Security and protection monitoring intelligent analysis method and system based on Internet of Things

Abstract

The invention relates to the technical field of the Internet of Things, and discloses a security monitoring intelligent analysis method and system based on the Internet of Things. The method comprises the following steps: performing view field distortion correction on a multi-camera video stream of the Internet of Things security and protection monitoring system to obtain corrected pixel coordinates; calculating target space coordinates from each camera to a target based on the corrected pixel coordinates; performing spectrum sensing and dynamic allocation in combination with the target space coordinates to obtain a transmission spectrum allocation scheme; and performing iterative optimization on the target space coordinates based on the transmission spectrum allocation scheme, and outputting a target positioning result of the Internet of Things security and protection monitoring system. According to the method, systematic errors in a traditional distance measurement method are effectively suppressed, the anti-interference capability in a complex electromagnetic environment is enhanced, on-demand dynamic allocation of spectrum resources is realized, and the real-time requirement of a security and protection monitoring system of the Internet of Things is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things, and particularly relates to a security monitoring intelligent analysis method and system based on Internet of Things. BACKGROUND

[0002] Traditional security monitoring systems mainly rely on single cameras for target detection and tracking, but face problems such as limited viewing angle and insufficient positioning accuracy in complex environments. Although existing multi-camera cooperative monitoring technology can expand the monitoring range to some extent, it often fails to meet the requirements of modern security monitoring for high accuracy, real-time performance and reliability due to the lack of effective multi-view information fusion mechanisms and accurate target positioning algorithms. At the same time, most existing multi-camera positioning algorithms are based on simple triangulation or stereo vision principles, lacking effective processing mechanisms for multipath effects in complex electromagnetic environments, which significantly reduces the positioning reliability in complex scenes such as occlusion and reflection. SUMMARY

[0003] The main purpose of the present application is to provide a security monitoring intelligent analysis method and system based on Internet of Things, which effectively suppresses the systematic errors in traditional ranging methods, enhances the anti-interference ability in complex electromagnetic environments, realizes the on-demand dynamic allocation of spectrum resources, and meets the real-time requirements of Internet of Things security monitoring systems.

[0004] To achieve the above purpose, the present application provides a security monitoring intelligent analysis method based on Internet of Things, comprising the following steps: Field of view distortion correction is performed on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates; Based on the corrected pixel coordinates, the target space coordinates of each camera to the target are calculated; Frequency spectrum sensing and dynamic allocation are performed in combination with the target space coordinates to obtain a transmission spectrum allocation scheme; Based on the transmission spectrum allocation scheme, the target space coordinates are iteratively optimized to output the target positioning result of the Internet of Things security monitoring system.

[0005] Optionally, in the first implementation manner of the first aspect of the present application, the field of view distortion correction on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates comprises: Pixel coordinates and brightness information of the target in the multi-camera video stream are extracted to obtain basic observation data of each camera; A video signal intensity measurement model containing light intensity attenuation coefficient and incident angle compensation is established based on the basic observation data, and the signal intensity attenuation value of the target under different viewing angles is calculated through the video signal intensity measurement model; extracting original feature data containing pixel coordinates, brightness values, signal-to-noise ratios and time stamps according to the signal strength attenuation values; performing field of view distortion correction on the original feature data to obtain corrected pixel coordinates.

[0006] Optionally, in a second implementation form of the first aspect of the present application, the field of view distortion correction on the original feature data to obtain corrected pixel coordinates comprises: extracting pixel coordinates of each camera from the original feature data and converting the pixel coordinates into first coordinates; establishing a continuous field of view distortion correction function containing radial distortion items and tangential distortion items based on the first coordinates and calculating distortion coefficients of each camera through the continuous field of view distortion correction function; performing distortion correction on the first coordinates according to the distortion coefficients to obtain second coordinates; converting the second coordinates back to the pixel coordinate system to obtain corrected pixel coordinates.

[0007] Optionally, in a third implementation form of the first aspect of the present application, the calculation of target space coordinates of each camera to the target based on the corrected pixel coordinates comprises: detecting a round-trip time delay of the video signal according to the corrected pixel coordinates and calculating an initial measured distance of each camera to the target based on the round-trip time delay; performing error correction on the initial measured distance to obtain an environment corrected distance; establishing a distance residual correction algorithm based on the environment corrected distance and calculating adaptive weight coefficients according to signal-to-noise ratios and time delay attenuation factors of each camera; performing multipath effect detection and path separation correction on the environment corrected distance and combining the adaptive weight coefficients to calculate weighted distance measurement results; performing position prediction based on the weighted distance measurement results to obtain target space coordinates.

[0008] Optionally, in a fourth implementation form of the first aspect of the present application, the multipath effect detection and path separation correction on the environment corrected distance and the combination of the adaptive weight coefficients to calculate weighted distance measurement results comprise: performing amplitude second derivative and phase jump analysis on the video signal corresponding to the environment corrected distance to obtain signal amplitude second derivative and phase jump; when the signal amplitude second derivative exceeds a preset signal amplitude threshold and the phase jump exceeds a preset phase threshold, detecting a multipath effect and outputting a multipath effect detection result; Based on the multipath effect detection result, a RAKE receiver principle is used for path separation on the video signal to obtain a separated signal of each propagation path and a corresponding path distance measurement value; A path weight coefficient of each propagation path is calculated according to the signal strength of each propagation path, and a comprehensive weight matrix is formed in combination with the adaptive weight coefficient; The path distance measurement value is weighted and fused with the comprehensive weight matrix to obtain a weighted distance measurement result.

[0009] Optionally, in the fifth implementation manner of the first aspect of the present application, the position prediction based on the weighted distance measurement result to obtain the target space coordinates comprises: A positioning residual value is calculated according to the weighted distance measurement result, and when the positioning residual value exceeds a residual threshold value, the weighted distance measurement result is input into a video interpolation neural network for position prediction to obtain predicted position data; The predicted position data is subjected to a time sequence consistency constraint to obtain a time sequence constrained position; A position confidence score is calculated based on the time sequence constrained position, and the time sequence constrained position and the confidence score are combined to output the target space coordinates.

[0010] Optionally, in the sixth implementation manner of the first aspect of the present application, the spectrum sensing and dynamic allocation in combination with the target space coordinates to obtain a transmission spectrum allocation scheme comprises: A distance relationship between each camera and the target is calculated according to the target space coordinates, and spectrum occupation sensing is performed to identify available spectrum state information; A position-aware spectrum resource mapping relationship is constructed based on the available spectrum state information and the target space coordinates; Real-time quality evaluation is performed on data transmission under the spectrum resource mapping relationship to obtain a quality evaluation index, and when the quality evaluation index falls below a preset standard, an intelligent spectrum switching is started by a spectrum reallocation process to output a transmission spectrum allocation scheme.

