A lamp networking theft prevention monitoring method and system based on all-around identification

Abstract

The application discloses a kind of based on all-around identification's lamp networking theft prevention monitoring method and system, it is related to lamp networking theft prevention monitoring technical field, including the multi-dimensional data of acquisition lamp networking equipment is preprocessed, and the abnormal probability value of multi-dimensional data is obtained;Introduce DS evidence theory, combine abnormal probability value, construct the basic probability assignment of multi-dimensional data, based on basic probability assignment, using improved Euclidean distance calculation evidence distance and support, again based on support, calculate weighting factor, and combine Deng entropy function, obtain Deng entropy after modification, obtain modified Deng entropy and fuse, obtain final value;The application significantly improves the precision of anomaly detection, reduces the false alarm rate, and can quickly respond and trigger emergency warning in the scene of power disconnection and illegal intrusion, thereby further improving the security and intelligent level of lamp networking equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of lamp networking theft prevention monitoring, and particularly relates to a lamp networking theft prevention monitoring method and system based on all-around identification. BACKGROUND

[0002] With the development of smart cities, lamp networking (IoLT) as an important branch of the Internet of Things has gradually become one of the key technologies in public facility management. Lamp networking not only enables intelligent control of lighting scenes such as street lamps, buildings, and parks, but also realizes device state monitoring, environmental perception, and security protection functions through the integration of various sensors and communication technologies. However, with the widespread application of lamp networking systems, the security of devices has become increasingly prominent. Traditional theft prevention mechanisms mainly rely on box door alarms or combined with built-in cameras for monitoring and alarm, but these methods have significant limitations. For example, the pure box door alarm method is extremely prone to false alarms due to false triggering; while the built-in camera can partially solve this problem, it is still difficult to effectively avoid the occurrence of missed reports when illegal intrusions are taken.

[0003] And the existing lamp networking theft prevention monitoring method based on all-around identification relies on a single data input source, making it difficult to cope with complex environmental changes, and has a high false alarm rate, and is slow to react and lacks flexibility in dealing with complex and dynamic scenarios. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a lamp networking theft prevention monitoring method and system based on all-around identification, which solves the problems of the existing lamp networking theft prevention monitoring method based on all-around identification, such as difficulty in coping with complex environmental changes, high false alarm rate, slow reaction, and lack of flexibility.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a lamp networking theft prevention monitoring method based on all-around identification, which includes,

[0008] Collecting multi-dimensional data of lamp networking devices for preprocessing and obtaining abnormal probability values of the multi-dimensional data;

[0009] The multi-dimensional data includes voltage time series signals, image data, and box door switch states;

[0010] The abnormal probability values of the multi-dimensional data are extracted through a feature extraction model and a convolutional neural network;

[0011] The abnormal probability value includes a voltage abnormal probability value, a behavior abnormal probability value and a box door abnormal probability value.

[0012] The DS evidence theory is introduced, the basic probability assignment of multi-dimensional data is constructed in combination with the abnormal probability value, the improved Euclidean distance is used to calculate the evidence distance and the support degree based on the basic probability assignment, the weighted factor is calculated based on the support degree, the Deng entropy is obtained by using the Deng entropy function, the Deng entropy is modified after the Deng entropy, the modified Deng entropy is fused to obtain the final value.

[0013] The final value refers to the final fused basic probability assignment.

[0014] The lamp networking theft prevention monitoring and early warning are carried out based on the final value, and the visual interface is used for display and the database is used for storage.

[0015] As a preferred scheme of the lamp networking theft prevention monitoring method based on the comprehensive recognition, wherein: the improved Euclidean distance is used to calculate the distance between any two basic probability assignments in the proposition subset, and the distance is defined as the evidence distance;

[0016] Based on the Euclidean evidence distance, the maximum distance in all evidence distances is selected, the support degree between the evidence distances is calculated, and the support degree is defined as the similarity;

[0017] Based on the similarity, a symmetric matrix is constructed, the credibility of each basic probability assignment is calculated and normalized, and the credibility is defined as the weighted factor;

[0018] Based on the basic probability assignment, the Deng entropy of each basic probability assignment is calculated by using the Deng entropy function, and then the Deng entropy is modified by using the weighted factor;

[0019] Based on the modified Deng entropy, the weight of each basic probability assignment is calculated for weighting to obtain a fusion value, and the operation is repeated to obtain the fusion values corresponding to the three abnormal probability values.

