Non-motor vehicle theft behavior detection method, device and equipment and storage medium

Abstract

The application discloses a non-motor vehicle theft behavior detection method and device, equipment and a storage medium, and belongs to the technical field of video monitoring. The method comprises the following steps: modeling a driver and a non-motor vehicle in a snapshot image in which the driver rides the non-motor vehicle, to obtain a driver biological feature library and a non-motor vehicle feature library; based on the driver biological feature library and the non-motor vehicle feature library, filing driver biological features and corresponding non-motor vehicle features belonging to the same driver, to obtain non-motor vehicle archives corresponding to the drivers respectively; clustering the non-motor vehicle archives, to obtain the number of clustering clusters of the non-motor vehicle archives; and if the number of clustering clusters is greater than a preset threshold, determining that the driver corresponding to the non-motor vehicle archives has non-motor vehicle theft behavior. The application aims to improve the non-motor vehicle theft behavior judgment accuracy and solve the technical problem of low non-motor vehicle theft behavior judgment accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of video monitoring, in particular to a non-motor vehicle theft behavior detection method, device, equipment and storage medium. BACKGROUND

[0002] Non-motor vehicles are widely used in people's daily travel due to their energy-saving, environmental protection, convenient travel, small size and flexibility. Non-motor vehicles greatly facilitate people's production and life, but also bring great pressure to the management of relevant departments. Mainly in three aspects, first, it is difficult to prevent, because the self-prevention consciousness of non-motor vehicle owners is not strong, and non-motor vehicles are small and easy to be stolen, resulting in non-motor vehicles being stolen; second, the range is wide; third, it is difficult to crack the case.

[0003] At present, by binding non-motor vehicles and non-motor vehicle owners one by one, if the current user of the non-motor vehicle is not equal to the non-motor vehicle owner bound thereto, it is determined that the current user of the non-motor vehicle has a theft behavior. However, in actual use, there are often behaviors of one vehicle and multiple riders, and mutual exchange of riding. In the above-mentioned situation, based on the one-to-one binding of non-motor vehicles and non-motor vehicle owners to determine whether there is a theft behavior, a misjudgment may occur, resulting in low accuracy of non-motor vehicle theft behavior judgment. SUMMARY

[0004] Therefore, the present application provides a non-motor vehicle theft behavior detection method, device, equipment and storage medium, aiming to improve the accuracy of non-motor vehicle theft behavior judgment and solve the technical problem of low accuracy of existing non-motor vehicle theft behavior judgment.

[0005] To achieve the above-mentioned purpose, the present application provides a non-motor vehicle theft behavior detection method, comprising the following steps:

[0006] Modeling the driving personnel in the snapshot image of the non-motor vehicle riding behavior and the non-motor vehicle respectively to obtain a driving personnel biological feature library and a non-motor vehicle feature library, wherein the driving personnel biological features in the driving personnel biological feature library correspond one by one to the non-motor vehicle features in the non-motor vehicle feature library;

[0007] Based on the driving personnel biological feature library and the non-motor vehicle feature library, the driving personnel biological features and the corresponding non-motor vehicle features belonging to the same driving personnel are archived to obtain the non-motor vehicle archives corresponding to each driving personnel;

[0008] Clustering the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives;

[0009] If the number of clustering clusters is greater than a preset threshold, it is determined that the driving personnel corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior.

[0010] Optionally, the step of clustering the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives comprises:

[0011] Calculating the similarity of the non-motor vehicle features in the non-motor vehicle archives to obtain an adjacency matrix;

[0012] Clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives.

[0013] Optionally, the step of clustering the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives comprises:

[0014] Taking the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives as the input of the first layer neural network of the GCN model to generate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archives;

[0015] Residual connecting the current layer neural network with the next layer neural network to extract part of the input features from the input features of the current layer neural network as residual features, wherein when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archives, and when the current layer neural network is not the first layer neural network of the GCN model, the input features include the residual features and the non-motor vehicle aggregated features;

[0016] Respectively taking the residual features, the non-motor vehicle aggregated features and the adjacency matrix as the input of the next layer neural network of the GCN model to regenerate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archives;

[0017] Returning to the step of residual connecting the current layer neural network with the next layer neural network to extract part of the input features from the input features of the current layer neural network as residual features until obtaining the number of clustering clusters of the non-motor vehicle archives output by the last layer neural network of the GCN model.

[0018] Optionally, the activation function in the GCN model is SeLU.

[0019] Optionally, the step of clustering the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives comprises:

[0020] Placing the non-motor vehicle features in the non-motor vehicle archives at each node of the adjacency matrix;

[0021] Random walk operation is performed on each node of the adjacency matrix until the average encoding length is the shortest, so as to obtain the cluster quantity of the non-motor vehicle archive.

[0022] Optionally, before the step of clustering the non-motor vehicle archive to obtain the cluster quantity of the non-motor vehicle archive, the method further comprises:

[0023] Based on the non-motor vehicle attribute model, the non-motor vehicle attributes in the snapshot image with the riding non-motor vehicle behavior are identified to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library.

[0024] The step of clustering the non-motor vehicle archive to obtain the cluster quantity of the non-motor vehicle archive comprises:

[0025] The non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes are clustered to obtain the cluster quantity of the non-motor vehicle archive.

[0026] Optionally, when the non-motor vehicle attribute is a non-motor vehicle orientation, the step of clustering the non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes to obtain the cluster quantity of the non-motor vehicle archive comprises:

[0027] Similarity calculation is performed on the non-motor vehicle features in the non-motor vehicle archive to obtain an adjacency matrix.