[0011] Optionally, in the seventh implementation manner of the first aspect of the present application, the position-aware spectrum resource mapping relationship constructed based on the available spectrum state information and the target space coordinates comprises: The spatial distance value between each camera and the target is calculated according to the target space coordinates, and a signal-to-interference-and-noise ratio parameter of each frequency band is extracted from the available spectrum state information; A quality evaluation matrix is generated through the product operation of the spatial distance value and the signal-to-interference-and-noise ratio parameter; establish a mathematical optimization model based on the quality evaluation matrix, with the goal of maximizing the total transmission capacity of the system, and set the constraint condition as each camera monopolizing one frequency band and each frequency band being allocated to only one camera, and solve to obtain a spectrum allocation decision matrix; implement a distance priority adjustment strategy on the spectrum allocation decision matrix to form a spectrum allocation matrix; convert the binary allocation identifier in the spectrum allocation matrix into a direct correspondence table of camera number and frequency band number, and establish a spectrum resource mapping relationship.

[0012] Optionally, in an eighth implementation manner of the first aspect of the present application, the iterative optimization of the target space coordinates based on the transmission spectrum allocation scheme outputs a target positioning result of the Internet of Things security monitoring system, and the method comprises the following steps: constructing a multi-objective optimization function according to a transmission quality evaluation index in the transmission spectrum allocation scheme and in combination with the positioning accuracy of the target space coordinates and the system processing time delay; inputting the target space coordinates as initial position parameters into a Levenberg-Marquardt iterative algorithm and performing iterative optimization in combination with the multi-objective optimization function; performing exponential moving average updating on the credibility weight of each camera in the iterative optimization process to determine a target position coordinate; calculating a spatial positioning accuracy and a confidence value based on the target position coordinate, combining target motion trajectory analysis and abnormal behavior detection to generate a target positioning result of the Internet of Things security monitoring system.

[0013] The present application also provides an Internet of Things-based security monitoring intelligent analysis system, which comprises: a field of view distortion correction module configured to perform field of view distortion correction on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates; a calculation module configured to calculate target space coordinates of each camera to a target based on the corrected pixel coordinates; a spectrum sensing and dynamic allocation module configured to perform spectrum sensing and dynamic allocation in combination with the target space coordinates to obtain a transmission spectrum allocation scheme; an iterative optimization module configured to perform iterative optimization of the target space coordinates based on the transmission spectrum allocation scheme to output a target positioning result of the Internet of Things security monitoring system.

[0014] In summary, the application can accurately quantify the attenuation characteristics of the signal in the propagation process by establishing a video signal intensity measurement model containing light intensity attenuation coefficient and incident angle compensation, effectively reducing the influence of environmental factors on signal measurement compared with the traditional simple pixel extraction method. The continuous field of view distortion correction function containing radial distortion term and tangential distortion term, especially the introduction of tangent function term, can accurately process the strong non-linear distortion of the edge area of the wide-angle monitoring camera, solving the technical problem that the traditional polynomial distortion model cannot effectively correct complex distortion. Through the distance residual correction algorithm combined with the adaptive weight mechanism, the signal-to-noise ratio and time delay attenuation factors are considered comprehensively, and the systematic error in the traditional ranging method is effectively suppressed, and the anti-interference ability in the complex electromagnetic environment is enhanced. The multipath effect detection mechanism based on the amplitude second derivative and phase jump analysis, combined with the path separation technology of RAKE receiver principle, effectively solves the technical bottleneck that the traditional method cannot handle multipath interference in complex propagation environment. The neural network architecture with timing consistency constraint, through the optimization combination of multiple activation functions and the timing continuity constraint, ensures that the position prediction result conforms to the physical law of target motion, avoiding the problem of discontinuous inter-frame prediction in traditional methods. Through the intelligent spectrum allocation strategy based on position information, the target spatial distribution is combined with the spectrum resource configuration, realizing the on-demand dynamic allocation of spectrum resources, and through the multi-objective iterative optimization algorithm, the positioning accuracy, processing delay and transmission quality are optimized, overcoming the problem of overall performance inconsistency caused by independent optimization of each subsystem in traditional methods, realizing the performance optimization of the system level. The Levenberg-Marquardt algorithm with adaptive damping factor and the exponential moving average weight update strategy significantly improve the convergence speed while ensuring the optimization accuracy, meeting the strict real-time requirements of the Internet of Things security monitoring system. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is an embodiment of the application based on the Internet of Things Security Monitoring Intelligent Analysis Method Step Schematic Diagram; Figure 2 is an embodiment of the application based on the Internet of Things Security Monitoring Intelligent Analysis System Structure Block Diagram.

[0016] The implementation of the application, functional characteristics and advantages will be further described with reference to the embodiments and accompanying drawings. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0018] Reference Figure 1The embodiment provides an intelligent security monitoring analysis method based on an Internet of Things, and comprises the following steps: S1, field of view distortion correction is performed on a multi-camera video stream of an Internet of Things security monitoring system to obtain corrected pixel coordinates; Among them, for the synchronous video stream data collected by the multi-camera, the target in the monitoring area is detected based on the frame difference method and the optical flow method, and the pixel coordinates and the corresponding brightness information of the target in each frame of image are obtained by combining the high-resolution imaging module to construct a basic observation data set. Based on the improved video signal intensity measurement model, wherein the exponential decay term is used to describe the decreasing characteristics of the target reflection signal intensity with the increase of the three-dimensional space distance, and the cos(θ) is introduced as the light incidence angle compensation factor to eliminate the influence of the change of the viewing angle on the brightness measurement, and the random noise distribution is introduced into the model to improve the adaptability of the model to the noise interference in the actual monitoring environment, and then the signal attenuation degree of the target under each camera angle is calculated according to the signal intensity measurement model, and the real light intensity is back calculated. The output of the above model is fused with the basic observation data to construct an original feature vector sequence containing pixel coordinates, brightness values, signal-to-noise ratios, time stamps, brightness gradient information and local texture features. The continuous field of view distortion correction function is applied to the pixel coordinates extracted in the original feature vector, the continuous field of view distortion correction function describes the composite distortion effect of the wide-angle camera in the edge area through the nonlinear combination of the radial term and the tangential term, realizes the reverse mapping correction of the pixel position in the normalized coordinate system, and then restores the corrected coordinates to the image plane, completes the correction process under the multi-camera angle, and outputs the corrected pixel coordinate data.

