lighting parameters 2-3 seconds in advance, completely solving the technical defects of the response lag of the traditional system, ensuring the timeliness and continuity of the lighting response. A buffer zone mechanism and a bidirectional lag determination algorithm are designed to effectively avoid the frequent switching problem when the intruder wanders near the hierarchical boundary, significantly improving the stability of the system operation and the service life of the lighting equipment, and improving the user experience. A lighting strategy optimization engine based on reinforcement learning is constructed, and the system can automatically identify threat patterns and optimize response strategies based on historical data, realizing a technical leap from fixed mode to adaptive learning, so that the lighting strategy can continuously evolve and optimize to adapt to different application environments and threat characteristics. Through the behavior pattern analysis algorithm, the system can accurately identify four threat types: false intruders, scouts, direct threats and device destroyers, and configure differentiated lighting response strategies for each type, realizing a technical improvement from unified response to precise classification, significantly improving the relevance of threat identification and handling. By establishing an energy consumption-effect evaluation matrix and a reward function mechanism, the system can find the optimal balance point between threat prevention effect and energy consumption, realizing intelligent energy-saving control of the lighting system, ensuring security effect and reducing operating costs.

[0118] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here.

[0119] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the whole or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of 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 method described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0120] The above-described and above-mentioned embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for alarm linkage response of a lighting system based on regional intrusion detection, characterized in that, The method comprises the following steps: locating an intruder according to detection signals of multiple sensors in a monitoring area to obtain fusion position data; creating predefined hierarchical boundary data according to the monitoring area and determining hierarchical area parameters according to the fusion position data and the predefined hierarchical boundary data; specifically comprising: taking a protection target as the center to pre-establish a peripheral warning layer, a medium-distance early warning layer, a close-distance alarm layer, a dangerous approach layer and a core protection layer and determine the predefined hierarchical boundary data; extracting the horizontal coordinate and the vertical coordinate of the intruder from the fusion position data and calculating the radial distance of the intruder to the protection target based on the horizontal coordinate and the vertical coordinate; comparing the radial distance with the predefined hierarchical boundary data to determine the current hierarchical identifier of the intruder, and searching for the corresponding hierarchical area parameters according to the current hierarchical identifier; performing trajectory prediction on the current motion state of the intruder according to the fusion position data to obtain trajectory prediction data, and performing bidirectional lag judgment across the hierarchical boundary in combination with the hierarchical area parameters to obtain hierarchical conversion control instructions; specifically comprising: comparing the radial distance of the spatial position at the next moment with the current position according to the trajectory prediction data to obtain a motion direction determination result; extracting the hierarchical boundary distance of the peripheral warning layer, the medium-distance early warning layer, the close-distance alarm layer and the dangerous approach layer from the hierarchical area parameters; performing subtraction operation on the hierarchical boundary distance and a preset target value to obtain an inner lag boundary distance, and performing addition operation on the hierarchical boundary distance and the preset target value to obtain an outer lag boundary distance; performing boundary check on the inner lag boundary distance and the outer lag boundary distance to obtain bidirectional lag boundary parameters; comparing the radial distance of the intruder with the bidirectional lag boundary parameters to obtain a hierarchical crossing confirmation state; generating a target lighting control strategy based on the fusion position data and the hierarchical conversion control instructions.

2. The zone intrusion detection based lighting system alarm linkage response method according to claim 1, wherein, The method comprises the following steps: acquiring a heat source detection signal of an infrared sensor, a motion recognition signal of a microwave sensor, a target classification signal of an image recognition sensor and a distance measurement signal of a laser ranging sensor in the monitoring area; adding a time stamp mark to the heat source detection signal, the motion recognition signal, the target classification signal and the distance measurement signal respectively and performing time stamp alignment to obtain time-synchronized detection signals; determining the azimuth angle and the pitch angle of the intruder relative to the protection target based on the time-synchronized detection signals, and performing trigonometric function operation based on the azimuth angle and the pitch angle to obtain a three-dimensional spatial position comprising a horizontal coordinate, a vertical coordinate and a height coordinate; performing difference calculation on the three-dimensional spatial positions of consecutive frames to obtain inter-frame displacement, and dividing the inter-frame displacement by a first time interval to obtain fusion position data.