[0028] Based on the non-motor vehicle orientation, a clustering threshold value between the non-motor vehicles used in the step of clustering the non-motor vehicle archive is determined;

[0029] Based on the clustering threshold value, the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive, clustering is performed to obtain the cluster quantity of the non-motor vehicle archive.

[0030] In addition, to achieve the above-mentioned purposes, the application further provides a non-motor vehicle theft behavior detection device, comprising:

[0031] A modeling module is configured to model the driving personnel and the non-motor vehicle in the snapshot image with the riding non-motor vehicle behavior respectively to obtain a driving personnel biological feature library and a non-motor vehicle feature library, wherein the driving personnel biological features in the driving personnel biological feature library correspond to the non-motor vehicle features in the non-motor vehicle feature library one by one.

[0032] an archiving module configured to archive, based on the driver biometric library and the non-motor vehicle feature library, driver biometrics and corresponding non-motor vehicle features belonging to the same driver, to obtain respective non-motor vehicle archives of the drivers;

[0033] a clustering module configured to cluster the non-motor vehicle archives, to obtain a number of clustering clusters of the non-motor vehicle archives;

[0034] a judging module configured to determine that the driver corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior if the number of clustering clusters is greater than a preset threshold.

[0035] Optionally, the clustering module comprises:

[0036] a first computing unit configured to perform similarity calculation on non-motor vehicle features in the non-motor vehicle archives, to obtain an adjacency matrix; and a first clustering unit configured to perform clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives, to obtain the number of clustering clusters of the non-motor vehicle archives.

[0037] and / or the first clustering unit is specifically configured to take the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives as inputs of a first layer neural network of a GCN model, to generate non-motor vehicle aggregated features corresponding to the non-motor vehicle archives; to perform residual connection between a current layer neural network and a next layer neural network, to extract part of input features from input features of the current layer neural network as residual features, wherein when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archives, and when the current layer neural network is not the first layer neural network of the GCN model, the input features comprise the residual features and the non-motor vehicle aggregated features; to take the residual features, the non-motor vehicle aggregated features and the adjacency matrix as inputs of the next layer neural network of the GCN model, to regenerate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archives; and to return to perform the step of performing residual connection between the current layer neural network and the next layer neural network, to extract part of input features from input features of the current layer neural network as residual features, until the number of clustering clusters of the non-motor vehicle archives output by a last layer neural network of the GCN model is obtained.

[0038] and / or an activation function in the GCN model in the first clustering unit is SeLU.

[0039] And / or the first clustering unit is further specifically configured to place non-motor vehicle features in the non-motor vehicle archive at nodes of the adjacency matrix; and perform random walk operation on the nodes of the adjacency matrix until the average encoding length is the shortest, to obtain the number of clustering clusters of the non-motor vehicle archive.

[0040] And / or the non-motor vehicle theft behavior detection device further comprises an identification module configured to identify non-motor vehicle attributes in the snapshot image of the non-motor vehicle riding behavior based on the non-motor vehicle attribute model, to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library; and the clustering module is further configured to cluster the non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes, to obtain the number of clustering clusters of the non-motor vehicle archive.

[0041] And / or when the non-motor vehicle attribute is a non-motor vehicle orientation, the clustering module comprises: a first threshold determination module configured to determine a clustering threshold between non-motor vehicles used in the step of clustering the non-motor vehicle archive based on the non-motor vehicle orientation; and a second clustering module configured to cluster based on the clustering threshold, the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive, to obtain the number of clustering clusters of the non-motor vehicle archive.

[0042] In addition, to achieve the above-mentioned purpose, the present application also provides a non-motor vehicle theft behavior detection device, which comprises a memory, a processor and a non-motor vehicle theft behavior detection program stored in the memory and executable on the processor, wherein the non-motor vehicle theft behavior detection program is configured to implement the steps of the non-motor vehicle theft behavior detection method as described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application also provides a storage medium, wherein the storage medium stores a non-motor vehicle theft behavior detection program, and the non-motor vehicle theft behavior detection program is executed by a processor to implement the steps of the non-motor vehicle theft behavior detection method as described above.

[0044] The application provides a non-motor vehicle theft behavior detection method, device, equipment and storage medium, compared with the prior art, which judges whether there is a theft behavior based on one-to-one binding of a non-motor vehicle and a non-motor vehicle owner, misjudgment is easy to occur, and the non-motor vehicle theft behavior judgment accuracy is low. In the application, a driver in a snapshot image of a riding non-motor vehicle behavior and a non-motor vehicle are modeled respectively to obtain a driver biological feature library and a non-motor vehicle feature library, wherein the driver biological features in the driver biological feature library correspond one-to-one to the non-motor vehicle features in the non-motor vehicle feature library; based on the driver biological feature library and the non-motor vehicle feature library, driver biological features and corresponding non-motor vehicle features belonging to the same driver are archived to obtain respective non-motor vehicle archives of the drivers; the non-motor vehicle archives are clustered to obtain a cluster number of the non-motor vehicle archives; and if the cluster number is greater than a preset threshold, it is determined that the driver corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior. In the application, whether there is a theft behavior is no longer judged based on one-to-one binding of a non-motor vehicle and a non-motor vehicle owner, but the drivers are archived, the non-motor vehicles ridden by the drivers are clustered, and when the number of non-motor vehicles ridden by the driver within a certain time (the cluster number) exceeds the preset threshold, it is determined that the driver has a non-motor vehicle theft behavior, which avoids the non-motor vehicle from being judged as stolen when the non-motor vehicle is ridden by multiple people or the non-motor vehicle is ridden by multiple people in turn, and improves the non-motor vehicle theft behavior judgment accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 It is a flowchart of the first embodiment of the non-motor vehicle theft behavior detection method of the application.