[0019] S2, based on the corrected pixel coordinates, the target space coordinates of each camera to the target are calculated; Specifically, based on the corrected pixel coordinates, the frame-level timestamps of the target are extracted from the video data of each camera, and the time delay of the multi-camera observation target is uniformly calibrated through a frame synchronization mechanism. The round-trip delay of the target video signal is detected, and the round-trip delay is converted into an initial measurement distance using the light propagation formula. In the conversion process, the systematic error compensation term of environmental factors such as temperature and humidity is considered. The error compensation term is calculated from the temperature, relative humidity, and empirical coefficient according to the offline calibration model, so that the initial ranging data is corrected to the environmental corrected distance. In order to offset the influence of multi-source error accumulation on the positioning accuracy, a distance residual correction algorithm combined with multiple cameras is constructed. The distance residual correction algorithm defines a residual function between the theoretical distance and the measured distance, and calculates the weight coefficient by combining the signal-to-noise ratio and the time delay attenuation degree of the camera. Among them, the camera with high signal-to-noise ratio contributes more to the result, while the path with time attenuation effect decays according to the exponential function, realizing adaptive weighted control. The environmental corrected distance is detected for multipath effect, and the direct, reflected and scattered paths are identified and separated according to the amplitude second derivative and phase mutation index of the signal. The RAKE receiving idea is used to estimate the distance contribution and weight of each path, and the multipath correction is performed on the environmental corrected distance based on the weight fusion strategy, to obtain the weighted distance measurement result under the multi-camera. All weighted distances are taken as input to construct a nonlinear least squares optimization model, and the three-dimensional position is solved to minimize the residual function, so as to predict and output the spatial coordinates of the target.

[0020] S3, combined with the target spatial coordinates, spectrum sensing and dynamic allocation are performed to obtain a transmission spectrum allocation scheme; It should be noted that the distance mapping relationship is established according to the target space coordinates and the position coordinates of each camera, the real-time distance between each camera and the target is calculated through the space Euclidean distance formula, and a distance matrix is constructed based on this. Through the improved energy detection algorithm, the frequency spectrum occupation sensing analysis is carried out on each frequency band. The energy detection algorithm introduces a dynamic noise power estimation mechanism on the basis of traditional received power statistics, and adjusts the decision threshold in real time, so as to adaptively identify the available frequency spectrum state information in different interference environments, including the current idle channel number, signal-to-noise ratio level and adjacent frequency interference index. Combined with the target space position and the spectrum occupation state, the position-aware spectrum resource mapping relationship is established in the multi-frequency band multi-subchannel resource pool, that is, the subchannels with low interference and wide bandwidth are preferentially allocated to the cameras with short distance to the target and high channel quality, so as to maximize the data throughput efficiency and transmission stability. Real-time quality evaluation is performed on the data transmission under the spectrum resource mapping relationship, the throughput, average delay and packet loss rate of each link are continuously monitored, and the quality evaluation index Q is generated through normalized weighted fusion. When the Q value of a certain link is lower than the set threshold, it is considered that the current spectrum configuration cannot meet the transmission requirements of the target data flow, and the spectrum reallocation process is triggered. In the reallocation process, a full-band scan is performed, the available spectrum state is updated, the channel is optimized based on the current space position and channel SINR level, the channel switching is completed through the intelligent spectrum switching mechanism without interrupting the current video stream transmission, and the transmission spectrum allocation scheme is output.

[0021] S4, based on the transmission spectrum allocation scheme, iteratively optimizing the target space coordinates, outputting the target positioning result of the Internet of Things security monitoring system.

[0022] Specifically, the link quality evaluation index in the transmission spectrum allocation scheme is introduced into the positioning optimization framework as an evaluation parameter, and a multi-objective optimization function is constructed by combining the initial positioning accuracy of the target spatial coordinates and the overall data processing delay of the system, wherein the multi-objective optimization function takes the positioning residual sum of squares as the main target, takes the normalized processing delay ratio and the transmission quality inverse factor as the secondary target, and is integrated into a global optimization index through weighting. The current estimated target three-dimensional spatial coordinates are input as initial variables into the improved Levenberg-Marquardt iterative optimization framework, and the numerical iteration of the residual function is performed based on the Jacobian matrix generated by automatic differentiation. The switching mechanism between gradient descent and second-order Newton method is adjusted by the dynamic damping factor to maintain the convergence speed while enhancing the stability of the algorithm, and a more optimal position solution is obtained. In each iteration process, the new credibility score of each camera is calculated in real time according to the signal-to-noise ratio, residual contribution, and transmission channel state, and the original weight is updated in an exponential moving average manner, so that the system can dynamically adjust the influence of each perspective on the final position judgment when processing continuous target tracking, and improve the reliability of the multi-camera fusion result. When the optimization iteration meets the residual convergence, the position change is small, or the maximum round condition is reached, the estimated target position coordinates are output, and the position estimation accuracy is calculated based on the weighted error propagation of all cameras, and the confidence score is output according to the residual normalized value. The speed and acceleration vectors of the continuous frame position coordinates are extracted, the target motion trend is analyzed by combining historical trajectory modeling and behavior pattern library, and if the features such as deviation, abnormal speed mutation or non-compliance with regional behavior rules are detected in the motion path, the abnormal behavior flag is triggered, and the target positioning information including spatial position, error evaluation, confidence level and behavior judgment result is output.

[0023] In one example, a field of view distortion correction is performed on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates, including: Pixel coordinates and brightness information of the target in the multi-camera video stream are extracted to obtain basic observation data of each camera; A video signal intensity measurement model including light intensity attenuation coefficients and incident angle compensation is established based on the basic observation data, and the signal intensity attenuation values of the target under different perspectives are calculated through the video signal intensity measurement model; Raw feature data including pixel coordinates, brightness values, signal-to-noise ratios, and timestamps are extracted according to the signal intensity attenuation values; The raw feature data is subjected to field of view distortion correction to obtain corrected pixel coordinates.

