A door opening alarm system for storage cabinets

Abstract

The present invention relates to the technical field of storage cabinet anti-theft alarm, and specifically discloses a door opening alarm system for storage cabinets, which obtains a variety of behavior monitoring data when the target storage cabinet is opened through a behavior monitoring module, and then uses a behavior analysis module to analyze the behavior deviation degree, and then performs weighted summation of multiple behavior deviation degrees and corresponding evaluation weights to obtain a behavior abnormality index, and finally updates the behavior evaluation standard or issues an alarm based on the behavior abnormality index, and at the same time updates the evaluation weight based on the cumulative number of times the behavior abnormality index does not exceed a preset abnormal threshold. Compared with the prior art, the evaluation weight and evaluation standard corresponding to each behavior monitoring data in the present invention will be dynamically adjusted with the number of times the target storage cabinet door is opened, solving the problem in the prior art that the storage cabinet door opening warning scheme cannot be combined with the storage cabinet usage frequency to dynamically adjust the behavior recognition strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-theft alarm for storage cabinets, and in particular to a door opening alarm system for storage cabinets. Background Art

[0002] The storage cabinet door opening alarm system is a key technology to ensure storage safety. Its main function is to detect and respond to unauthorized door opening in a timely manner. The system monitors in real time and immediately issues an alarm once an abnormality is detected, effectively preventing property losses and illegal intrusions, while providing important records for subsequent investigations.

[0003] The application of behavior recognition technology has further improved the intelligence level of the storage cabinet door opening alarm. By analyzing and identifying user behavior, the system can distinguish between legal and illegal operations, reduce false alarms, improve the accuracy of alarms and response efficiency, and ensure the safety and order of the storage environment.

[0004] However, before the application of behavior recognition technology, it is necessary to learn the behavior of each user, but storage cabinets are obviously not a high-frequency tool. In practice, when behavior recognition technology is applied to the field of storage cabinet security, it is easy to have large behavior learning errors due to insufficient data samples, which affects the accuracy of the alarm. Therefore, people need a storage cabinet door opening warning solution that can dynamically adjust the behavior recognition strategy based on the frequency of storage cabinet use. Summary of the invention

[0005] The purpose of the present invention is to provide a door opening alarm system for a storage cabinet to solve the following technical problems:

[0006] The storage cabinet door opening warning solution in the prior art cannot dynamically adjust the behavior recognition strategy in combination with the storage cabinet usage frequency.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A door opening alarm system for a storage cabinet, comprising:

[0009] The behavior monitoring module is used to obtain various behavior monitoring data when the target storage cabinet is opened. Each behavior monitoring data corresponds to a behavior evaluation standard and an evaluation weight, wherein the initial evaluation weight of each behavior monitoring data is zero;

[0010] A behavior analysis module is used to obtain the behavior deviation degree corresponding to each behavior monitoring data based on the behavior evaluation standard and the behavior monitoring data;

[0011] The learning warning module is used to perform weighted summation of multiple behavior deviations and corresponding evaluation weights to obtain the behavior anomaly index when the target storage cabinet is opened;

[0012] If the behavior abnormality index does not exceed the preset abnormality threshold, the behavior evaluation standard is updated based on the behavior monitoring data, and the evaluation weight is updated based on the cumulative number of times the behavior abnormality index does not exceed the preset abnormality threshold;

[0013] If the behavior abnormality index exceeds the preset abnormality threshold, an abnormal door opening alarm will be issued.

[0014] As a further solution of the present invention: updating the evaluation weight based on the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold, including:

[0015] Obtain the cumulative number of times that the behavior anomaly index does not exceed the preset anomaly threshold, and the preset upper limit value of the target evaluation weight, where the target evaluation weight is the current evaluation weight to be updated;

[0016] Update the target evaluation weight based on the following formula:

[0017] ;

[0018] Among them, W is the updated target evaluation weight, A is the preset upper limit value of the target evaluation weight, e is a natural constant, k is a preset growth rate control parameter, and n is the cumulative number of times the behavior anomaly index does not exceed the preset anomaly threshold.