[0046] Figure 2 It is a detailed flowchart of step S30 in the first embodiment of the non-motor vehicle theft behavior detection method of the application.

[0047] Figure 3 It is a first scene involved in the non-motor vehicle theft behavior detection method of the application.

[0048] Figure 4 It is a device structure diagram of a hardware running environment involved in the embodiment scheme of the application.

[0049] The implementation of the object of the application, functional features and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0050] It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0051] The embodiment of the present application provides a non-motor vehicle theft behavior detection method, as shown in the figure. Figure 1 In an embodiment of the non-motor vehicle theft behavior detection method of the present application, the method comprises the following steps:

[0052] Step S10, modeling the driver in the snapshot image of the riding non-motor vehicle behavior and the non-motor vehicle respectively, obtaining a driver biological feature library and a non-motor vehicle feature library, wherein the driver biological feature in the driver biological feature library corresponds to the non-motor vehicle feature in the non-motor vehicle feature library one by one.

[0053] Step S20, based on the driver biological feature library and the non-motor vehicle feature library, filing the driver biological feature and the corresponding non-motor vehicle feature belonging to the same driver, obtaining the non-motor vehicle archives corresponding to each driver;

[0054] Step S30, clustering the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives.

[0055] Step S40, if the number of clustering clusters is greater than a preset threshold, it is determined that the driver corresponding to the non-motor vehicle archives has non-motor vehicle theft behavior.

[0056] The embodiment aims to improve the accuracy of non-motor vehicle theft behavior judgment.

[0057] Specifically, in the present application, instead of judging whether there is theft behavior based on one-to-one binding of non-motor vehicles and non-motor vehicle owners, the driver is filed, the non-motor vehicles ridden by the driver are clustered, and when the number of non-motor vehicles ridden by the driver within a certain time (the number of clustering clusters) exceeds a preset threshold, it is determined that the driver has non-motor vehicle theft behavior, which avoids the judgment of theft when the non-motor vehicle is ridden by multiple people or the non-motor vehicle is ridden by multiple people, improves the accuracy of non-motor vehicle theft behavior judgment, and is more robust for one-car multi-riding (one family riding one car) and changing new cars.

[0058] Further, in the present application, the GCN model is used to cluster the non-motor vehicle archives to obtain the clustering cluster number of the non-motor vehicle archives, that is, to obtain the number of non-motor vehicles ridden by the driving personnel within a certain time corresponding to the non-motor vehicle archives, and then to judge whether the driving personnel has the non-motor vehicle theft behavior. Compared with the existing GCN model, the GCN model used in the present application connects the current layer neural network and the next layer neural network in a residual manner. On the one hand, for each layer neural network, the original non-motor vehicle features are retained to a certain extent while learning new non-motor vehicle features, which ensures the balance between the learned new non-motor vehicle features and the shallow non-motor vehicle features. On the other hand, the over-smoothing problem of the GCN model itself is alleviated, the depth of the GCN model is expanded, and then the accuracy of the output clustering cluster number is improved, and then the non-motor vehicle theft behavior judgment accuracy is improved. In the present application, the activation function is replaced from ReLU to SeLU to achieve better convergence of the GCN model.

[0059] Further, in the present application, the infomap algorithm (putting the non-motor vehicle features in the non-motor vehicle archives into each node of the adjacency matrix; performing random walk operation on each node of the adjacency matrix until the average encoding length is the shortest, to obtain the clustering cluster number of the non-motor vehicle archives) is used to cluster the non-motor vehicle archives to obtain the clustering cluster number of the non-motor vehicle archives. Compared with using the GCN model to cluster the non-motor vehicle archives, using the infomap algorithm to cluster the non-motor vehicle archives does not need to perform model training, reduces the consumption of network resources, and reduces the clustering time.

[0060] Further, in the present application, the non-motor vehicle attributes in the snapshot image of the non-motor vehicle riding behavior are identified to obtain a non-motor vehicle attribute library. The non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library. The non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes are clustered to obtain the clustering cluster number of the non-motor vehicle archives. That is, in the present application, the non-motor vehicle attribute features are clustered together with the non-motor vehicle features, which increases the features used for clustering, and then improves the clustering performance, and then improves the accuracy of the output clustering cluster number, and then improves the non-motor vehicle theft behavior judgment accuracy.

[0061] As an example, the non-motor vehicle attribute is a non-motor vehicle orientation, and the appearance of different orientations of the non-motor vehicle is very different, which will cause the same-orientation electric vehicles to be clustered into one category, but the different-orientation electric vehicles to be seriously split under the same clustering threshold. That is, in this application, the clustering threshold between the non-motor vehicles used in the step of determining the clustering of the non-motor vehicle archives is determined based on the non-motor vehicle orientation, so as to ensure the recall rate of the final clustering result, thereby improving the accuracy of the number of output clustering clusters, and further improving the accuracy of the judgment of the non-motor vehicle theft behavior.

[0062] The specific steps are as follows:

[0063] In step S10, the driver and the non-motor vehicle in the snapshot image in which the non-motor vehicle is ridden are modeled respectively to obtain a driver biometric feature library and a non-motor vehicle feature library, wherein the driver biometric features in the driver biometric feature library correspond one-to-one to the non-motor vehicle features in the non-motor vehicle feature library.