[0024] In this example, in a multi-camera array, the collected video data is processed frame by frame, and the motion target contour in the current frame is identified by background modeling combined with inter-frame difference and optical flow detection algorithm, and the contour area is pixel counted to extract the two-dimensional pixel position coordinates and brightness distribution value of the target in the image frame. The brightness information is obtained by image gray value or from RGB channel weighted transformation, and at the same time, the time stamp information is combined to mark the time when the target appears in each frame of image, to form the basic observation data. Combined with the spatial installation parameters and optical axis orientation of the camera, a video signal intensity measurement model for modeling the actual light attenuation process is constructed in the reference calibration environment. The video signal intensity measurement model considers the distance loss effect experienced by the reflected light from the target surface to the camera, that is, the farther the target, the lower the brightness, and also considers the brightness loss caused by the angle deviating from the main optical axis of the camera, that is, the more oblique the angle, the weaker the light intensity received by the camera. On this basis, non-ideal noise disturbance caused by atmosphere, dust, optical lens material, etc. in the actual environment is superimposed to fit the real attenuation trend of the target brightness value in the actual use scene. By applying the video signal intensity measurement model to the image data collected by each camera, the signal attenuation intensity of the target in the current frame is inversely deduced under the premise of keeping the original pixel coordinates unchanged, and the signal-to-noise ratio value is calculated by combining the image brightness value of each pixel point and the local background noise evaluation parameter, to represent the clarity of the brightness feature of the target area in the current video frame. The signal-to-noise ratio value is fused with the original pixel coordinates, image brightness, signal-to-noise ratio, and corresponding frame timestamp data to construct an original feature vector. Since the multi-camera system uses a wide-angle lens to expand the field of view, the image edges are prone to obvious distortion. The original pixel coordinates are subjected to field of view distortion correction processing. The correction process converts each pixel coordinate in the image plane to a normalized image coordinate according to the intrinsic matrix of the lens and the distortion parameters extracted from the calibration image, and then performs nonlinear offset compensation on the normalized coordinate according to the pre-constructed continuous field of view distortion function. In the compensation process, the image stretching effect caused by the radial distortion of the lens is considered, and the image offset distortion caused by the tangential distortion is corrected, while avoiding the discontinuous fitting phenomenon of traditional segmented polynomial model in strong distortion area. When the normalization correction is completed, the coordinate system is projected back to the original image space to obtain the corrected pixel coordinates after the angle consistency correction, and the coordinates are reassembled into the original feature vector to output a unified observation feature data set containing the corrected pixel coordinates, light intensity, noise weight, and timestamp.

[0025] In one example, the original feature data is subjected to field of view distortion correction to obtain corrected pixel coordinates, comprising: extracting the pixel coordinates of each camera from the original feature data, and converting the pixel coordinates to a first coordinate; a continuous field distortion correction function including a radial distortion term and a tangential distortion term is established based on the first coordinates, and distortion coefficients of each camera are calculated through the continuous field distortion correction function; distortion correction is performed on the first coordinates according to the distortion coefficients to obtain second coordinates; The second coordinates are converted back to the pixel coordinate system to obtain corrected pixel coordinates.

[0026] In this example, pixel coordinate information detected under each camera view is extracted from the feature data of the original multi-camera video stream. The pixel coordinates are the position markers of the target in the two-dimensional image plane within the image frame, represented as an integer pixel coordinate pair on the horizontal and vertical axes. The original pixel coordinates are substituted into the camera internal parameter model for conversion. According to the principal point position and focal length value obtained through camera calibration, the pixel plane coordinates are mapped to normalized imaging plane coordinates, i.e., first coordinates. The first coordinates are in the camera imaging coordinate system and are dimensionless real numbers, representing the theoretical position of the target in the lens imaging geometric model. The conversion process eliminates the interference of hardware parameters such as pixel size, resolution, and image size on geometric modeling, allowing each camera to perform distortion modeling based on a unified geometric reference system under different imaging settings. Based on the first coordinates, a continuous field distortion correction function including a radial distortion term and a tangential distortion term is established according to the actual imaging distortion problem of wide-angle lenses or fisheye lenses. The distortion correction function uniformly models the entire field through a continuous and derivable function form. The radial distortion term is used to simulate the image radial stretching effect caused by the increasing distance between the target and the imaging center, which is reflected as the nonlinear shift of the image edge outward or inward. The tangential distortion term is used to compensate for the image distortion caused by lens installation tilt or manufacturing errors, which is manifested as the misalignment drift of image pixels in asymmetric directions. When constructing the continuous field distortion correction function, a set of distortion coefficients in the function is fitted using a nonlinear least error optimization method based on the pairs of calibration points with known physical positions in the calibration image and their observed pixel positions in the image, and the smooth continuity and physical interpretability of the model in the edge region of the image are ensured to improve the effectiveness of distortion compensation at extreme viewing angles. The first coordinates are substituted into the continuous distortion correction function for coordinate correction. For each coordinate point, the offset vector of the coordinate point is calculated according to the geometric offset model described by the continuous distortion correction function, and the offset result is applied to the original first coordinate position to obtain the second coordinates after distortion correction. The second coordinates are converted from the normalized coordinate system back to the pixel coordinate system of the original image. Through the restoration of the focal length scaling and the principal point coordinate offset, the second coordinates are mapped back to the standard pixel point position in the image frame to obtain the corrected pixel coordinates.

[0027] In one example, target space coordinates of each camera to the target are calculated based on the corrected pixel coordinates, including: The round-trip delay of the video signal is detected according to the corrected pixel coordinates, and an initial measured distance of each camera to the target is calculated based on the round-trip delay; An error correction is performed on the initial measured distance to obtain an environment corrected distance; A distance residual correction algorithm is established based on the environment corrected distance, and an adaptive weight coefficient is calculated according to a signal-to-noise ratio and a time delay attenuation factor of each camera; A multi-path effect detection and path separation correction are performed on the environment corrected distance, and a weighted distance measurement result is calculated in combination with the adaptive weight coefficient; A position prediction is performed based on the weighted distance measurement result to obtain a target space coordinate.

