Anti-theft monitoring and early warning method and device for unmanned store

Abstract

The invention discloses an anti-theft monitoring and early warning method and device for an unmanned store, and relates to the technical field of anti-theft early warning. The method comprises the steps of performing information collection on a target customer, and establishing a target customer label based on an information collection result; performing video acquisition on the shopping behavior to obtain a target customer behavior sequence; calling a target consumption behavior analysis model, and performing suspicion degree judgment on the target customer behavior sequence to obtain a target customer suspicion degree; when the suspicion degree exceeds a suspicion degree threshold value, a target customer label is sent to a monitoring control terminal, and when the target customer leaves the unmanned selling shop, a commodity taking list and a checkout commodity list are acquired; and judging whether the commodity taking list is matched with the checkout commodity list or not, if not, activating an early warning device of the unmanned store to give an alarm, and starting an anti-theft plan. The technical problem that in the prior art, an unmanned store is difficult to accurately recognize the theft behavior is solved, and the technical effect of improving the theft behavior recognition accuracy is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-theft warning, and particularly relates to an anti-theft monitoring and warning method and device for an unmanned vending store. Background Art

[0002] With the continuous progress of technology and the change of consumption habits, the unmanned vending store, as a new retail model, has gradually emerged. The unmanned vending store is welcomed by more and more consumers for its convenience and high efficiency. The unmanned vending store usually adopts automated equipment and technologies, without manual operation. Customers can purchase goods by scanning codes to achieve self-service shopping and checkout. However, the unmanned vending store also faces some challenges in terms of safety and management. Due to the lack of personnel supervision and the safety facilities of traditional stores, these stores are prone to becoming targets of theft activities. Traditional unmanned vending stores are difficult to accurately identify theft behaviors, unable to give early warnings in time, and prone to causing losses. Summary of the Invention

[0003] The embodiments of the present application provide an anti-theft monitoring and warning method and device for an unmanned vending store, which solve the technical problem that it is difficult for existing unmanned vending stores to accurately identify theft behaviors.

[0004] In view of the above problems, the embodiments of the present application provide an anti-theft monitoring and warning method and device for an unmanned vending store.

[0005] In the first aspect of the embodiments of the present application, an anti-theft monitoring and warning method for an unmanned vending store is provided. The method includes: Deploy information collection devices in the unmanned vending store. When a target customer enters the unmanned vending store, collect information about the target customer, and establish a target customer label based on the information collection result; Perform real-time video collection on the shopping behavior of the target customer through the distributed camera network of the unmanned vending store, map and associate the video collection result with the target customer label, and extract action features from the video collection result to obtain a target customer behavior sequence; Retrieve a target consumption behavior analysis model from a preset model library based on the target customer label, judge the suspicious degree of the target customer behavior sequence, and obtain the target customer suspicious degree; When the target customer suspicious degree exceeds the suspicious degree threshold, send the target customer label to the monitoring and control terminal. When the target customer leaves the unmanned vending store, the monitoring and control terminal obtains a list of taken goods and a list of checked-out goods based on the video collection result; Judge whether the list of taken goods matches the list of checked-out goods. When they do not match, activate the warning device of the unmanned vending store to give an alarm and start an anti-theft plan.

[0006] In the second aspect of the embodiments of the present application, an anti-theft monitoring and warning device for an unmanned vending store is provided. The device includes: An information collection module, which is used to deploy information collection devices in the unmanned vending store. When a target customer enters the unmanned vending store, the information collection module collects information about the target customer and establishes a target customer label based on the information collection result. A behavior sequence acquisition module, which is used to perform real-time video collection on the shopping behavior of the target customer through the distributed camera network of the unmanned vending store, map and associate the video collection result with the target customer label, and extract action features from the video collection result to obtain the target customer behavior sequence. A suspiciousness judgment module, which is used to retrieve a target consumption behavior analysis model from a preset model library based on the target customer label, judge the suspiciousness of the target customer behavior sequence, and obtain the target customer suspiciousness. A monitoring module, which is used to send the target customer label to the monitoring control terminal when the target customer suspiciousness exceeds the suspiciousness threshold. When the target customer leaves the unmanned vending store, the monitoring control terminal obtains a list of items taken and a list of items checked out based on the video collection result. An early warning module, which is used to judge whether the list of items taken matches the list of items checked out. When they do not match, the early warning device of the unmanned vending store is activated to give an alarm and start an anti-theft plan.