3. The zone intrusion detection based lighting system alarm linkage response method according to claim 1, wherein, The method comprises the following steps: Determine the coordinate difference between adjacent frames according to the fusion position data, and calculate the velocity sequence data by dividing the coordinate difference between adjacent frames by the second time interval; Calculate the acceleration vector based on the difference between adjacent velocity vectors in the velocity sequence data, and determine the trajectory curvature according to the acceleration vector; According to the trajectory curvature, the motion behavior of the intruder is identified as a straight line approach, a curve around, a jump movement or a wandering observation mode, and a trajectory curve fitting is performed to obtain a trajectory fitting result; According to the trajectory fitting result, the current position coordinates, the current approach speed and the acceleration trend are substituted into the second-order kinematics prediction equation to calculate the spatial position of the intruder at the next time and the estimated time of reaching each level boundary, and the trajectory prediction data is obtained. The trajectory prediction data and the hierarchical region parameters are bidirectional hysteresis judged across the level boundary to obtain a level conversion control instruction.

4. The zone-based intrusion detection based lighting system alarm linkage response method of claim 1, wherein, The numerical comparison of the radial distance of the intruder with the bidirectional hysteresis boundary parameter obtains a level crossing confirmation state, which includes: Find the corresponding current level boundary numerical combination in the bidirectional hysteresis boundary parameter; Compare the radial distance of the intruder with the inner hysteresis boundary distance and the outer hysteresis boundary distance in the current level boundary numerical combination respectively, record the boundary crossing determination result as inner boundary crossing when the radial distance is less than the inner hysteresis boundary distance, record the boundary crossing determination result as outer boundary crossing when the radial distance is greater than the outer hysteresis boundary distance, and record the boundary crossing determination result as boundary inside stay when the radial distance is between the two boundaries. The boundary crossing determination result and the motion direction determination result are combined and logically judged to determine the level crossing confirmation state, which includes confirming crossing upgrade, confirming crossing downgrade and keeping the current level.

5. The zone-based intrusion detection based lighting system alarm linkage response method of claim 1, wherein, The target lighting control strategy is generated based on the fusion position data and the level conversion control instruction, which includes: Construct the reinforcement learning state vector of the intruder according to the fusion position data, and define the lighting brightness adjustment, color temperature change and frequency flicker setting in the level conversion control instruction as the reinforcement learning action space; Construct the reinforcement learning environment based on the reinforcement learning state vector and the reinforcement learning action space; In the reinforcement learning environment, calculate the reward function value based on the threat prevention success rate and the corresponding energy consumption data after executing the level conversion control instruction, and use the reward function value as the feedback signal of the reinforcement learning algorithm to obtain the action execution reward evaluation index; According to the action execution reward evaluation index, configure corresponding brightness percentage, color temperature Kelvin value and frequency flicker frequency parameters for the intruder in each level to obtain the target lighting control strategy.

6. The zone-based intrusion detection based lighting system alarm linkage response method according to claim 5, wherein, In the reinforcement learning environment, calculate the reward function value based on the threat prevention success rate and the corresponding energy consumption data after executing the level conversion control instruction, and use the reward function value as the feedback signal of the reinforcement learning algorithm to obtain the action execution reward evaluation index, which includes: In the reinforcement learning environment, a threat blocking success rate after the hierarchical transition control instruction is executed is statistically performed, and current consumption and voltage values of the lighting device during execution of the hierarchical transition control instruction are monitored to obtain corresponding energy consumption data; A difference value of the threat blocking success rate minus a product of an energy consumption weight coefficient and the energy consumption data is taken as a reward function value; A feedback signal of a reinforcement learning algorithm is generated according to a positive or negative nature and a value size of the reward function value, and an action execution reward evaluation index is calculated according to the feedback signal.

7. A zone-based intrusion detection based lighting system alarm linkage response system, characterized in that, The method is used for executing the alarm linkage response of the regional intrusion detection-based lighting system as claimed in any one of claims 1-6, and the regional intrusion detection-based lighting system alarm linkage response system comprises: A positioning module is configured to position an intruder according to detection signals of a plurality of sensors in a monitoring area to obtain fusion position data; A creating module is configured to create predefined hierarchical boundary data according to the monitoring area, and determine layered area parameters according to the fusion position data and the predefined hierarchical boundary data; A prediction module is configured to perform trajectory prediction on a current motion state of the intruder according to the fusion position data to obtain trajectory prediction data, and perform bidirectional hysteresis determination across the hierarchical boundary in combination with the layered area parameters to obtain a hierarchical transition control instruction; A generation module is configured to generate a target lighting control strategy based on the fusion position data and the hierarchical transition control instruction.

Images