[0064] As an example, the snapshot image is an image captured by a daily monitoring device. For example, a monitoring camera on the road or a monitoring camera in the garage. The snapshot image in which the non-motor vehicle is ridden is part of the above snapshot images. The snapshot image is an image captured by a daily monitoring device. Even if the thief avoids the monitoring camera when stealing the non-motor vehicle, it is difficult to avoid the daily monitoring device during riding, thereby expanding the detection range of the non-motor vehicle theft behavior.

[0065] The process of obtaining the snapshot image in which the non-motor vehicle is ridden from all snapshot images captured by the daily monitoring device includes:

[0066] The position relationship between the human body and the non-motor vehicle is obtained by detecting the snapshot image through a detection algorithm.

[0067] Based on the position relationship between the human body and the non-motor vehicle, it is determined whether the non-motor vehicle is ridden.

[0068] If the non-motor vehicle is ridden, the snapshot image is determined to be the snapshot image in which the non-motor vehicle is ridden.

[0069] As an example, the detection algorithm can be a FAST feature point detection algorithm. The position information of the human body and the position information of the non-motor vehicle in the snapshot image are detected by the FAST feature point detection algorithm. If the distance between the position of the human body and the position of the non-motor vehicle in the snapshot image is less than a threshold value or both the position of the human body and the position of the non-motor vehicle in the snapshot image are within a preset area range, it is determined that the non-motor vehicle is ridden.

[0070] As an example, the driver biological features include face features and body features. Therefore, the driver and the non-motor vehicle in the snapshot image where the non-motor vehicle riding behavior exists are respectively modeled, and the driver biological feature library and the non-motor vehicle feature library can be specifically as follows:

[0071] The face and / or body in the snapshot image where the non-motor vehicle riding behavior exists are respectively modeled, and the face feature library, the body feature library and the non-motor vehicle feature library are obtained.

[0072] The face features in the face feature library correspond to the non-motor vehicle features one by one, the body features in the body feature library correspond to the non-motor vehicle features one by one, and the face features, the body features and the non-motor vehicle features corresponding to the same snapshot image where the non-motor vehicle riding behavior exists are corresponding.

[0073] In step S20, based on the driver biological feature library and the non-motor vehicle feature library, the driver biological features and the corresponding non-motor vehicle features belonging to the same driver are archived, and the non-motor vehicle archives corresponding to each driver are obtained.

[0074] As an example, the driver biological features and the corresponding non-motor vehicle features belonging to the same driver are archived based on a clustering algorithm.

[0075] The clustering algorithm can be DBSCAN, HAC, Infomap, spectral clustering, etc.

[0076] As an example, the driver biological feature library includes the face feature library and the body feature library. In this example, based on the driver biological feature library and the non-motor vehicle feature library, the driver biological features and the corresponding non-motor vehicle features belonging to the same driver are archived, and the non-motor vehicle archives corresponding to each driver are obtained. Specifically, it can be as follows:

[0077] The face features in the face feature library are clustered and archived based on a clustering algorithm, and the initial driver archives are obtained.

[0078] According to the face features in the initial driver archives, the body features corresponding to the face features are obtained from the body feature library and supplemented into the initial driver archives.

[0079] And based on the body features, the body features with a similarity greater than a similarity threshold are obtained from the body feature library and supplemented into the initial driver archives, and the driver archives are obtained.

[0080] Based on the facial and / or body features in the driver's file, non-motorized vehicle features corresponding to the facial and / or body features in the driver's file are retrieved from the non-motorized vehicle feature database to form a non-motorized vehicle file.

[0081] Step S30: Cluster the non-motorized vehicle files to obtain the number of clusters of the non-motorized vehicle files.

[0082] As an example, there are various clustering algorithms for clustering the non-motorized vehicle files. These algorithms can be traditional clustering algorithms, such as DBSCAN, HAC, Infomap, and spectral clustering; or they can be deep clustering algorithms, such as clustering models based on convolutional neural networks or graph convolutional networks.

[0083] Among them, such as Figure 2 As shown, the step of clustering the non-motorized vehicle files to obtain the number of clusters for the non-motorized vehicle files includes the following steps S31-S32:

[0084] Step S31: Calculate the similarity of non-motorized vehicle features in the non-motorized vehicle file to obtain the adjacency matrix;

[0085] As an example, the similarity calculation of non-motorized vehicle features in the non-motorized vehicle file to obtain the adjacency matrix can specifically be as follows:

[0086] Based on KNN graph construction, the similarity between pairwise non-motorized vehicle features in the non-motorized vehicle file is obtained, thereby obtaining the adjacency matrix.

[0087] As an example, when using the GCN model to cluster non-motorized vehicle records, post-processing is performed on the adjacency matrix after it is obtained to make the aggregated features more accurate. For example, the number of neighbors for each sample (non-motorized vehicle feature) can be limited, and only the top k neighbors can be aggregated. Another example is to filter the similarity of neighbors by a threshold, and aggregate the features only if the similarity between two samples (non-motorized vehicle features) is greater than the threshold.

[0088] Step S32: Based on the adjacency matrix and the non-motorized vehicle features in the non-motorized vehicle file, perform clustering to obtain the number of clusters in the non-motorized vehicle file.

[0089] As an example, the GCN model is used to cluster the non-motorized vehicle files based on the adjacency matrix and the non-motorized vehicle features in the non-motorized vehicle files to obtain the number of clusters in the non-motorized vehicle files. Figure 3 The diagram shows a flowchart of the GCN model clustering process in this application. The parts enclosed by rectangles in the diagram represent the residual connections and residual features between layers.

[0090] Figure 3 The figure shows the flowchart of the GCN model clustering process in the present application. The part in the rectangular frame in the figure is the residual connection between layers and the residual feature.