[0007] One or more technical solutions provided in the present application have at least the following technical effects or advantages: Information collection devices are deployed in the unmanned vending store. When a target customer enters the unmanned vending store, information about the target customer is collected, and a target customer label is established based on the information collection result. Then, real-time video collection is performed on the shopping behavior of the target customer through the distributed camera network in the unmanned vending store. The video collection result will be associated with the target customer label, and action features will be extracted to obtain the target customer behavior sequence. Then, a target consumption behavior analysis model is retrieved from a preset model library based on the target customer label, the suspiciousness of the target customer behavior sequence is judged, and the target customer suspiciousness is obtained. When the target customer suspiciousness exceeds the suspiciousness threshold, the target customer label is sent to the monitoring control terminal. When the target customer leaves the unmanned vending store, the monitoring control terminal obtains a list of items taken and a list of items checked out based on the video collection result. Judge whether the list of items taken matches the list of items checked out. When they do not match, the early warning device of the unmanned vending store is activated to give an alarm and start an anti-theft plan. This solves the technical problem in the prior art that it is difficult for unmanned vending stores to accurately identify theft behaviors, and achieves the technical effect of improving the accuracy of theft behavior identification. Description of the Drawings

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0009] Figure 1 It is a schematic flow chart of a theft prevention monitoring and early warning method for an unmanned vending store provided by an embodiment of the present application; Figure 2 It is a schematic structural diagram of a theft prevention monitoring and early warning device for an unmanned vending store provided by an embodiment of the present application.

[0010] Explanation of reference numerals: Information acquisition module 11, behavior sequence acquisition module 12, suspiciousness judgment module 13, monitoring module 14, early warning module 15. Specific embodiments

[0011] By providing a theft prevention monitoring and early warning method and device for an unmanned vending store in the embodiments of the present application, the technical problem in the prior art that it is difficult for an unmanned vending store to accurately identify theft behaviors is solved.

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope protected by the present application.

[0013] It should be noted that the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0014] Embodiment 1 As Figure 1 shown, the embodiments of the present application provide a theft prevention monitoring and early warning method for an unmanned vending store. Among them, the method includes: Deploy information acquisition devices in the unmanned vending store. When a target customer enters the unmanned vending store, collect information about the target customer, and establish a target customer label based on the information acquisition result; An information collection device is arranged in the unmanned vending store. When a target customer enters the unmanned vending store, the information collection device will collect information about the target customer. The information collection device is used to capture and record information such as the images and behavior data of customers. Based on the information collection results, a label of the target customer is established for identification and management.

[0015] Furthermore, establishing the label of the target customer includes: Presetting information collection dimensions, forming an information collection device based on the information collection dimensions, and distributing it in the unmanned vending store in different areas. Among them, the information collection dimensions at least include facial features, age, gender, and height; Assigning a unique tracking identifier to each customer entering the unmanned vending store, and real-time monitoring the tracking identifiers of each customer in the unmanned vending store; If it is monitored that there is a customer without an assigned tracking identifier, determine it as the target customer, trigger the information collection device corresponding to the position information of the target customer, and collect information about the target customer to obtain the information collection results; Based on the information collection results, construct a multi-dimensional feature vector of the target customer, use the multi-dimensional feature vector as the label of the target customer, and assign a tracking identifier to the target customer.

[0016] The information collection dimensions include facial features, age, gender, and height. An information collection device is formed according to the information collection dimensions. The information collection device includes a facial recognition camera, a height measurement device, etc. The information collection device is distributed in different areas of the unmanned vending store. For example, a facial recognition camera is installed in the entrance area to capture the information of customers entering the store; cameras can also be installed in the shelf area and the checkout area to monitor the shopping behavior and settlement process of customers in real time. When a customer enters the unmanned vending store, a unique tracking identifier needs to be assigned to each customer to monitor the activity track of the customer in real time throughout the store. When it is monitored that there is a customer without an assigned tracking identifier, that is, the target customer, the information collection device corresponding to the position information of this customer will be immediately triggered to obtain the information collection results. Based on the information collection results, construct a multi-dimensional feature vector of the target customer, including information such as facial features, age, gender, and height, to form a comprehensive description of the target customer. Use the multi-dimensional feature vector as the label of the target customer and assign a unique tracking identifier for identification and tracking.

[0017] Through the distributed camera network of the unmanned vending store, real-time video collection of the shopping behavior of the target customer is carried out, the video collection results are mapped and associated with the target customer label, and action feature extraction is performed on the video collection results to obtain the target customer behavior sequence; The distributed camera network in the unmanned store covers key areas, such as the shelf area, checkout area, and main aisles in the store, ensuring that the location and number of cameras can capture the shopping behavior of the target customer. When the target customer enters the store, the customer is identified and locked through the tracking mark. Subsequently, the distributed camera network begins to collect real-time video of the customer's shopping behavior. The camera automatically adjusts the shooting angle and focal length according to the customer's movement trajectory to ensure that the customer's behavior can always be clearly captured. While the video is being collected, the video collection results are mapped and associated with the target customer label, that is, each video is associated with the corresponding customer label. After completing the video collection and mapping association, the video collection results are subjected to motion feature extraction. Computer vision and deep learning technologies, such as posture estimation and motion recognition, can be used to extract the target customer's motion information from the video, thereby obtaining the target customer behavior sequence. The target customer behavior sequence includes information such as the target customer's movement path, residence time, and shopping actions in the store.