[0028] In this example, according to the corrected pixel coordinates, the time synchronization module integrated in the joint camera system is used to obtain the timestamp data when each camera detects the target, and combined with the set start and end events of video signal transmission and return, the round-trip propagation delay of the target image in the video signal processing flow is determined. The round-trip delay represents the complete path delay of the video signal from the camera to the target and then back to the camera. The measurement is based on the network time protocol with microsecond-level precision for time synchronization processing between multiple cameras to ensure the time domain consistency of the data source. After obtaining the round-trip delay, the initial measurement distance between the camera and the target is preliminarily calculated based on the actual speed standard of the propagation medium, which is light or electromagnetic signal. At the same time, the initial measurement distance value is corrected to eliminate the influence of environmental errors, including systematic errors caused by temperature fluctuations, relative humidity changes, and sensor working state in the signal propagation path. Combined with the current collected environmental temperature and humidity data and the preset error compensation model, the initial distance value is corrected in the first order to obtain a more accurate environmental correction distance in the actual scene. Based on the environmental correction distances obtained by multiple cameras, a distance residual correction algorithm suitable for non-ideal observation conditions is constructed. The distance residual correction algorithm takes a certain estimated position in the three-dimensional space as the center, calculates the residual value between the theoretical distance from the center position to each camera and the measured environmental correction distance, and finds the optimal position estimation point of the target by minimizing the sum of squares of residuals. According to the signal-to-noise ratio level in the current image frame of each camera and the time delay attenuation experienced during signal sampling, the adaptive weight coefficient of each camera is dynamically calculated. The weight reflects the comprehensive index of image clarity, data freshness and channel stability. The higher the value, the stronger the credibility of the camera to the final position solution. While constructing the residual function, the multipath effect that may exist in the environmental correction distance is identified and stripped. In complex scenes, there are not only one return path for the signal, but also direct, reflected and diffracted propagation modes. At this time, the time series differential analysis of the signal intensity curve and the monitoring of the phase jump are carried out to judge whether there are multiple propagation paths. If it meets the set jump threshold and amplitude change rule, it is considered that there is a multipath superposition phenomenon, and the signal path separation processing is carried out accordingly. By weighting and fusing the distance measurement values of each path, the interference of non-main path on the distance measurement result is effectively eliminated. Combined with the adaptive weight coefficient, the distance measurement result after multipath separation is weighted and integrated to obtain the weighted distance measurement result. The weighted distance result corresponding to each camera is input into the three-dimensional coordinate inversion module, and the nonlinear minimization method is used to iteratively solve the optimal coordinate position of the target that satisfies all weighted distance constraints in space. At the same time, the calculation process is dynamically adjusted by combining the position residual change trend and the signal consistency evaluation mechanism.

[0029] In one example, the environmental correction distance is detected for multipath effect and path separation correction, and the weighted distance measurement result is calculated by combining the adaptive weight coefficient, including: The video signal corresponding to the environmental correction distance is subjected to amplitude second derivative and phase jump analysis to obtain signal amplitude second derivative and phase jump; When the signal amplitude second derivative exceeds a preset signal amplitude threshold and the phase jump exceeds a preset phase threshold, a multipath effect is detected and a multipath effect detection result is output; Based on the multipath effect detection result, a RAKE receiver principle is used to separate paths for the video signal to obtain separated signals of each propagation path and corresponding path distance measurement values; Path weight coefficients of each propagation path are calculated according to signal strengths of each propagation path, and a comprehensive weight matrix is formed in combination with adaptive weight coefficients; The path distance measurement values are weighted and fused with the comprehensive weight matrix to obtain a weighted distance measurement result.

[0030] In this example, based on the environmental correction distance and the corresponding video signal, the time-domain amplitude curve and phase trajectory of each signal are analyzed, and for the continuous samples of the signal intensity changing over time, the second derivative operation of the amplitude value is performed to obtain the acceleration characteristics of the local change, and the phase values of adjacent time points in the phase domain are differentially processed to determine whether the phase continuity of the signal in the propagation process encounters a mutation. If the second derivative of the amplitude exceeds the preset signal amplitude threshold value in a certain period of time, and the phase jump value also exceeds the set phase mutation threshold value at the same time, it indicates that the signal is disturbed by multiple propagation paths, and the typical multipath superposition phenomenon occurs, marking that the frame or the path exists multipath effect. Based on the multipath effect detection result, the mixed signal is processed by the RAKE receiver mechanism, the entire received composite signal is regarded as being superimposed by a plurality of sub-paths with different time delays and different attenuation coefficients, and a plurality of receiving "fingers" are set to capture these signal components with correlation but small delay difference on the time axis, so as to separate the composite signal into a plurality of independent propagation paths in the signal space, each path corresponding to an obvious delay peak and a corresponding energy peak. Combined with the time stamp and the correction distance information in the early stage, the independent path propagation time and the path distance measurement value converted therefrom are calculated for each separated path. The signal intensity of each path is counted, and the signal intensity is used as an important basis for path reliability. The path weight coefficient is generated through the normalization processing of the signal amplitude of each path, reflecting the contribution degree of each path in the overall ranging result. The higher the signal intensity, the more likely the path is the main path or the main reflection path, the higher the distance reliability, and the greater the weight. According to the adaptive weight coefficient of the camera, a comprehensive weight matrix under the condition of multiple cameras and multiple paths is constructed together with the path weight, which horizontally describes the contribution intensity of different propagation paths and vertically describes the overall reliability of the result by different cameras, forming a weight distribution dynamic response system. The path distance measurement values and the corresponding comprehensive weights are multiplied in a matrix relationship, and all weighted distance values are summed and normalized to obtain the weighted comprehensive distance measurement result.

[0031] In one example, based on the weighted distance measurement result, a position prediction is performed to obtain a target space coordinate, including: According to the weighted distance measurement result, a positioning residual value is calculated, and when the positioning residual value exceeds a residual threshold value, the weighted distance measurement result is input into a video interpolation neural network for position prediction to obtain predicted position data; The predicted position data is subjected to a time sequence consistency constraint to obtain a time sequence constrained position; A position confidence score is calculated based on the time sequence constrained position, and the time sequence constrained position and the confidence score are combined to output the target space coordinate.