[0020] Based on the fusion value, the conflict coefficient is calculated, the fusion is performed by using the Dempster combination rule in combination with the conflict coefficient to obtain the final value.

[0021] As a preferred scheme of the lamp networking theft prevention monitoring method based on all-around recognition provided in the application, wherein: the lamp networking theft prevention monitoring early warning based on the final value uses a field method to set an alarm threshold value, when the final value is greater than or equal to the alarm threshold value, an alarm is triggered, and after the energy storage short message emergency sending mechanism is activated, an alarm signal of “potential power failure damage risk” is generated and sent to the management personnel, otherwise the monitoring continues.

[0022] The activated energy storage short message emergency sending mechanism uses a super capacitor as an energy storage unit, and installs a low-power processor and a planar antenna for the lamp networking device.

[0023] The energy storage unit provides power for the low-power processor, so that when the main power is cut off, enough power can be provided for the planar antenna, and the signal is amplified based on the radio frequency planar antenna driving circuit to complete the sending of a single frame of wireless signal and emergency message.

[0024] The short message includes device identification, alarm timestamp, image data and total harmonic distortion change record.

[0025] As a preferred scheme of the lamp networking theft prevention monitoring method based on all-around recognition provided in the application, wherein: the display through the visual interface means that the short message is displayed using a visual tool, and red is used as the alarm color and green is used as the continue monitoring color.

[0026] As a preferred scheme of the lamp networking theft prevention monitoring method based on all-around recognition provided in the application, wherein: the storage using a database means that the alarm information is stored using a database, and after storage, the alarm information is added with the unique ID of the lamp networking device.

[0027] As a preferred scheme of the lamp networking theft prevention monitoring method based on all-around recognition provided in the application, wherein: the pre-processing of the multi-dimensional data of the lamp networking device means that the voltage time sequence signal of the power supply network in the lamp networking device is collected through a voltage sensor.

[0028] The image data around the lamp networking device is collected through a multi-view endoscope camera, and the box door opening and closing state of the lamp networking device is collected through a door magnetic switch.

[0029] All data is denoised and normalized, the voltage time sequence signal is segmented using a window function, and the frequency domain is converted after applying short-time Fourier transform to obtain the frequency spectrum, including the fundamental wave and harmonic component.

[0030] As a preferred scheme of the lamp networking theft prevention monitoring method based on all-around recognition provided in the application, wherein: the abnormal probability value of the multi-dimensional data is obtained based on the fundamental wave and harmonic component, square average values are respectively obtained, and root values of the square average values are obtained by taking square roots, to define effective values, a THD calculation formula is combined with the effective values to calculate a harmonic distortion rate;

[0031] Based on the spectrum, the standard deviation is obtained after the mean value is obtained, and based on the spectrum, the spectral density change value is calculated, and the harmonic distortion rate and the standard deviation are combined to form an input vector;

[0032] Based on the input vector, a convolutional neural network is used to obtain the voltage abnormal probability value ;

[0033] Based on the image data, a deep convolutional neural network is used to obtain the abnormal probability value of each behavior in the image, and the sum of the abnormal probability values of the behaviors is obtained to obtain the behavior abnormal probability value ;

[0034] Further based on the image data, an ROI detection technology is used to determine whether a person is close, and a door magnetic switch is combined to determine whether the opening and closing of the box door is authorized, an encoding technology is used to encode the determination result, when there is no person close and no authorization, the box door is opened, the encoding value is set to 1, otherwise 0, and the result is defined as the box door abnormal probability value ;

[0035] Based on the voltage, behavior and box door abnormal probability values, the BPA function form of the DS evidence theory to which each data belongs is constructed to obtain the basic probability assignment of each data, and a universal set and a proposition subset are constructed.

[0036] The voltage proposition set includes normal fluctuation and abnormal fluctuation, the image proposition set includes face covering behavior, normal, latent behavior and fast moving behavior, and the box door proposition set includes box door opening and box door closing.