[0019] As a further solution of the present invention: the multiple behavior monitoring data include storage cabinet opening time, and the behavior evaluation standard corresponding to the storage cabinet opening time includes a standard opening time period; updating the behavior evaluation standard based on the behavior monitoring data includes:

[0020] Obtain the actual storage cabinet opening time and the historical storage cabinet opening time record, wherein the actual storage cabinet opening time is the storage cabinet opening time for which the behavior deviation degree is to be calculated this time, and the historical storage cabinet opening time record is the time record of each time the user opened the storage cabinet in the past;

[0021] Clustering the actual storage cabinet opening time and the historical storage cabinet opening time records to obtain multiple clusters;

[0022] According to the maximum time period range covered by the historical storage cabinet opening time records in each cluster, multiple standard opening time periods are obtained.

[0023] As a further solution of the present invention: based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including:

[0024] Compare the actual storage cabinet opening time with multiple standard opening time periods;

[0025] If the actual storage cabinet opening time is within a standard opening time period, the behavior deviation corresponding to the actual storage cabinet opening time is zero;

[0026] If the actual storage cabinet opening time is not within any of the standard opening time periods, then the deviation time between the actual storage cabinet opening time and the endpoint of each standard start time period is calculated;

[0027] According to the minimum deviation time, the behavior deviation degree corresponding to the actual storage cabinet opening time is obtained.

[0028] As a further solution of the present invention: the multiple behavior monitoring data include password input interval data, and the behavior evaluation standard corresponding to the password input interval data includes a standard interval feature vector; updating the behavior evaluation standard based on the behavior monitoring data includes:

[0029] Acquire the actual password input interval data and obtain the input interval accumulation vector, wherein the actual password input interval data is the password input interval data of the behavior deviation degree to be calculated this time, and the element value in the input interval accumulation vector represents the cumulative value of the interval length between two adjacent characters in the password input by the user each time in the past and this time;

[0030] The cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold is obtained, and the standard interval feature vector is obtained based on the ratio of the input interval cumulative vector to the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold.

[0031] As a further solution of the present invention: based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including:

[0032] According to the actual password input interval data, an actual input interval vector is obtained, wherein the element value in the actual input interval vector represents the interval length between two adjacent characters in the password input by the user this time;

[0033] According to the difference between the actual input interval vector and the standard interval feature vector, the behavior deviation degree corresponding to the actual password input interval data is obtained.

[0034] As a further solution of the present invention: the multiple behavior monitoring data include person position data, and the behavior evaluation standard corresponding to the person position data includes a standard position feature vector; updating the behavior evaluation standard based on the behavior monitoring data includes:

[0035] Acquire actual person position data to obtain an actual position feature vector, wherein the actual person position data is the person position data of the behavior deviation degree to be calculated this time;

[0036] Obtaining an initial standard position feature vector, where the initial standard position feature vector is the standard position feature vector to be updated;

[0037] The average vector of the initial standard position feature vector and the actual position feature vector is calculated to obtain the standard position feature vector.

[0038] As a further solution of the present invention: the person position data is a monitoring image when the user opens the target storage cabinet; the actual person position data is obtained to obtain the actual position feature vector, including:

[0039] Acquire the actual person position data, and perform image recognition on the actual person position data based on the preset YOLO algorithm to obtain an output vector;

[0040] An actual position feature vector is obtained according to the output vector, wherein each element in the actual position feature vector is used to represent the central coordinates of the character and the size of the character's bounding box when the user opens the target storage cabinet.

[0041] As a further solution of the present invention: based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including:

[0042] According to the difference between the actual position feature vector and the standard position feature vector, the behavior deviation degree corresponding to the actual person position data is obtained.

[0043] Beneficial effects of the present invention:

[0044] The present invention provides a door opening alarm system for a storage cabinet, which obtains a variety of behavior monitoring data when a target storage cabinet is opened through a behavior monitoring module, and then uses a behavior analysis module to obtain the behavior deviation corresponding to each behavior monitoring data based on the behavior evaluation standard according to the behavior monitoring data, and then performs weighted summation on multiple behavior deviations and corresponding evaluation weights to obtain a behavior anomaly index when the target storage cabinet is opened this time. If the behavior anomaly index does not exceed a preset abnormal threshold, the behavior evaluation standard is updated based on the behavior monitoring data, and the evaluation weight is updated based on the cumulative number of times the behavior anomaly index does not exceed the preset abnormal threshold. If the behavior anomaly index exceeds the preset abnormal threshold, an abnormal door opening alarm is performed. Compared with the prior art, although the present invention also uses behavioral evaluation for early warning, the present invention obtains a behavioral anomaly index through weighted synthesis of multiple behavioral monitoring data, and the evaluation weight and evaluation standard corresponding to each behavior monitoring data will be dynamically adjusted with the number of times the target storage cabinet is opened. This means that when the user opens the cabinet door less frequently, the conditions for evaluating behavioral deviations are relatively loose, and each evaluation weight is also relatively low. At this time, even if the standard for evaluating behavioral anomalies is not accurate enough, the system will not obtain a higher behavioral anomaly index, thereby reducing the probability of misjudgment and improving user experience. As the number of sample records of users opening storage cabinets increases, the standard for evaluating behavioral anomalies will become more and more accurate, but the standard for evaluating behavioral deviations will gradually become more stringent, achieving an effect of becoming more accurate the more it is used, thereby solving the problem in the prior art that the storage cabinet door opening early warning scheme cannot be combined with the frequency of storage cabinet use to dynamically adjust the behavior recognition strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention will be further described below in conjunction with the accompanying drawings.

[0046] Figure 1 It is a structural schematic diagram of the door opening alarm system for storage cabinets of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] See also Figure 1 The present invention is a door opening alarm system for a storage cabinet, comprising:

[0049] The behavior monitoring module 110 is used to obtain a variety of behavior monitoring data when the target storage cabinet is opened, each behavior monitoring data corresponds to a behavior evaluation standard and an evaluation weight, wherein the initial evaluation weight of each behavior monitoring data is zero;

[0050] The behavior analysis module 120 is used to obtain the behavior deviation degree corresponding to each behavior monitoring data according to the behavior monitoring data based on the behavior evaluation standard;

[0051] The learning warning module 130 is used to perform weighted summation of multiple behavior deviations and corresponding evaluation weights to obtain a behavior abnormality index when the target storage cabinet is opened this time;

[0052] If the behavior abnormality index does not exceed the preset abnormality threshold, the behavior evaluation standard is updated based on the behavior monitoring data, and the evaluation weight is updated based on the cumulative number of times the behavior abnormality index does not exceed the preset abnormality threshold;

[0053] If the behavior abnormality index exceeds the preset abnormality threshold, an abnormal door opening alarm will be issued.

[0054] It should be noted that the door opening alarm system for the storage cabinet can be separated from the storage cabinet itself and additionally set in a remote server. The door opening alarm system for the storage cabinet and the storage cabinet can be regarded as two independent systems. The door opening alarm system for the storage cabinet in the present invention is only used for door opening alarm, and does not participate in the decision-making process of the storage cabinet opening, that is, whether the storage cabinet door is opened is still controlled by the storage cabinet itself or other facilities (such as verifying the password entered by the user, verifying the NFC signal, verifying the face data, etc.). It is for this reason that all the data mentioned in this article, such as behavior monitoring data, the cumulative number of times that the behavior abnormality index does not exceed the preset abnormal threshold (that is, the total number of times that the behavior assessment has been passed in the past), and the historical storage cabinet opening time record, input interval cumulative vector, initial standard position feature vector and other data that will be involved later also correspond to a potential condition, that is, these data are all obtained on the premise that the user successfully opens the target storage cabinet (such as the input interval data obtained on the premise that the user enters the correct password, and the opening time obtained on the premise that the user successfully opens the cabinet door by swiping the card), and no additional explanation will be given later.

[0055] Compared with the prior art, although the present invention also uses behavioral evaluation for early warning, the present invention obtains a behavioral anomaly index through weighted synthesis of multiple behavioral monitoring data, and the evaluation weight and evaluation standard corresponding to each behavior monitoring data will be dynamically adjusted with the number of times the target storage cabinet is opened. This means that when the user opens the cabinet door less frequently, the conditions for evaluating behavioral deviations are relatively loose, and each evaluation weight is also relatively low. At this time, even if the standard for evaluating behavioral anomalies is not accurate enough, the system will not obtain a higher behavioral anomaly index, thereby reducing the probability of misjudgment and improving user experience. As the number of sample records of users opening storage cabinets increases, the standard for evaluating behavioral anomalies will become more and more accurate, but the standard for evaluating behavioral deviations will gradually become more stringent, achieving an effect of becoming more accurate the more it is used, thereby solving the problem in the prior art that the storage cabinet door opening early warning scheme cannot be combined with the frequency of storage cabinet use to dynamically adjust the behavior recognition strategy.