[0032] In this example, the initial position estimate of the target in space is constructed according to the weighted distance measurement result, and the initial position estimate value is compared with the actual observation constraint obtained by multi-camera multi-path fusion, and a three-dimensional residual value is calculated, that is, the deviation between the position supported by the ranging observation and the geometric inversion estimation result, and the positioning residual value is numerically considered as a stability indicator of the current positioning result. When the positioning residual value is lower than the preset residual threshold, it means that the current system ranging result is more reliable, which is directly used for downstream modules. If the positioning residual value exceeds the set threshold, it means that there are non-ideal factors such as ranging abnormality, target shielding, noise interference or complex environmental reflection in the current scene. The current weighted ranging result is input into the video interpolation neural network structure for time series-based video interpolation prediction. The video interpolation neural network includes multiple layers of nonlinear feature mapping layers and time window processing structures with attention mechanism, which can perform spatial interpolation prediction on the current frame after receiving the target spatial trajectory at multiple historical time points. It is suitable for handling abnormal situations such as insufficient frame rate, severe ranging fluctuation or short-term loss of target. It extracts the spatiotemporal continuity features of target movement through temporal feature mapping, and constructs a reasonable prediction of the current time spatial position based on this, to obtain predicted position data. The predicted position data is subjected to temporal consistency constraint, and by analyzing the displacement trend between the predicted position and the spatial position at the last time, it is evaluated whether the speed and acceleration meet the physically acceptable continuous motion model. If the prediction result shows discontinuity in speed change, direction deviation, etc., position smoothing, speed adjustment or interpolation fusion operations are performed to generate time series constraint position data that conforms more to the actual physical motion law. The position confidence score is calculated based on the time series constraint position, and the confidence score model is constructed to quantify the credibility of the current time series constraint position. The scoring process combines multiple factors such as the compression degree of the residual, the matching degree of the prediction bias and the historical trajectory consistency, and the average level of signal-to-noise ratio in the current time window for comprehensive evaluation, and outputs the normalized confidence score to identify the credibility level of the spatial coordinates. The confidence score and the corresponding three-dimensional time series constraint position are jointly output to form the target spatial coordinate result containing the position coordinate value and the confidence level.

[0033] In one example, spectrum sensing and dynamic allocation are combined with target spatial coordinates to obtain a transmission spectrum allocation scheme, including: According to the distance relationship between each camera and the target calculated based on the target spatial coordinates, spectrum occupation sensing is performed to identify available spectrum state information; Based on the available spectrum state information and the target spatial coordinates, a spectrum resource mapping relationship for position sensing is constructed; Real-time quality evaluation is performed on data transmission under the spectrum resource mapping relationship to obtain a quality evaluation index. When the quality evaluation index falls below the preset standard, the intelligent spectrum switching is started by starting the spectrum reallocation process to output the transmission spectrum allocation scheme.

[0034] In this example, taking the target space coordinates as the reference point, combined with the known camera fixed installation coordinates, the actual physical distance between each camera and the target at the current time is determined by the spatial Euclidean distance calculation method, and the camera target distance relationship matrix of the current frame is constructed according to the physical distance. Perform spectrum occupation perception, scan all the frequency bands currently available in the area where the communication module is located one by one, and perform sample power statistics on each sub-band by an improved energy detection algorithm. The energy detection algorithm detects the average level of received signal strength and introduces noise power estimation and dynamic threshold adjustment mechanism to adapt to the characteristics of dynamic and severe spectrum resource changes and multiple interference sources in the Internet of Things scene, identify each spectrum unit in the current time in idle, congestion or weak interference state, and generate a spectrum state information table containing center frequency, bandwidth, availability identifier and interference strength. Fuse the spectrum state information with the target space coordinates to construct a spectrum resource mapping relationship with spatial location as the core constraint, preferentially match the cameras with close target distance, clear field of view and high channel quality to the current idle or high signal-to-noise ratio spectrum unit to form a one-to-one or one-to-many allocation relationship, and mark the sub-band number and pre-allocation result obtained by each camera through the spectrum resource mapping matrix. In the actual data transmission process, real-time quality evaluation is performed on each communication link, with throughput, average delay and packet loss rate as the main input indicators, and a unified quality evaluation index is formed through weighted fusion to reflect the transmission reliability and link stability under the current spectrum allocation scheme. When some sub-bands cause the link quality to decrease significantly due to increased interference, target movement or environmental obstruction, the quality evaluation value tends to approach the preset warning lower limit. When the quality evaluation index of a certain link falls below the set threshold, the spectrum reallocation process is triggered. Rescan all available frequency bands and update the spectrum state table, and on this basis, perform intelligent spectrum switching mechanism according to the latest target space coordinates and link state. The switching mechanism will preferentially select low-interference, low-load and minimum-latency available frequency bands and complete dynamic binding adjustment of data channels. In order to ensure that the video stream does not lose frames, the entire spectrum switching process is completed within milliseconds, and through interruption suppression and data caching mechanism, smooth transition of consecutive frames is ensured. Finally, the updated spectrum allocation scheme is output, reflecting the frequency band number and channel parameters currently used by each camera, as well as the corresponding target space position, signal-to-noise ratio information and spectrum state update timestamp.

[0035] In one example, based on the available spectrum state information and target space coordinates, a location-aware spectrum resource mapping relationship is constructed, including: According to the target space coordinates, the spatial distance values between each camera and the target are calculated, and the signal-to-interference-and-noise ratio parameters of each frequency band are extracted from the available spectrum state information; A quality evaluation matrix is generated by a product operation of the spatial distance values and the signal-to-interference-and-noise ratio parameters; A mathematical optimization model is established based on the quality evaluation matrix, with the goal of maximizing the total transmission capacity of the system, and constraint conditions are set as each camera monopolizing one frequency band and each frequency band being allocated to only one camera, and a spectrum allocation decision matrix is obtained by solving; A distance priority adjustment strategy is implemented on the spectrum allocation decision matrix to form a spectrum allocation matrix; A direct correspondence table of camera numbers and frequency band numbers is converted from the binary allocation identifiers in the spectrum allocation matrix to establish a spectrum resource mapping relationship.

[0036] In this example, based on the target space coordinates, the position data of all deployed cameras are jointly calculated, the real space distance values between each camera and the space target are calculated using the Euclidean geometric model, reflecting the accessibility of the target to each camera and the signal propagation path length, and serving as a parameter for evaluating channel loss; at the same time, the channel quality information of the current available frequency band is extracted in the latest spectrum sensing operation, and the signal-to-interference-and-noise ratio parameter is used as a key indicator to reflect the transmission stability and anti-interference ability of the frequency band at the current time, the effective signal-to-interference-and-noise ratio values of each frequency band in the current environment are extracted, and a signal-to-interference-and-noise ratio parameter vector is constructed. The camera-target distance value and the signal-to-interference-and-noise ratio of the corresponding frequency band are multiplied one by one to generate a two-dimensional quality evaluation matrix, each row of the matrix corresponds to a camera, each column corresponds to a frequency band, and each matrix element represents the expected communication quality indicator of the camera in the current frequency band. The larger the indicator value, the better the channel conditions and the shorter the transmission path in the real scene. Based on the quality evaluation matrix, a mathematical optimization model is constructed to maximize the total transmission capacity, the mathematical optimization model takes the mapping combination of the camera-frequency band as the variable, and the element value of the quality evaluation matrix as the benefit indicator. The objective function is the sum of the quality indicator values of all selected combinations. The constraint conditions include: each camera can only be bound to one frequency band, each frequency band can only be allocated to one camera, the binding relationship between all cameras and frequency bands constitutes a one-to-one decision matrix, and the elements in the matrix are binary values. The value of 1 represents that the camera uses the frequency band, and the value of 0 represents that it is not allocated. By solving the mathematical optimization model, an initial binding scheme between the optimal camera and the frequency band is obtained, which maximizes the transmission potential of the current spectrum resource. Based on the spatial distance data of the target and the camera, a distance priority adjustment strategy is implemented, which prioritizes the high-quality frequency bands of cameras with short distances to the target and small signal propagation path losses based on the initial allocation results, and optimizes the allocation of frequency bands of redundant or long-path cameras through local exchange, to obtain a spectrum allocation matrix. The binary allocation identifiers in the spectrum allocation matrix are mapped one by one to the direct correspondence between the camera number and the frequency band number to construct a spectrum resource mapping table, which marks the specific frequency band number used by each camera in the current system, the corresponding channel characteristics and spatial position parameters.