[0037] In a second aspect, the application provides a lamp networking theft prevention monitoring system based on all-around recognition, comprising,

[0038] The acquisition and acquisition module is used for collecting multi-dimensional data of the lamp networking device for preprocessing, and obtaining the abnormal probability value of the multi-dimensional data;

[0039] The calculation and fusion module is used for introducing the DS evidence theory, combining the abnormal probability value, constructing the basic probability assignment of the multi-dimensional data, using the improved Euclidean distance based on the basic probability assignment to calculate the evidence distance and the support degree, and based on the support degree, calculating the weighting factor, and combining the Deng entropy function to obtain the Deng entropy after modification to obtain the modified Deng entropy for fusion to obtain the final value.

[0040] Display and storage module, based on the final value, carry out lamp networking theft prevention warning, and display through visual interface and store through database.

[0041] In a third aspect, the present application provides a computer device, comprising a memory and a processor, and the memory stores a computer program, wherein: the computer program is executed by the processor to implement any step of the all-around identification based lamp networking theft prevention monitoring method according to the first aspect of the present application.

[0042] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein: the computer program is executed by the processor to implement any step of the all-around identification based lamp networking theft prevention monitoring method according to the first aspect of the present application.

[0043] The present application has the beneficial effects that: the abnormal probability value of data is corrected and fused by the Deng entropy function, the confidence evaluation of data is optimized, and the conflict problem between multi-source information is solved by the improved Euclidean distance, which makes the present application significantly improve the accuracy of anomaly detection, reduce the false positive rate, and quickly respond and trigger emergency warning in the scene of power disconnection and illegal intrusion, thereby further improving the security and intelligent level of the lamp networking device. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows: obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0045] Fig. 1 The flowchart of the all-around identification based lamp networking theft prevention monitoring method in embodiment 1.

[0046] Fig. 2 The structural diagram of the all-around identification based lamp networking theft prevention monitoring system in embodiment 1.

[0047] Fig. 3 The flowchart of the final value based lamp networking theft prevention warning in embodiment 1. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0049] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present application.

[0050] Secondly, the "one embodiment" or "embodiment" referred to herein is intended to encompass a particular feature, structure, or characteristic described herein, which can be included in at least one implementation of the present application. The appearances of the "in one embodiment" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another.

[0051] Embodiment 1, refer to Figs. 1-3 As a first embodiment of the present application, the embodiment provides a lamp networking theft prevention monitoring method based on all-around recognition, comprising the following steps:

[0052] S1, collecting multi-dimensional data of lamp networking equipment for preprocessing, and obtaining abnormal probability value of multi-dimensional data;

[0053] Specifically, collecting multi-dimensional data of lamp networking equipment for preprocessing means collecting voltage time sequence signals of power supply network in lamp networking equipment through a voltage sensor;

[0054] Collecting image data around the lamp networking equipment through a multi-lens endoscope camera, and collecting the box door opening and closing state of the lamp networking equipment through a door magnetic switch;

[0055] All data are denoised and normalized, the voltage time sequence signals are segmented using a window function, and the frequency domain conversion is applied after the short-time Fourier transform, to obtain the frequency spectrum, including the fundamental wave and harmonic components.

[0056] Through the collaborative input of the three types of information sources of voltage, image and door magnetic, the coverage recognition of the three types of damage signs of "abnormal power supply", "suspicious behavior" and "physical intervention" can be effectively realized, thereby constructing a multi-dimensional redundant error mechanism, significantly reducing the judgment deviation caused by the missing information and the misjudgment rate. The original physical signal often has high-frequency noise, electromagnetic interference and imaging error. Through normalization and filtering denoising in the preprocessing process, more identifiable input is provided for the deep neural network to learn the discriminative features, and the overall robustness is improved. The frequency domain features such as harmonic content, spectral density change and THD can more sensitively reflect the nonlinear disturbance caused by device operation, power supply attack, etc.