[0056] Further, in a preferred embodiment, in the above-mentioned learning warning module, updating the evaluation weight based on the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold specifically includes:

[0057] Obtain the cumulative number of times that the behavior anomaly index does not exceed the preset anomaly threshold, and the preset upper limit value of the target evaluation weight, where the target evaluation weight is the current evaluation weight to be updated;

[0058] Update the target evaluation weight based on the following formula:

[0059] ;

[0060] Among them, W is the updated target evaluation weight, A is the preset upper limit value of the target evaluation weight, e is a natural constant, k is a preset growth rate control parameter, and n is the cumulative number of times the behavior anomaly index does not exceed the preset anomaly threshold.

[0061] In the initial state, because the system is still unaware of the user's behavioral habits, all evaluation weights are zero, so that the final calculated behavior anomaly index is also zero, avoiding the occurrence of early warning errors. The significance of the above formula is that as the number of times the user successfully opens the cabinet door increases, the number of samples obtained by the system increases. At this time, the evaluation weight will gradually increase and continue to approach the preset upper limit value, achieving the effect of accurate judgment. The preset upper limit value is the theoretical importance ratio of multiple behavior monitoring data set by humans (the sum of the preset upper limits of multiple behavior monitoring data is 1). Through the above formula, the evaluation weight will continue to approach the preset upper limit value but cannot actually reach the preset upper limit value, so that the evaluation weight will always retain the margin for continued adjustment. The present invention also provides a more detailed embodiment for more clearly explaining the above content: avoiding errors in the calculation of the behavior anomaly index caused by the actual evaluation weight exceeding the element and the upper limit value, which is convenient for subsequent use.

[0062] Further, in a preferred embodiment, the multiple behavior monitoring data include storage cabinet opening time, and the behavior evaluation standard corresponding to the storage cabinet opening time includes a standard opening time period. Based on this, in the learning warning module 130 of this embodiment, the behavior evaluation standard is updated based on the behavior monitoring data, specifically including:

[0063] Obtain the actual storage cabinet opening time and the historical storage cabinet opening time record, wherein the actual storage cabinet opening time is the storage cabinet opening time for which the behavior deviation degree is to be calculated this time, and the historical storage cabinet opening time record is the time record of each time the user opened the storage cabinet in the past;

[0064] Clustering the actual storage cabinet opening time and the historical storage cabinet opening time records to obtain multiple clusters;

[0065] According to the maximum time period range covered by the historical storage cabinet opening time records in each cluster, multiple standard opening time periods are obtained.

[0066] The time when the user opens the storage cabinet door can be used as a reference for behavior evaluation to a certain extent. The standard opening time period is the reasonable time for the user to open the cabinet door. If the user opens the storage cabinet outside the standard opening time period, it can be regarded as abnormal behavior. In this embodiment, the standard opening time period is obtained by clustering the historical storage cabinet opening time records, which can not only obtain multiple standard opening time periods, but also flexibly obtain the floating duration of different standard opening time periods according to the different distribution of the historical storage cabinet opening time records in the clustering cluster, which greatly improves the accuracy.

[0067] Further, in a preferred embodiment, in the behavior analysis module 120, based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including:

[0068] Compare the actual storage cabinet opening time with multiple standard opening time periods;

[0069] If the actual storage cabinet opening time is within a standard opening time period, the behavior deviation corresponding to the actual storage cabinet opening time is zero;

[0070] If the actual storage cabinet opening time is not within any of the standard opening time periods, then the deviation time between the actual storage cabinet opening time and the endpoint of each standard start time period is calculated;

[0071] According to the minimum deviation time, the behavior deviation degree corresponding to the actual storage cabinet opening time is obtained.

[0072] The significance of the above method is that if the user opens the target storage cabinet within the standard opening time period, the behavior is characterized accurately. At this time, the behavior deviation can be directly set to zero to reduce the computing pressure.

[0073] Further, in a preferred embodiment, the multiple behavior monitoring data include password input interval data, and the behavior evaluation standard corresponding to the password input interval data includes a standard interval feature vector. Based on this, in the learning warning module 130 of this embodiment, the behavior evaluation standard is updated based on the behavior monitoring data, specifically including:

[0074] Acquire the actual password input interval data and obtain the input interval accumulation vector, wherein the actual password input interval data is the password input interval data of the behavior deviation degree to be calculated this time, and the element value in the input interval accumulation vector represents the cumulative value of the interval length between two adjacent characters in the password input by the user each time in the past and this time;

[0075] The cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold is obtained, and the standard interval feature vector is obtained based on the ratio of the input interval cumulative vector to the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold.