[0037] In one example, based on the transmission spectrum allocation scheme, the target space coordinates are iteratively optimized, and the target positioning result of the Internet of Things security monitoring system is output, including: According to the transmission quality evaluation index in the transmission spectrum allocation scheme, a multi-objective optimization function is constructed in combination with the positioning accuracy of the target space coordinates and the system processing time delay; The target space coordinates are input as initial position parameters into the Levenberg-Marquardt iterative algorithm, and a multi-objective optimization function is combined to perform iterative optimization; An exponential moving average update is performed on the credibility weight of each camera in the iterative optimization process to determine the target position coordinates. Based on the target position coordinates, the spatial positioning accuracy and confidence value are calculated, combined with target motion trajectory analysis and abnormal behavior detection, to generate the target positioning result of the Internet of Things security monitoring system.

[0038] In this example, the real-time transmission quality evaluation index in the transmission spectrum allocation scheme is extracted and standardized, the transmission quality evaluation index includes multiple key parameters such as the throughput, transmission delay and packet loss rate of the camera link, and a transmission quality function value is formed by weighted fusion; at the same time, the initial value of the target space coordinate obtained based on the weighted distance measurement under the current frame is introduced, and the positioning error is measured by combining the ranging residual sum of squares corresponding to the space coordinate position, and the processing time between the current frame image processing and the positioning output is counted as a system time delay parameter and introduced into the optimization framework. The above three core performance indicators, i.e. the transmission quality evaluation function, the positioning accuracy index and the system processing time delay, are constructed as a joint multi-objective optimization function, balancing the synergy in communication stability, spatial positioning accuracy and response speed, so that the optimization process not only pursues the minimum spatial error, but also can suppress the problems of too high time delay and low spectrum use efficiency. The current target space coordinate is input as the initial position parameter into the Levenberg-Marquardt iterative optimization algorithm, which has the convergence characteristics of gradient descent and Gauss-Newton method, and is suitable for processing nonlinear least squares problems. By dynamically adjusting the damping factor, the convergence stability can be maintained when far away from the optimal solution, and the convergence rate can be improved when close to the minimum value. In each iteration, the optimization function calculates the joint partial derivative of the current position with respect to the three sub-targets, updates the position coordinates until the optimization function value converges or the residual change is less than the set threshold; after each iteration, camera credibility evaluation is performed based on the current predicted position and the observation data of each camera, and the camera weight is updated combined with the signal-to-noise ratio, image integrity and trajectory continuity in the observation frame, and the weight is smoothed by the exponential moving average method, so that the credibility can evolve dynamically over time instead of suddenly changing, so that the positioning contribution of stable visual angle is gradually strengthened and the interference of high error channels is weakened in multiple iterations, and the global optimal position solution is stably output. Based on the error propagation model in multiple dimensions, the uncertainty of the global optimal position solution in X, Y and Z space dimensions is estimated and synthesized into a spatial positioning accuracy value, and the position confidence score of the current frame is calculated according to the final residual value and the historical error mean value. The higher the score value, the more reliable the position solution is under the conditions of ranging, geometry and communication. The target position coordinates are compared and analyzed with the historical trajectory, the behavior model of the target in the continuous time period is modeled by constructing the motion speed, acceleration and direction change sequence, and if a sudden behavior such as abnormal acceleration, path twisting or deviation from the conventional area trajectory is detected, the abnormal behavior recognition module is triggered by setting a behavior threshold, and the current position, positioning accuracy, confidence score and behavior state identifier are output, generating the target positioning result of the Internet of Things security monitoring system.

[0039] Referring to Figure 2 , the embodiment provides an Internet of Things-based security monitoring intelligent analysis system, comprising: A field of view distortion correction module 1 is configured to perform field of view distortion correction on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates; A calculation module 2 is configured to calculate target space coordinates of each camera to the target based on the corrected pixel coordinates; A spectrum sensing and dynamic allocation module 3 is configured to perform spectrum sensing and dynamic allocation in combination with the target space coordinates to obtain a transmission spectrum allocation scheme; An iterative optimization module 4 is configured to perform iterative optimization on the target space coordinates based on the transmission spectrum allocation scheme to output a target positioning result of the Internet of Things security monitoring system.

[0040] In the embodiment, the specific implementation of each unit in the system embodiment is described above in the method embodiment, which will not be repeated here.

[0041] It should be noted that in this paper, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, system, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, system, article or method. Without more limitations, the element defined by the statement "including a" does not exclude the existence of other identical elements in the process, system, article or method including the element.