[0057] Further, obtaining the abnormal probability value of multi-dimensional data means that based on the fundamental wave and harmonic components, the square average value is respectively taken in the square root to obtain the respective root mean square value, which is defined as the effective value. The harmonic distortion rate is calculated using the THD calculation formula combined with the effective value, and the formula is:

[0058]

[0059] In the formula, represents the harmonic distortion rate, represents the fundamental effective value, represents the effective value of the first harmonic component;

[0060] Based on the spectrum, the mean is obtained, and then the standard deviation is taken, and further based on the spectrum, the spectral density change value is calculated, and the harmonic distortion rate and the standard deviation are combined to form an input vector;

[0061] The formula for calculating the spectral density change value is:

[0062]

[0063] In the formula, represents the spectral density change value at time , represents the highest frequency range value (taken in the power standard range) of the sum, represents the lowest frequency range value (taken in the power standard range) of the sum, represents the spectral value at frequency at time , represents the spectral value at frequency at time , represents the absolute value, represents the time interval;

[0064] Based on the input vector, a convolutional neural network is used to obtain a voltage anomaly probability value ;

[0065] Based on the image data, a deep convolutional neural network is used to obtain the anomaly probability value of each behavior in the image by summing up the anomaly probability values of each behavior in the image, to obtain a behavior anomaly probability value ;

[0066] The construction and training content of the convolutional neural network and the deep convolutional neural network are as follows:

[0067] A 1D-CNN network architecture is used to construct an extraction model, and the input is set to the input vector and the output is set to the voltage anomaly probability;

[0068] A convolutional neural network is constructed by using a VGG network architecture, and the input is set to the image data and the output is set to the behavior anomaly probability;

[0069] ​​​The extraction and convolution neural network models are trained by using a training data set respectively, a loss function is defined, and the parameters of the extraction and convolution neural network models are iteratively optimized by using an Adam optimizer until the loss converges;

[0070] By inputting the voltage and image data into the neural network dedicated to the structure, the interference problem between modalities is avoided, and the optimal features are extracted in the most suitable modeling space for each type of data, improving the sensitivity and discrimination of the "electrical disturbance behavior" and "personnel behavior anomaly". The voltage signal modeling and image behavior modeling are structured and processed separately, and are unified in the form of "abnormal probability", providing a standardized and unified interface for subsequent DS evidence theory fusion, confidence evaluation and Deng entropy correction, thereby laying a good foundation for subsequent early warning;

[0071] Further based on image data, ROI detection technology is used to determine whether there is personnel close, and in combination with the door magnetic switch, it is determined whether the opening and closing of the box door is authorized, and the judgment result is encoded using coding technology. When there is no personnel close and unauthorized, the box door is opened, the code value is set to 1, otherwise it is 0, and the result is defined as the box door abnormal probability value ;

[0072] Based on the voltage, behavior and box door abnormal probability values, the BPA function form of the DS evidence theory to which each data belongs is constructed, the basic probability assignment of each data is obtained, and the universal set and proposition subset are constructed.

[0073] The universal set formula is:

[0074]

[0075] In the formula, represents the universal set, represents the abnormal proposition set, represents the normal proposition set.

[0076] The proposition subset formula is:

[0077]

[0078] In the formula, represents the basic probability assignment of the abnormal proposition set of the th data, represents the abnormal probability of the th data, represents the basic probability assignment of the normal proposition set of the th data, represents the basic probability assignment of the uncertain proposition set of the th data.

[0079] The voltage proposition set includes normal fluctuation and abnormal fluctuation, the image proposition set includes face covering behavior, normal, latent behavior, fast moving behavior, and the box door proposition set includes box door opening and box door closing.

[0080] The frequency spectrum features are extracted by Fourier transform, and the input vector is constructed by combining the THD and the spectrum density change amount parameters. The voltage anomaly probability is output by the 1D-CNN model. This step strengthens the modeling capability of the perturbation signal, especially in the initial power failure, peak interference and other extremely short time effect scenarios, and can still maintain a high recognition rate. The image recognition channel uses a VGG deep network to model the behavior features of the person, combines with the ROI detection to lock the target area, and realizes global anomaly judgment by weighted sum of behavior probability. This mechanism can distinguish complex behavior patterns such as latency, covering, and fast intrusion, and reduce the false alarm rate caused by image blocking. The box door state is modeled by combining the door magnetic switch and the personnel proximity state coding, and the box door anomaly probability value is obtained by logical judgment and coding mapping, so as to realize the dual verification mechanism of “physical behavior + visual behavior” and enhance the judgment credibility of the recognition.