[0091] The specific process of clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive by using the GCN model to obtain the number of clustering clusters of the non-motor vehicle archive includes:

[0092] Step A1, taking the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive as the input of the first layer neural network of the GCN model, generating the non-motor vehicle aggregated features corresponding to the non-motor vehicle archive;

[0093] Step A2, performing residual connection between the current layer neural network and the next layer neural network, extracting part of the input features from the input features of the current layer neural network as residual features, wherein when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archive, and when the current layer neural network is not the first layer neural network of the GCN model, the input features include the residual features and the non-motor vehicle aggregated features;

[0094] Step A3, taking the residual features, the non-motor vehicle aggregated features and the adjacency matrix as the input of the next layer neural network of the GCN model respectively, and regenerating the non-motor vehicle aggregated features corresponding to the non-motor vehicle archive;

[0095] Step A4, return to step A2: performing residual connection between the current layer neural network and the next layer neural network, extracting part of the input features from the input features of the current layer neural network as residual features, until the number of clustering clusters of the non-motor vehicle archive output by the last layer neural network of the GCN model is obtained.

[0096] As an example, the activation function in the GCN model is SeLU.

[0097] As an example, the infomap algorithm is used to cluster based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive to obtain the number of clustering clusters of the non-motor vehicle archive.

[0098] The specific process of clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive by using the infomap algorithm to obtain the number of clustering clusters of the non-motor vehicle archive includes:

[0099] Step B1, placing non-motor vehicle features in the non-motor vehicle archive into each node of the adjacency matrix;

[0100] Step B2, performing random walk operation on each node of the adjacency matrix until the average encoding length is the shortest, obtaining the number of clustering clusters of the non-motor vehicle archive.

[0101] In the present example, the average encoding length can be expressed by the formula:

[0102]

[0103] wherein N represents the real number of clustering clusters, Y represents the preset number of clustering clusters, a represents a node in the clustering cluster i, p α represents the probability of random walk visiting the a node, represents the probability of random walk continuing to walk in the clustering cluster i, represents the probability of random walk jumping out of the clustering cluster i to other clusters, represents the probability of random walk across different clustering clusters, and P represents the transition probability matrix.

[0104] Step S40, if the number of clustering clusters is greater than the preset threshold, it is determined that the non-motor vehicle archive corresponds to a driver who has non-motor vehicle theft behavior.

[0105] It should be noted that the preset threshold can be set according to actual conditions, which is not limited in the present embodiment.

[0106] As an example, before comparing the number of clustering clusters with the preset threshold, the non-motor vehicles in each clustering cluster are identified by a deep learning algorithm, and if the identification result is a shared non-motor vehicle, the clustering cluster is deleted, and the number of clustering clusters obtained in step 30 is reduced by 1.

[0107] The application provides a non-motor vehicle theft behavior detection method, device, equipment and storage medium. Compared with the prior art, the non-motor vehicle and the owner of the non-motor vehicle are one-to-one bound to determine whether there is a theft behavior, which is easy to misjudge, resulting in low accuracy of non-motor vehicle theft behavior judgment. In the application, the driver in the snapshot image of the non-motor vehicle riding behavior and the non-motor vehicle are modeled respectively to obtain a driver biological feature library and a non-motor vehicle feature library, wherein the driver biological feature in the driver biological feature library corresponds to the non-motor vehicle feature in the non-motor vehicle feature library one by one. Based on the driver biological feature library and the non-motor vehicle feature library, the driver biological features and the corresponding non-motor vehicle features belonging to the same driver are archived to obtain the non-motor vehicle archives corresponding to the drivers respectively. The non-motor vehicle archives are clustered to obtain the cluster number of the non-motor vehicle archives. If the cluster number is greater than a preset threshold, it is determined that the driver corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior. In the application, instead of determining whether there is a theft behavior based on the one-to-one binding of the non-motor vehicle and the owner of the non-motor vehicle, the driver is archived, and the non-motor vehicle ridden by the driver is clustered. When the number of non-motor vehicles ridden by the driver within a certain time (cluster number) exceeds the preset threshold, it is determined that the driver has a non-motor vehicle theft behavior, which avoids the non-motor vehicle being judged as stolen when it is ridden by multiple people or when it is ridden by multiple people, and improves the accuracy of non-motor vehicle theft behavior judgment.

[0108] Further, based on the first embodiment of the application, another embodiment of the application is provided, in which the step of clustering the non-motor vehicle archives to obtain the cluster number of the non-motor vehicle archives further includes:

[0109] Step 50, identifying the non-motor vehicle attributes in the snapshot image of the non-motor vehicle riding behavior based on the non-motor vehicle attribute model to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library.

[0110] As an example, the non-motor vehicle attributes include non-motor vehicle orientation, non-motor vehicle color, and non-motor vehicle style (pedal electric vehicle, pedal electric vehicle, etc.).

[0111] As an example, the non-motor vehicle attribute model can be obtained by training Resnet, HRnet and other networks.

[0112] In this embodiment, step S30, clustering the non-motor vehicle archives to obtain the cluster number of the non-motor vehicle archives, includes:

[0113] clustering the non-motor vehicle features in the non-motor vehicle feature library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle profile.