[0076] The password in the above process is the password required to open the target storage cabinet. The above process is essentially to count the interval between two characters each time the user enters the password, and calculate the average value to obtain the standard interval feature vector. When the method described in this embodiment is adopted, the system only needs to maintain the input interval cumulative vector and the cumulative number of times the behavior abnormality index does not exceed the preset abnormal threshold each time the user enters the password, without saving the interval record each time the password is entered, which greatly reduces the storage pressure of the system.

[0077] Further, in a preferred embodiment, in the behavior analysis module 120, based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, specifically including:

[0078] According to the actual password input interval data, an actual input interval vector is obtained, wherein the element value in the actual input interval vector represents the interval length between two adjacent characters in the password input by the user this time;

[0079] According to the difference between the actual input interval vector and the standard interval feature vector, the behavior deviation degree corresponding to the actual password input interval data is obtained.

[0080] In the above process, the difference between the actual input interval vector and the standard interval feature vector can be expressed in any existing way, such as Euclidean distance, Manhattan distance, etc. Then, by linearly changing the difference value, the behavior deviation degree can be obtained.

[0081] Further, in a preferred embodiment, the multiple behavior monitoring data include person position data, and the behavior evaluation standard corresponding to the person position data includes a standard position feature vector. Based on this, in the learning warning module 130 of this embodiment, the behavior evaluation standard is updated based on the behavior monitoring data, specifically including:

[0082] Acquire actual person position data to obtain an actual position feature vector, wherein the actual person position data is the person position data of the behavior deviation degree to be calculated this time;

[0083] Obtaining an initial standard position feature vector, where the initial standard position feature vector is the standard position feature vector to be updated;

[0084] The average vector of the initial standard position feature vector and the actual position feature vector is calculated to obtain the standard position feature vector.

[0085] In the above process, the position feature vector is a feature vector that represents the relative position of the user and the storage cabinet when the user opens the storage cabinet. The specific structure of the position feature vector is different depending on the specific form of the person's position data. For example, if the person's position data is infrared detection data, the position feature vector can be data in the form of point cloud coordinates.

[0086] In a preferred embodiment, the person position data is a monitoring image when the user opens the target storage cabinet. The above process: obtaining the actual person position data and obtaining the actual position feature vector includes:

[0087] Acquire the actual person position data, and perform image recognition on the actual person position data based on the preset YOLO algorithm to obtain an output vector;

[0088] An actual position feature vector is obtained according to the output vector, wherein each element in the actual position feature vector is used to represent the central coordinates of the character and the size of the character's bounding box when the user opens the target storage cabinet.

[0089] The YOLO (You Only Look Once) algorithm is a real-time target detection system that can predict the location and category of targets in an image through a single forward propagation. Unlike traditional target detection methods, YOLO treats the target detection task as a regression problem, dividing the image into grid cells, each of which is responsible for predicting the target whose center point falls within the grid. Each grid predicts multiple bounding boxes and category probabilities, thereby achieving fast and accurate target detection.

[0090] The YOLO algorithm generates a vector at output, which contains the predicted information of the target in the image. Specifically, each grid cell outputs a vector containing multiple elements, each of which has a specific meaning, including the bounding box coordinates: each grid predicts multiple bounding boxes, each of which consists of four coordinate values, usually (x, y, width, height) or (center_x, center_y, width, height), indicating the center point coordinates of the bounding box and the width and height. That is, the center coordinates of the character and the size of the character's bounding box when the user opens the target storage cabinet in this embodiment.

[0091] This embodiment utilizes the characteristics of the output results of the YOLO algorithm itself, and directly uses the elements in the vector output by the YOLO algorithm as the position feature vector, so that the entire system not only has the accuracy of the YOLO algorithm, but also greatly reduces the difficulty of implementing the system and has good feasibility.

[0092] Further, in a preferred embodiment, in the behavior analysis module 120, based on the behavior evaluation standard, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including:

[0093] According to the difference between the actual position feature vector and the standard position feature vector, the behavior deviation degree corresponding to the actual person position data is obtained.