[0081] S2, introduce DS evidence theory, combine with anomaly probability value, construct basic probability assignment of multi-dimensional data, based on basic probability assignment, use improved Euclidean distance to calculate evidence distance and support, then based on support, calculate weighting factor, combine with Deng entropy function, get Deng entropy after modification, get modified Deng entropy for fusion, get final value;

[0082] Specifically, the improved Euclidean distance is used to calculate the evidence distance and the support, and then based on the support, the weighting factor is calculated, and the Deng entropy function is combined to obtain the Deng entropy after modification, and the modified Deng entropy is fused to obtain the final value. Based on the proposition subset, the improved Euclidean distance is used to calculate the distance between any two basic probability assignments in the proposition subset, which is defined as the evidence distance, and the formula is:

[0083]

[0084] In the formula, the evidence distance between the basic probability assignments of the first and the second data, the Euclidean distance square, the proposition subset is a subset of the universal set , the basic probability assignment of the first data in the proposition subset , the basic probability assignment of the first data in the proposition subset , the basic probability assignment of the first data in the proposition subset .​

[0085] The existing Jousselme distance needs to introduce a weight matrix when calculating the Jaccard similarity, and the traditional Euclidean distance is easy to treat all subset weights equally, and does not consider the structural characteristics of the conflict contribution of each subset, and in a high-order proposition set (such as a large set), a large number of subsets are "compound propositions", and simple difference may cover up the structural differences;

[0086] The traditional Euclidean distance formula is:

[0087]

[0088] The present application is improved on this basis, first, the set And any two basic probability assignments are regarded as vectors on the subset of propositions, then introduce the scaling factor To ensure that the distance function satisfies the non-negativity and symmetry of the metric space, and finally get the improved Euclidean distance formula;

[0089] Based on the Euclidean evidence distance, select the maximum distance in all evidence distances, calculate the support degree between evidence distances, define as similarity, the formula is:

[0090]

[0091] In the formula, The support degree between the evidence distances of the first And the Data, Indicates the maximum distance selected;

[0092] Based on the similarity, a symmetric matrix is constructed, the credibility of each basic probability assignment is calculated and normalized, and is defined as a weighted factor, the formula is:

[0093]

[0094] In the formula, The credibility of the basic probability assignment, The basic probability assignment of the first Data, Indicates the total number of data;

[0095] Based on the basic probability assignment, the Deng entropy of each basic probability assignment is calculated using the Deng entropy function, and then the Deng entropy is modified using the weighted factor;

[0096] The formula for calculating the Deng entropy of each basic probability assignment is:

[0097]

[0098] wherein, denotes the Deng entropy of the basic probability assignment of the i th data, denotes the subset of the proposition, is the subset of the universal set, denotes the empty set, denotes the logarithm function (base 2), denotes the power set size of the subset of the proposition,

[0099] The formula for correcting the Deng entropy using the weighting factor is:

[0100]

[0101] wherein, denotes the Deng entropy of the basic probability assignment of the i th data after correction,

[0102] Based on the corrected Deng entropy, the weight of each basic probability assignment is calculated for weighting to obtain a fusion value, and the operation is repeated to obtain fusion values corresponding to the three anomaly probabilities;

[0103] The formula for calculating the weight of each basic probability assignment is:

[0104]

[0105] wherein, denotes the weight of the Deng entropy of the i th data after correction, denotes the total number of data; The formula for the fusion value is:

[0106]

[0107]

[0108] wherein, denotes the fusion value of the subset of the proposition of the i th data, Based on the fusion value, a conflict coefficient is calculated, and the fusion is performed using the Dempster combination rule in combination with the conflict coefficient to obtain a final value;

[0109] The formula for the conflict coefficient is:

[0110]

[0111]

[0112] wherein, denotes the conflict coefficient,​​​​​​​ This represents the intersection operation. Representing a subset of propositions , , The intersection, To represent the empty set, Indicates the first A subset of propositions for data The fusion value, Indicates the first A subset of propositions for data The fusion value;

[0113] The fusion formula is as follows:

[0114]

[0115] In the formula, This represents the final value.