[0114] As an example, when the non-motor vehicle attribute is non-motor vehicle color, the step of clustering the non-motor vehicle features in the non-motor vehicle feature library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle profile comprises:

[0115] dividing the non-motor vehicle features in the non-motor vehicle feature library based on the non-motor vehicle color to obtain a plurality of groups of non-motor vehicle features;

[0116] performing the following steps on each group of non-motor vehicle features: performing similarity calculation on the non-motor vehicle features to obtain an adjacency matrix; and clustering based on the adjacency matrix and the non-motor vehicle features to obtain the number of clustering clusters;

[0117] adding the number of clustering clusters corresponding to each group of non-motor vehicle features to obtain the number of clustering clusters of the non-motor vehicle profile.

[0118] As an example, when the non-motor vehicle attribute is non-motor vehicle style (scooter, pedal car, etc.), the step of clustering the non-motor vehicle features in the non-motor vehicle feature library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle profile comprises:

[0119] adding the non-motor vehicle attribute to the non-motor vehicle features to obtain integrated non-motor vehicle features;

[0120] performing similarity calculation on the integrated non-motor vehicle features to obtain an adjacency matrix;

[0121] clustering based on the adjacency matrix and the integrated non-motor vehicle features to obtain the number of clustering clusters of the non-motor vehicle profile.

[0122] As an example, when the non-motor vehicle attribute is non-motor vehicle orientation, the step of clustering the non-motor vehicle features in the non-motor vehicle feature library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle profile comprises:

[0123] performing similarity calculation on the non-motor vehicle features in the non-motor vehicle profile to obtain an adjacency matrix;

[0124] determining a clustering threshold between non-motor vehicles used in the step of clustering the non-motor vehicle profile based on the non-motor vehicle orientation;

[0125] Clustering is performed based on the clustering threshold, the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive to obtain the number of clustering clusters of the non-motor vehicle archive.

[0126] For example, there are non-motor vehicles A, B and C, wherein non-motor vehicles A and B are forward-facing, and non-motor vehicle C is backward-facing. When clustering, the clustering threshold between non-motor vehicles A and B is set to 0.8, and the clustering threshold between non-motor vehicles A and C and between non-motor vehicles B and C is set to 0.6. That is, the smaller the orientation difference between non-motor vehicles, the larger the clustering threshold used by the non-motor vehicles when clustering.

[0127] In this embodiment, the non-motor vehicle attributes in the snapshot image of the riding non-motor vehicle behavior are recognized to obtain a non-motor vehicle attribute library. The non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library. The non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes are clustered to obtain the number of clustering clusters of the non-motor vehicle archive. That is, in this embodiment, the non-motor vehicle attribute features are clustered together with the non-motor vehicle features, which increases the features used for clustering, thereby improving the clustering performance, thereby improving the accuracy of the output clustering cluster number, and thereby improving the accuracy of the non-motor vehicle theft behavior judgment.

[0128] Referring to Figure 4 , Figure 4 is a device structure schematic diagram of a hardware running environment involved in the embodiment scheme of the present application.

[0129] As Figure 4 shown, the non-motor vehicle theft behavior detection device can include a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 can include a display screen (Display) and an input unit such as a keyboard (Keyboard). The optional user interface 1003 can also include a standard wired interface and a wireless interface. The network interface 1004 can optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WIreless-FIdelity, WI-FI) interface). The memory 1005 can be a high-speed random access memory (RAM) memory, or a stable non-volatile memory (NVM), such as a magnetic disk memory. The memory 1005 can also be an independent storage device from the aforementioned processor 1001.

[0130] Those skilled in the art can understand that Figure 4 The structure shown in the figure does not constitute a limitation on the non-motor vehicle theft behavior detection device, and can include more or fewer components than the figure, or combine certain components, or different component arrangements.

[0131] As Figure 4 As shown, the memory 1005 as a storage medium can include an operating system, a data storage module, a network communication module, a user interface module, and a non-motor vehicle theft behavior detection program.

[0132] In Figure 4 The network interface 1004 in the non-motor vehicle theft behavior detection device is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the non-motor vehicle theft behavior detection device of the present application can be arranged in the non-motor vehicle theft behavior detection device, and the non-motor vehicle theft behavior detection device calls the non-motor vehicle theft behavior detection program stored in the memory 1005 through the processor 1001 to realize the steps of the graph neural network optimization method described in any one of the above embodiments.

[0133] The specific implementation of the non-motor vehicle theft behavior detection device of the present application is basically the same as that of the above-mentioned non-motor vehicle theft behavior detection method, and will not be repeated here.

[0134] The present application also provides a non-motor vehicle theft behavior detection device, comprising:

[0135] A modeling module is configured to model a driver and a non-motor vehicle in a snapshot image in which the non-motor vehicle is ridden, to obtain a driver biometric feature library and a non-motor vehicle feature library, wherein the driver biometric features in the driver biometric feature library correspond one-to-one to the non-motor vehicle features in the non-motor vehicle feature library;

[0136] An archiving module is configured to archive driver biometric features and corresponding non-motor vehicle features belonging to the same driver based on the driver biometric feature library and the non-motor vehicle feature library, to obtain a non-motor vehicle archive corresponding to each driver;

[0137] A clustering module is configured to cluster the non-motor vehicle archives, to obtain a number of clusters of the non-motor vehicle archives;

[0138] A judgment module is configured to determine that a driver corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior if the number of clusters is greater than a preset threshold.