[0094] Similarly, in the above process, the difference between the actual position feature vector and the standard position feature vector can also be expressed in any existing manner, such as Euclidean distance, Manhattan distance, etc. Then, by linearly changing the difference value, the behavior deviation degree can be obtained.

[0095] Finally, it should be noted that in reality, human activities are random. No matter what kind of behavior monitoring data is used for behavior evaluation, it is difficult to avoid the occurrence of additional situations. For example, employees leaving get off work early due to emergencies may cause abnormal evaluation of the storage cabinet opening time. Users entering the storage cabinet opening password while using a mobile phone may cause abnormal evaluation of the password input interval data. Users may cause abnormal evaluation of the person's location data due to changes in their own physiological state and clothing. At this point, because the present invention performs weighted analysis based on a variety of behavior monitoring data to obtain a behavior abnormality index, the present invention can accommodate the above randomness to obtain a more accurate warning result. In practice, different preset upper limits of evaluation weights can also be set according to the randomness of different behavior monitoring data to further improve the accuracy of the final warning.

[0096] In addition, the specific abnormal door opening alarm method used in practice can be flexibly set according to the specific situation. For example, any existing method such as sending a warning message to the user or locking the target storage cabinet and issuing an alarm prompt sound can be used.

[0097] The present invention provides a door opening alarm system for a storage cabinet, which obtains a variety of behavior monitoring data when a target storage cabinet is opened through a behavior monitoring module, and then uses a behavior analysis module to obtain the behavior deviation corresponding to each behavior monitoring data based on the behavior evaluation standard according to the behavior monitoring data, and then performs weighted summation on multiple behavior deviations and corresponding evaluation weights to obtain a behavior anomaly index when the target storage cabinet is opened this time. If the behavior anomaly index does not exceed a preset abnormal threshold, the behavior evaluation standard is updated based on the behavior monitoring data, and the evaluation weight is updated based on the cumulative number of times the behavior anomaly index does not exceed the preset abnormal threshold. If the behavior anomaly index exceeds the preset abnormal threshold, an abnormal door opening alarm is performed. Compared with the prior art, although the present invention also uses behavioral evaluation for early warning, the present invention obtains a behavioral anomaly index through weighted synthesis of multiple behavioral monitoring data, and the evaluation weight and evaluation standard corresponding to each behavior monitoring data will be dynamically adjusted with the number of times the target storage cabinet is opened. This means that when the user opens the cabinet door less frequently, the conditions for evaluating behavioral deviations are relatively loose, and each evaluation weight is also relatively low. At this time, even if the standard for evaluating behavioral anomalies is not accurate enough, the system will not obtain a higher behavioral anomaly index, thereby reducing the probability of misjudgment and improving user experience. As the number of sample records of users opening storage cabinets increases, the standard for evaluating behavioral anomalies will become more and more accurate, but the standard for evaluating behavioral deviations will gradually become more stringent, achieving an effect of becoming more accurate the more it is used, thereby solving the problem in the prior art that the storage cabinet door opening early warning scheme cannot be combined with the frequency of storage cabinet use to dynamically adjust the behavior recognition strategy.

[0098] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A door opening alarm system for a storage cabinet, characterized in that: include: The behavior monitoring module is used to obtain various behavior monitoring data when the target storage cabinet is opened. Each behavior monitoring data corresponds to a behavior evaluation standard and an evaluation weight, wherein the initial evaluation weight of each behavior monitoring data is zero; A behavior analysis module is used to obtain the behavior deviation degree corresponding to each behavior monitoring data based on the behavior evaluation standard and the behavior monitoring data; The learning warning module is used to perform weighted summation of multiple behavior deviations and corresponding evaluation weights to obtain the behavior anomaly index when the target storage cabinet is opened; If the behavior abnormality index does not exceed the preset abnormality threshold, the behavior evaluation standard is updated based on the behavior monitoring data, and the evaluation weight is updated based on the cumulative number of times the behavior abnormality index does not exceed the preset abnormality threshold; If the behavior abnormality index exceeds the preset abnormality threshold, an abnormal door opening alarm will be issued; The multiple behavior monitoring data include password input interval data, and the behavior evaluation standard corresponding to the password input interval data includes a standard interval feature vector; updating the behavior evaluation standard based on the behavior monitoring data includes: Acquire the actual password input interval data and obtain the input interval accumulation vector, wherein the actual password input interval data is the password input interval data of the behavior deviation degree to be calculated this time, and the element value in the input interval accumulation vector represents the cumulative value of the interval length between two adjacent characters in the password input by the user each time in the past and this time; Obtaining the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold, and obtaining a standard interval feature vector based on the ratio of the input interval cumulative vector to the cumulative number of times that the behavior abnormality index does not exceed the preset abnormality threshold; Based on the behavior evaluation criteria, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including: According to the actual password input interval data, an actual input interval vector is obtained, wherein the element value in the actual input interval vector represents the interval length between two adjacent characters in the password input by the user this time; According to the difference between the actual input interval vector and the standard interval feature vector, the behavior deviation degree corresponding to the actual password input interval data is obtained.