[0116] The BPA difference between all proposition subsets is calculated by improved Euclidean distance, and the power set mapping structure is introduced to preserve the topological structure information of the proposition set at the semantic level. The improvement not only preserves the simplicity and symmetry of the Euclidean distance, but also enhances the discrimination ability between high-order compound propositions through the square metric form, and reflects the relative consistency degree of information through the normalization of the maximum evidence distance. The obtained support degree can intuitively evaluate the relative consistency degree of information, and provide data support for subsequent credibility allocation. The support degree is used to construct a symmetric matrix, and the credibility of each piece of evidence is calculated and normalized to obtain a weighting factor. The factor is used to adjust the weight of different information sources in the final entropy fusion, ensuring that the information fusion result not only preserves the influence of the dominant signal, but also is not disturbed by low-confidence evidence. Compared with the traditional average weighting mechanism, the method avoids the "dilution" of strong conflict data, and improves the robustness of the fusion. The uncertainty of each basic probability assignment is quantified by Deng entropy. As a structural entropy, Deng entropy can describe the information fuzziness caused by non-specific proposition subsets, and is superior to Shannon entropy and Hartley entropy in expressing multiple propositions. The influence degree of the overall uncertainty of each source information can be dynamically adjusted by modifying the Deng entropy with the weighting factor. The BPA after entropy modification not only reflects the uncertainty of the proposition itself, but also integrates the confidence state of the related support data, so that the entropy value has stronger context relevance and discrimination ability. The weight corresponding to each proposition subset is calculated based on the modified Deng entropy, and is used to obtain the preliminary fusion value by weighting. The fusion value not only contains the confidence probability, but also embeds the structural entropy and weight information, realizing the deep coupling expression of probability and fuzziness. Finally, the intersection degree of propositions between different information sources is quantified by the conflict coefficient, and information fusion is performed based on the Dempster combination rule to calculate the final fusion decision value, so that the invention can significantly improve the recognition accuracy and response efficiency of the lamp networking anti-theft system in dealing with complex damage behaviors.

[0117] S3, based on the final value, carry out lamp networking anti-theft monitoring and early warning, and display through a visual interface and store in a database.

[0118] Specifically, based on the final value, the lamp networking anti-theft monitoring and early warning refers to setting an alarm threshold using the field method. When the final value is greater than or equal to the alarm threshold, the alarm is triggered, and after activating the energy storage short message emergency sending mechanism, the "potential power failure damage risk" alarm signal is generated and sent to the management personnel. Otherwise, continue to monitor.

[0119] Activating the energy storage short message emergency sending mechanism refers to using a super capacitor as an energy storage unit, and installing a low-power processor and a planar antenna for the lamp networking device.

[0120] The energy storage unit provides power for the low-power processor, so that when the main power supply is cut off, enough power can be provided for the planar antenna, and the signal is amplified based on the radio frequency planar antenna driving circuit to complete the sending of a single frame of wireless signal and emergency message (the processing process of short message sending is controlled by the low-power processor, and the low-power processor stores and distributes energy through cooperation with the energy storage unit to ensure that short message sending can still be maintained in the case of power failure or power outage);

[0121] The short message includes device identification, alarm timestamp, image data and total harmonic distortion change record.

[0122] Through the combination of super capacitor + low-power MCU + planar antenna, the application can still independently complete the early warning sending in the moment of main power interruption, solve the passive defect of "power failure paralysis" of the existing lamp networking system, bind the final fusion probability value through the alarm threshold defined by the field method, trigger the alarm only when the fusion reasoning reaches high confidence output, significantly reduce the false alarm rate, improve the system reliability, the short message sent contains image frame abstract and power side THD change record in addition to device identification and timestamp, forming an "structured + minimized" emergency data format, which can still transmit key safety information completely under the condition of network resource shortage, and the structure adopted can be adapted to various lamp networking terminals, without relying on continuous power supply, with the characteristics of self-adaptation and low maintenance, so that the application can be applied to deployment requirements in remote, unattended or high damage risk scenarios.

[0123] Further, the short message is displayed using a visualization tool, and red is used as the alarm color and green is used as the continue monitoring color.

[0124] Red is used as the visual highlight identification of the alarm event, and green represents the normal monitoring state, both of which have good cognitive difference and cross-cultural universality, and conform to the color indication specification in the ISO standard that "red = warning, green = safety".

[0125] Further, the alarm information is stored using a database, and after storage, the unique ID of the lamp networking device is added to the alarm information.

[0126] By introducing the database structured management and device unique ID mechanism in the alarm information storage process, the performance of the application in data accuracy, manageability and traceability is significantly improved.