[0139] In a possible implementation of the present application, the clustering module comprises:

[0140] a first computing unit, configured to perform similarity calculation on the non-motor vehicle features in the non-motor vehicle archive to obtain an adjacency matrix;

[0141] and / or the first clustering unit is specifically configured to take the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive as inputs of a first layer neural network of a GCN model to generate non-motor vehicle aggregated features corresponding to the non-motor vehicle archive; perform residual connection between a current layer neural network and a next layer neural network to extract part of input features from input features of the current layer neural network as residual features, wherein when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archive, and when the current layer neural network is not the first layer neural network of the GCN model, the input features include the residual features and the non-motor vehicle aggregated features; take the residual features, the non-motor vehicle aggregated features and the adjacency matrix as inputs of the next layer neural network of the GCN model to regenerate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archive; and return to perform the step of performing residual connection between the current layer neural network and the next layer neural network to extract part of input features from input features of the current layer neural network as residual features until the clustering cluster number of the non-motor vehicle archive output by a last layer neural network of the GCN model is obtained;

[0142] and / or the activation function in the GCN model in the first clustering unit is SeLU;

[0143] and / or the first clustering unit is further specifically configured to place the non-motor vehicle features in the non-motor vehicle archive at nodes of the adjacency matrix; and perform random walk operation on the nodes of the adjacency matrix until the average encoding length is the shortest to obtain the clustering cluster number of the non-motor vehicle archive;

[0144] and / or the non-motor vehicle theft behavior detection device further comprises an identification module configured to identify non-motor vehicle attributes in a snapshot image with a riding non-motor vehicle behavior based on a non-motor vehicle attribute model to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library; and the clustering module is further configured to cluster the non-motor vehicle features in the non-motor vehicle feature library and the associated non-motor vehicle attributes to obtain the clustering cluster number of the non-motor vehicle archive;

[0145] And / or when the non-motor vehicle attribute is a non-motor vehicle orientation, the clustering module comprises: a first threshold determination module configured to determine a clustering threshold between non-motor vehicles used in the clustering of the non-motor vehicle archives based on the non-motor vehicle orientation; and a second clustering module configured to cluster based on the clustering threshold, the adjacency matrix, and the non-motor vehicle features in the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives.

[0146] The non-motor vehicle theft detection device embodiment of the present application is basically the same as the above-mentioned non-motor vehicle theft detection method embodiments, and will not be repeated here.

[0147] The embodiment of the present application provides a storage medium, and the storage medium stores one or more programs, and the one or more programs can also be executed by one or more processors to implement the steps of the non-motor vehicle theft detection method described in any of the above.

[0148] The storage medium embodiment of the present application is basically the same as the above-mentioned non-motor vehicle theft detection method embodiments, and will not be repeated here.

[0149] The present application also provides a computer program product, including a computer program, which is executed by a processor to implement the steps of the above-mentioned non-motor vehicle theft detection method.

[0150] The computer program product embodiment of the present application is basically the same as the above-mentioned non-motor vehicle theft detection method embodiments, and will not be repeated here.

[0151] It should be noted that in this paper, the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the sentence "including a…" does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0152] The above-mentioned embodiment number of the present application is only for description, not representing the pros and cons of the embodiments.

[0153] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, can also be through hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art contribution can be embodied in the form of software products, the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc) as described above, including a number of instructions to make a terminal device (may be a mobile phone, computer, server, or network equipment, etc.) executes the method described in various embodiments of the present application.

[0154] The above is only the preferred embodiment of the present application, not therefore limit the patent scope of the present application, any equivalent structure or equivalent process transformation using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A non-motor vehicle theft behavior detection method, characterized in that, The non-motor vehicle theft behavior detection method comprises the following steps: The driver in the snapshot image of the riding non-motor vehicle behavior and the non-motor vehicle are modeled respectively to obtain a driver biological feature library and a non-motor vehicle feature library, wherein the driver biological features in the driver biological feature library correspond one-to-one to the non-motor vehicle features in the non-motor vehicle feature library; Based on the driver biological feature library and the non-motor vehicle feature library, the driver biological features and the corresponding non-motor vehicle features belonging to the same driver are archived to obtain the non-motor vehicle archives corresponding to the respective drivers; The non-motor vehicle archives are clustered to obtain the cluster number of the non-motor vehicle archives; If the cluster number is greater than a preset threshold, it is determined that the driver corresponding to the non-motor vehicle archives has non-motor vehicle theft behavior; The step of clustering the non-motor vehicle archives to obtain the cluster number of the non-motor vehicle archives comprises: Similarity calculation is performed on the non-motor vehicle features in the non-motor vehicle archives to obtain an adjacency matrix; Based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives, clustering is performed to obtain the cluster number of the non-motor vehicle archives; The step of clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives to obtain the cluster number of the non-motor vehicle archives comprises: The adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives are taken as the input of the first layer neural network of a GCN model to generate non-motor vehicle aggregated features corresponding to the non-motor vehicle archives; The current layer neural network is connected in residual connection with the next layer neural network, and part of the input features of the current layer neural network is extracted as residual features, wherein when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archives, and when the current layer neural network is not the first layer neural network of the GCN model, the input features include the residual features and the non-motor vehicle aggregated features; The residual features, the non-motor vehicle aggregated features and the adjacency matrix are taken as the input of the next layer neural network of the GCN model respectively to regenerate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archives; The step of connecting the current layer neural network in residual connection with the next layer neural network and extracting part of the input features of the current layer neural network as residual features is returned to be executed until the cluster number of the non-motor vehicle archives output by the last layer neural network of the GCN model is obtained.

2. The non-motor vehicle theft detection method of claim 1, wherein, The activation function in the GCN model is SeLU.

3. The non-motor vehicle theft detection method of claim 1, wherein, The step of clustering based on the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archives to obtain the cluster number of the non-motor vehicle archives comprises: The non-motor vehicle features in the non-motor vehicle archives are placed at each node of the adjacency matrix; Random walk operation is performed on each node of the adjacency matrix until the average encoding length is the shortest to obtain the cluster number of the non-motor vehicle archives.