2. The door opening alarm system for a storage cabinet according to claim 1, characterized in that: The evaluation weight is updated based on the cumulative number of times that the behavior anomaly index does not exceed the preset anomaly threshold, including: Obtain the cumulative number of times that the behavior anomaly index does not exceed the preset anomaly threshold, and the preset upper limit value of the target evaluation weight, where the target evaluation weight is the current evaluation weight to be updated; Update the target evaluation weight based on the following formula: ； Among them, W is the updated target evaluation weight, A is the preset upper limit value of the target evaluation weight, e is a natural constant, k is a preset growth rate control parameter, and n is the cumulative number of times the behavior anomaly index does not exceed the preset anomaly threshold.

3. The door opening alarm system for a storage cabinet according to claim 1, characterized in that: The various behavior monitoring data include the storage cabinet opening time, and the behavior evaluation standard corresponding to the storage cabinet opening time includes the standard opening time period; the behavior evaluation standard is updated based on the behavior monitoring data, including: Obtain the actual storage cabinet opening time and the historical storage cabinet opening time record, wherein the actual storage cabinet opening time is the storage cabinet opening time for which the behavior deviation degree is to be calculated this time, and the historical storage cabinet opening time record is the time record of each time the user opened the storage cabinet in the past; Clustering the actual storage cabinet opening time and the historical storage cabinet opening time records to obtain multiple clusters; According to the maximum time period range covered by the historical storage cabinet opening time records in each cluster, multiple standard opening time periods are obtained.

4. The door opening alarm system for a storage cabinet according to claim 3, characterized in that: Based on the behavior evaluation criteria, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including: Compare the actual storage cabinet opening time with multiple standard opening time periods; If the actual storage cabinet opening time is within a standard opening time period, the behavior deviation corresponding to the actual storage cabinet opening time is zero; If the actual storage cabinet opening time is not within any of the standard opening time periods, then the deviation time between the actual storage cabinet opening time and the endpoint of each standard start time period is calculated; According to the minimum deviation time, the behavior deviation degree corresponding to the actual storage cabinet opening time is obtained.

5. The door opening alarm system for a storage cabinet according to claim 1, characterized in that: The various behavior monitoring data include person location data, and the behavior evaluation criteria corresponding to the person location data include a standard location feature vector; updating the behavior evaluation criteria based on the behavior monitoring data includes: Acquire actual person position data to obtain an actual position feature vector, wherein the actual person position data is the person position data of the behavior deviation degree to be calculated this time; Obtaining an initial standard position feature vector, where the initial standard position feature vector is the standard position feature vector to be updated; The average vector of the initial standard position feature vector and the actual position feature vector is calculated to obtain the standard position feature vector.

6. The door opening alarm system for a storage cabinet according to claim 5, characterized in that: The person position data is the monitoring image when the user opens the target storage cabinet; the actual person position data is obtained to obtain the actual position feature vector, including: Acquire the actual person position data, and perform image recognition on the actual person position data based on the preset YOLO algorithm to obtain an output vector; An actual position feature vector is obtained according to the output vector, wherein each element in the actual position feature vector is used to represent the central coordinates of the character and the size of the character's bounding box when the user opens the target storage cabinet.

7. The door opening alarm system for a storage cabinet according to claim 5, characterized in that: Based on the behavior evaluation criteria, the behavior deviation degree corresponding to each behavior monitoring data is obtained according to the behavior monitoring data, including: According to the difference between the actual position feature vector and the standard position feature vector, the behavior deviation degree corresponding to the actual person position data is obtained.

Images