[0127] The embodiment also provides a lamp networking anti-theft monitoring system based on all-around recognition, comprising:

[0128] The acquisition and acquisition module is used for collecting multi-dimensional data of the lamp networking device for preprocessing, and acquiring the abnormal probability value of the multi-dimensional data.

[0129] The computing fusion module is used to introduce the DS evidence theory, combine the abnormal probability value, construct the basic probability assignment of the multi-dimensional data, calculate the evidence distance and the support degree based on the basic probability assignment and the improved Euclidean distance, calculate the weighted factor based on the support degree, and obtain the modified Deng entropy after the Deng entropy is obtained by combining the Deng entropy function, to obtain the final value.

[0130] The display and storage module performs lamp networking theft prevention monitoring and early warning based on the final value, and displays through a visual interface and stores in a database.

[0131] The embodiment also provides a computer device suitable for the lamp networking theft prevention monitoring method based on all-around recognition, which comprises a memory and a processor.

[0132] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device which are connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, a trackball or a touchpad arranged on the shell of the computer device. In addition, the input device can also be an external keyboard, a touchpad or a mouse, etc.

[0133] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for implementing the omnibearing identification-based lamp networking theft prevention monitoring method. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk, or an optical disk.

[0134] It should be noted that the above embodiment is only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A omnibearing recognition-based lamp networking theft prevention monitoring method, characterized in that: Comprising, Collecting multi-dimensional data of the lamp networking device for preprocessing; The multi-dimensional data comprises voltage time series signals, image data and box door switch states; An abnormal probability value of the multi-dimensional data is obtained by squaring and averaging the fundamental wave and harmonic components respectively, taking the square root to obtain respective root mean square values, defining the values as effective values, using a THD calculation formula combined with the effective values to calculate a harmonic distortion rate; Based on the frequency spectrum, an average value is obtained and then a standard deviation is taken, and further based on the frequency spectrum, a spectral density change value is calculated, and then combined with the harmonic distortion rate and the standard deviation to form an input vector; Based on the input vector, a convolutional neural network is used to obtain a voltage anomaly probability value ; Based on the image data, using a deep convolutional neural network to obtain the abnormal probability value of each behavior in the image is summed to obtain the behavior abnormal probability value ; Further based on image data, using ROI detection technology, judge whether there is a person close, and combine the door magnetic switch to judge whether the opening and closing of the box door is authorized, encode the judgment result using coding technology, when there is no person close and unauthorized, the box door is opened, the coding value is set to 1, otherwise 0, and the result is defined as the box door abnormal probability value ; Based on the voltage, behavior and box door abnormal probability values, a BPA function form of a DS evidence theory to which each data belongs is constructed to obtain a basic probability assignment of each data, and a universal set and a proposition subset are further constructed; The voltage proposition set comprises normal fluctuations and abnormal fluctuations, the image proposition set comprises face occlusion behaviors, normal behaviors, latent behaviors and fast moving behaviors, and the box door proposition set comprises a box door opening and a box door closing; An improved Euclidean distance is used to calculate a distance between any two basic probability assignments in the proposition subset, which is defined as an evidence distance, and the formula is: In the formula, Indicates the first and Evidence distance between the basic probability assignments of the data points This represents the square of the Euclidean distance. Representing a subset of propositions It's the complete series. a subset of Indicates the first A subset of propositions for data Basic probability assignment, Indicates the first A subset of propositions for data The basic probability assignment; Based on the Euclidean evidence distance, the maximum distance in all evidence distances is selected, and then a support degree between the evidence distances is calculated, which is defined as a similarity, and the formula is: wherein represents the first and support between the evidence distances of the data, represents the maximum distance selected; Based on the similarity, a symmetric matrix is constructed, a credibility of each basic probability assignment is calculated for normalization, which is defined as a weighting factor, and the formula is: wherein represents the reliability of the basic probability assignment, represents the basic probability assignment of the th data, represents the total number of data; Based on the basic probability assignment, a Deng entropy of each basic probability assignment is calculated using a Deng entropy function, and then the Deng entropy is corrected using the weighting factor; The formula for calculating the Deng entropy of each basic probability assignment is: wherein denotes the Deng entropy of the basic probability assignment of the th data, denotes the power set size of the subset of the universal set denotes the empty set, denotes the logarithm function, denotes the power set size of the subset of the universal set The formula for correcting the Deng entropy using the weighting factor is: wherein denotes the Deng entropy of the basic probability assignment of the i-th data after revision; denotes the Deng entropy of the basic probability assignment of the i-th data after revision; Based on the corrected Deng entropy, a weight of each basic probability assignment is calculated for weighting to obtain a fusion value, and the operation is repeated to obtain fusion values corresponding to the three abnormal probability values; The formula for calculating the weight of each basic probability assignment is: wherein represents the weight of the modified Deng entropy of the i th data; The formula for the fusion value is: wherein represents the fusion value of the proposition subset of the first data . Based on the fusion value, a conflict coefficient is calculated, and a fusion is performed using a Dempster combination rule combined with the conflict coefficient to obtain a final value; The formula for the conflict coefficient is: In the formula, Indicates the conflict coefficient. This represents the intersection operation. Representing a subset of propositions , , The intersection, Represents the empty set. Indicates the first A subset of propositions for data The fusion value, Indicates the first A subset of propositions for data The fusion value; The fusion formula is: In the formulae, denotes the final value; The final value is a final fused basic probability assignment; Based on the final value, lamp networking anti-theft monitoring and early warning is performed, that is, an alarm threshold is set using a field method, when the final value is greater than or equal to the alarm threshold, an alarm is triggered, a low-energy short message emergency sending mechanism is activated, and an "potential power failure damage risk" alarm signal is generated and sent to a management personnel, otherwise, the monitoring is continued; The low-energy short message emergency sending mechanism is activated, that is, a super capacitor is used as an energy storage unit, and a low-power processor and a planar antenna are installed on the lamp networking device; The energy storage unit provides power for the low-power processor, so that when the main power is cut off, sufficient power can be provided for the planar antenna, and based on a radio frequency planar antenna driving circuit, a signal is amplified, a single frame wireless signal and an emergency message are sent, and are displayed through a visual interface and stored through a database; The short message comprises a device identifier, an alarm timestamp, image data and a change record of a total harmonic distortion rate.