4. The non-motor vehicle theft detection method of claim 1, wherein, The step of clustering the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives further comprises: identifying non-motor vehicle attributes in the snapshot images of the riding non-motor vehicle behavior based on the non-motor vehicle attribute model to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle attribute library; The step of clustering the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives comprises: clustering the non-motor vehicle features in the non-motor vehicle attribute library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle archives.

5. The non-motor vehicle theft detection method of claim 4, wherein, When the non-motor vehicle attribute is the non-motor vehicle orientation, the step of clustering the non-motor vehicle features in the non-motor vehicle attribute library and their associated non-motor vehicle attributes to obtain the number of clustering clusters of the non-motor vehicle archives comprises: calculating the similarity of the non-motor vehicle features in the non-motor vehicle archives to obtain an adjacency matrix; determining a clustering threshold between the non-motor vehicles used in the step of clustering the non-motor vehicle archives based on the non-motor vehicle orientation; clustering the non-motor vehicle features in the non-motor vehicle archives based on the clustering threshold, the adjacency matrix, and the non-motor vehicle features to obtain the number of clustering clusters of the non-motor vehicle archives.

6. A non-motor vehicle theft detection apparatus, characterized by, The non-motor vehicle theft behavior detection device comprises: a modeling module configured to model the drivers and non-motor vehicles in the snapshot images of the riding non-motor vehicle behavior to obtain a driver biometric feature library and a non-motor vehicle feature library, wherein the driver biometric features in the driver biometric feature library correspond one-to-one to the non-motor vehicle features in the non-motor vehicle feature library; an archiving module configured to archive the driver biometric features and their corresponding non-motor vehicle features belonging to the same driver based on the driver biometric feature library and the non-motor vehicle feature library to obtain non-motor vehicle archives corresponding to each driver; a clustering module configured to cluster the non-motor vehicle archives to obtain the number of clustering clusters of the non-motor vehicle archives; a judgment module configured to determine that the driver corresponding to the non-motor vehicle archives has a non-motor vehicle theft behavior if the number of clustering clusters is greater than a preset threshold; The clustering module comprises: a first calculation unit configured to calculate the similarity of the non-motor vehicle features in the non-motor vehicle archives to obtain an adjacency matrix; and a first clustering unit configured to cluster the non-motor vehicle features in the non-motor vehicle archives based on the adjacency matrix and the non-motor vehicle features to obtain the number of clustering clusters of the non-motor vehicle archives. and the first clustering unit is specifically configured to take the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive as inputs of a first layer neural network of the GCN model, to generate non-motor vehicle aggregated features corresponding to the non-motor vehicle archive; to perform residual connection between a current layer neural network and a next layer neural network, to extract part of input features from the input features of the current layer neural network as residual features, wherein, when the current layer neural network is the first layer neural network of the GCN model, the input features are the non-motor vehicle features in the non-motor vehicle archive, and when the current layer neural network is not the first layer neural network of the GCN model, the input features include the residual features and the non-motor vehicle aggregated features; to take the residual features, the non-motor vehicle aggregated features and the adjacency matrix as inputs of the next layer neural network of the GCN model respectively, to regenerate the non-motor vehicle aggregated features corresponding to the non-motor vehicle archive; and to return to perform the step of performing residual connection between a current layer neural network and a next layer neural network, to extract part of input features from the input features of the current layer neural network as residual features, until the number of clustering clusters of the non-motor vehicle archive output by the last layer neural network of the GCN model is obtained.

7. The non-motor vehicle theft behavior detection device according to claim 6, wherein and / or the activation function in the GCN model in the first clustering unit is SeLU; and / or the first clustering unit is further specifically configured to place the non-motor vehicle features in the non-motor vehicle archive at each node of the adjacency matrix; and to perform random walk operation on each node of the adjacency matrix until the average encoding length is the shortest, to obtain the number of clustering clusters of the non-motor vehicle archive; and / or the non-motor vehicle theft behavior detection device further comprises an identification module configured to identify non-motor vehicle attributes in a snapshot image of a riding non-motor vehicle behavior based on a non-motor vehicle attribute model, to obtain a non-motor vehicle attribute library, wherein the non-motor vehicle attributes in the non-motor vehicle attribute library are associated with the non-motor vehicle features in the non-motor vehicle feature library; and the clustering module is further configured to cluster the non-motor vehicle features in the non-motor vehicle attribute library and the associated non-motor vehicle attributes, to obtain the number of clustering clusters of the non-motor vehicle archive; and / or when the non-motor vehicle attribute is a non-motor vehicle orientation, the clustering module comprises: a first threshold determination module configured to determine a clustering threshold between non-motor vehicles used in the step of clustering the non-motor vehicle archive based on the non-motor vehicle orientation; and a second clustering module configured to cluster based on the clustering threshold, the adjacency matrix and the non-motor vehicle features in the non-motor vehicle archive, to obtain the number of clustering clusters of the non-motor vehicle archive.

8. A non-motor vehicle theft behavior detection apparatus, characterized by, The device comprises a memory, a processor and a non-motor vehicle theft behavior detection program stored on the memory and executable on the processor, and the non-motor vehicle theft behavior detection program is configured to implement the steps of the non-motor vehicle theft behavior detection method according to any one of claims 1 to 5.

9. A storage medium, characterized by The storage medium has stored thereon a non-motor vehicle theft behavior detection program, which, when executed by the processor, implements the steps of the non-motor vehicle theft behavior detection method according to any one of claims 1 to 5.

Images