[0042] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An Internet of Things-based security monitoring intelligent analysis method, characterized in that, The method comprises the following steps: Field of view distortion correction is performed on the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates; Target space coordinates of each camera to the target are calculated based on the corrected pixel coordinates; Spectrum sensing and dynamic allocation are combined based on the target space coordinates to obtain a transmission spectrum allocation scheme; The target space coordinates are iteratively optimized based on the transmission spectrum allocation scheme, and a target positioning result of the Internet of Things security monitoring system is output. 2.The IoT-based security monitoring intelligent analysis method according to claim 1, characterized in that, The method comprises the following steps: Pixel coordinates and brightness information of the target in the multi-camera video stream are extracted to obtain basic observation data of each camera; A video signal intensity measurement model containing light intensity attenuation coefficients and incident angle compensation is established based on the basic observation data, and signal intensity attenuation values of the target under different viewing angles are calculated through the video signal intensity measurement model; Original feature data containing pixel coordinates, brightness values, signal-to-noise ratios and time stamps are extracted according to the signal intensity attenuation values; Field of view distortion correction is performed on the original feature data to obtain corrected pixel coordinates. 3.The IoT-based security monitoring intelligent analysis method according to claim 2, characterized in that, The method comprises the following steps: Pixel coordinates of each camera are extracted from the original feature data, and the pixel coordinates are converted into first coordinates; A continuous field of view distortion correction function containing radial distortion terms and tangential distortion terms is established based on the first coordinates, and distortion coefficients of each camera are calculated through the continuous field of view distortion correction function; The first coordinates are corrected according to the distortion coefficients to obtain second coordinates; The second coordinates are converted back to the pixel coordinate system to obtain corrected pixel coordinates. 4.The IoT-based security monitoring intelligent analysis method according to claim 1, characterized in that, The method comprises the following steps: Round-trip time delays of video signals are detected according to the corrected pixel coordinates, and initial measurement distances of each camera to the target are calculated based on the round-trip time delays; Error correction is performed on the initial measurement distances to obtain environment corrected distances; A distance residual correction algorithm is established based on the environment corrected distances, and adaptive weight coefficients are calculated according to the signal-to-noise ratios and time delay attenuation factors of each camera; Multi-path effect detection and path separation correction are performed on the environment corrected distances, and weighted distance measurement results are calculated in combination with the adaptive weight coefficients; Position prediction is performed based on the weighted distance measurement results to obtain target space coordinates. 5.The IoT-based security monitoring intelligent analysis method according to claim 4, characterized in that, The method comprises the following steps: Amplitude second derivatives and phase jumps of video signals corresponding to the environment corrected distances are analyzed to obtain signal amplitude second derivatives and phase jumps; When the signal amplitude second derivative exceeds a preset signal amplitude threshold and the phase jump exceeds a preset phase threshold, a multi-path effect is detected, and a multi-path effect detection result is output. Based on the multipath effect detection result, a RAKE receiver principle is used for path separation on the video signal to obtain a separated signal of each propagation path and a corresponding path distance measurement value; A path weight coefficient of each propagation path is calculated according to the signal strength of each propagation path, and a comprehensive weight matrix is formed in combination with the adaptive weight coefficient; The path distance measurement value is weighted and fused with the comprehensive weight matrix to obtain a weighted distance measurement result. 6.The IoT-based security monitoring intelligent analysis method according to claim 5, characterized in that, The position prediction is performed based on the weighted distance measurement result to obtain a target spatial coordinate, including: A positioning residual value is calculated according to the weighted distance measurement result, and when the positioning residual value exceeds a residual threshold value, the weighted distance measurement result is input into a video interpolation neural network for position prediction to obtain predicted position data; The predicted position data is subjected to a time sequence consistency constraint to obtain a time sequence constrained position; A position confidence score is calculated based on the time sequence constrained position, and the time sequence constrained position and the confidence score are combined to output a target spatial coordinate. 7.The IoT-based security monitoring intelligent analysis method according to claim 1, characterized in that, The frequency spectrum sensing and dynamic allocation are performed in combination with the target spatial coordinate to obtain a transmission frequency spectrum allocation scheme, including: A distance relationship between each camera and the target is calculated according to the target spatial coordinate, and spectrum occupation sensing is performed to identify available spectrum state information; A position-aware spectrum resource mapping relationship is constructed based on the available spectrum state information and the target spatial coordinate; Real-time quality evaluation is performed on data transmission under the spectrum resource mapping relationship to obtain a quality evaluation index, and when the quality evaluation index falls below a preset standard, an intelligent spectrum switching is started to execute a spectrum reallocation process, and a transmission frequency spectrum allocation scheme is output. 8.The IoT-based security monitoring intelligent analysis method according to claim 7, characterized in that, The position-aware spectrum resource mapping relationship is constructed based on the available spectrum state information and the target spatial coordinate, including: The spatial distance value between each camera and the target is calculated according to the target spatial coordinate, and the signal-to-interference-and-noise ratio parameter of each frequency band is extracted from the available spectrum state information; A quality evaluation matrix is generated through the product operation of the spatial distance value and the signal-to-interference-and-noise ratio parameter; A mathematical optimization model is established based on the quality evaluation matrix with the goal of maximizing the total transmission capacity of the system, and the constraint condition is set as each camera exclusively occupying one frequency band and each frequency band being allocated to only one camera, and a spectrum allocation decision matrix is obtained by solving; A distance priority adjustment strategy is implemented on the spectrum allocation decision matrix to form a spectrum allocation matrix; The binary allocation identifier in the spectrum allocation matrix is converted into a direct correspondence table of camera number and frequency band number to establish a spectrum resource mapping relationship. 9.The IoT-based security monitoring intelligent analysis method according to claim 1, characterized in that, The target positioning result of the Internet of Things security monitoring system is output based on the iterative optimization of the target spatial coordinate according to the transmission frequency spectrum allocation scheme, including: A multi-objective optimization function is constructed according to the transmission quality evaluation index in the transmission frequency spectrum allocation scheme, in combination with the positioning accuracy of the target spatial coordinate and the system processing time delay; The Levenberg-Marquardt iterative algorithm is input with the target spatial coordinate as an initial position parameter, and the iterative optimization is performed in combination with the multi-objective optimization function; An exponential moving average update is performed on the credibility weight of each camera in the iterative optimization process to determine the target position coordinates; Based on the target position coordinates, the spatial positioning accuracy and confidence value are calculated, combined with target motion trajectory analysis and abnormal behavior detection, to generate the target positioning result of the Internet of Things security monitoring system.

10. An Internet of Things-based security monitoring intelligent analysis system, characterized in that, The steps for implementing the Internet of Things-based security monitoring intelligent analysis method of any one of claims 1-9 include: A field of view distortion correction module for field of view distortion correction of the multi-camera video stream of the Internet of Things security monitoring system to obtain corrected pixel coordinates; A calculation module for calculating the target spatial coordinates of each camera to the target based on the corrected pixel coordinates; A spectrum sensing and dynamic allocation module for spectrum sensing and dynamic allocation in combination with the target spatial coordinates to obtain a transmission spectrum allocation scheme; An iterative optimization module for iterative optimization of the target spatial coordinates based on the transmission spectrum allocation scheme to output the target positioning result of the Internet of Things security monitoring system.

Images