2. The omnidirectional recognition based lamp networking theft prevention monitoring method of claim 1, wherein: The display through the visual interface refers to displaying the short message using a visual tool, and using red as the alarm color and green as the continue monitoring color. 3.The omnidirectional recognition based anti-theft monitoring method for lamp networking according to claim 2, characterized in that: The storage using the database refers to storing the alarm information using the database, and adding the unique ID of the lamp networking device to the alarm information after storage.

4. The omnidirectional recognition based lamp networking theft prevention monitoring method of claim 3, wherein: The preprocessing of the multi-dimensional data of the lamp networking device refers to collecting the voltage time sequence signal of the power supply grid in the lamp networking device through a voltage sensor; Collecting image data around the lamp networking device through a multi-lens endoscope camera, and collecting the box door opening and closing state of the lamp networking device through a door magnetic switch; All data are denoised and normalized, the voltage time sequence signal is segmented using a window function, and the frequency domain conversion is applied after short-time Fourier transform to obtain the frequency spectrum, including the fundamental wave and harmonic components.

5. A theft monitoring system based on omnidirectional recognition and lamp networking, based on the theft monitoring method based on omnidirectional recognition and lamp networking of any one of claims 1-4, characterized in that: It comprises, The acquisition module is used for preprocessing the multi-dimensional data of the lamp networking device and acquiring the abnormal probability value of the multi-dimensional data; The calculation fusion module is used for introducing the DS evidence theory, combining the abnormal probability value, constructing the basic probability assignment of the multi-dimensional data, calculating the evidence distance and support degree using the improved Euclidean distance based on the basic probability assignment, calculating the weighted factor based on the support degree, and obtaining the Deng entropy after modification by combining the Deng entropy function, obtaining the modified Deng entropy for fusion, and obtaining the final value. The display and storage module performs lamp networking anti-theft monitoring and early warning based on the final value, and displays through a visual interface and stores using a database.

6. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor executes the computer program to realize the steps of the lamp networking anti-theft monitoring method based on comprehensive identification according to any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the lamp networking anti-theft monitoring method based on comprehensive identification according to any one of claims 1-4.

Images