[0018] Based on the target customer tag, a target consumption behavior analysis model is retrieved from a preset model library, and a suspiciousness judgment is made on the target customer behavior sequence to obtain the target customer suspicion; The preset model library includes multiple standard consumer behavior analysis models. According to the multi-dimensional feature information in the target customer label, the matching target consumer behavior analysis model is quickly located in the preset model library, and the target customer behavior sequence is input into the target consumer behavior analysis model for analysis to obtain the target customer suspicion. The target customer suspicion indicates the degree of deviation between the target customer behavior and normal consumer behavior, and is used to determine whether the target customer has abnormal behavior.

[0019] Further, the method comprises: Collect historical customer shopping videos, and obtain a sample customer set and a sample customer behavior sequence set based on the historical customer shopping videos, wherein each sample customer has a corresponding sample customer label; Marking the suspiciousness of the sample customer behavior sequences in the sample customer behavior sequence set to obtain a sample customer suspiciousness set; Performing cluster analysis on the sample customer labels to obtain multiple label clusters, extracting the centers of the multiple label clusters to generate multiple model identification labels; Dividing the sample customer behavior sequence set and the sample customer suspicion degree set based on the multiple label clusters to obtain multiple training sample sets; The multiple training sample sets are respectively subjected to supervised training to obtain multiple standard consumer behavior analysis models, and the multiple standard consumer behavior analysis models are mapped one by one with multiple model identification labels to form a preset model library.

[0020] Collect historical customer shopping videos, which record the shopping processes and behaviors of customers. From the historical customer shopping videos, extract representative customer samples to form a sample customer set. At the same time, through video analysis technology, extract the behavior sequences of each sample customer during the shopping process to form a sample customer behavior sequence set. Each sample customer has a corresponding sample customer label, including characteristics such as age, gender, and shopping habits. Based on the store's operation experience and professional knowledge, identify the suspiciousness of each behavior sequence in the sample customer behavior sequence set to obtain a sample customer suspiciousness set. Use the K-means algorithm to perform clustering analysis on the sample customer labels, and group similar labels into a clustering cluster. For example, group the sample customer labels of similar age ranges into the same label clustering cluster. Extract the center from each label clustering cluster to generate multiple model identification labels, which are used to represent different types of customer behavior patterns. According to the multiple model identification labels, divide the sample customer behavior sequence set and the sample customer suspiciousness set to obtain multiple training sample sets, ensuring that the customer behavior sequences and suspiciousness identifications in each training sample set match the corresponding label clustering clusters. Perform supervised learning on each training sample set to train multiple standard consumer behavior analysis models, with each model corresponding to a model identification label. Map the multiple model identification labels to the corresponding standard consumer behavior analysis models one by one to form a preset model library, which is used to analyze the behavior of new customers and evaluate their suspiciousness.

[0021] Furthermore, retrieving a target consumer behavior analysis model from the preset model library based on the target customer label includes: Obtain multiple model identification labels in the preset model library, calculate the label similarity with the target customer label, and obtain the maximum label similarity; Determine the target model identification label based on the maximum label similarity, and extract the corresponding standard consumer behavior analysis model from the preset model library according to the target model identification label as the target consumer behavior analysis model.

[0022] Obtain multiple model identification labels from the preset model library, calculate the label similarity between the target customer label and each model identification label. By comparing the similarities of the target customer label and the model identification labels in these features, a quantitative index can be obtained to represent the similarity between the labels. After obtaining all the label similarities, find the maximum label similarity, that is, the label with the highest similarity to the target customer label. Determine the target model identification label based on the maximum label similarity, that is, the model identification label with the highest similarity to the target customer label. According to the determined target model identification label, find the corresponding standard consumer behavior analysis model in the preset model library, and this model will be used as the target consumer behavior analysis model.

[0023] Furthermore, the method further includes: Extract a standard consumption behavior analysis model corresponding to the target customer label from a preset model library; Based on the standard consumption behavior analysis model, copy and generate a target consumption behavior analysis model for the target customer; Destroy the target consumption behavior analysis model when the target customer leaves the unmanned store.

[0024] According to the target customer label, extract a standard consumption behavior analysis model corresponding to the target customer label from a preset model library, and use the extracted standard consumption behavior analysis model as a basis to copy and generate a target consumption behavior analysis model for the target customer. This target model will be used to analyze and evaluate the consumption behavior of the target customer. When the target customer leaves the unmanned store, the target consumption behavior analysis model will be destroyed in a timely manner to ensure that personal data and behavior information of the target customer are no longer held and used. In this way, both accurate analysis and monitoring of the shopping behavior of the target customer can be achieved, and the privacy and data security of the customer can be ensured.

[0025] When the suspiciousness of the target customer exceeds the suspiciousness threshold, send the target customer label to a monitoring and control terminal. When the target customer leaves the unmanned store, the monitoring and control terminal obtains a list of items taken and a list of items checked out based on the video capture results; When it is determined through the target consumption behavior analysis model that the suspiciousness of the target customer exceeds the set suspiciousness threshold, the target customer label will be immediately sent to the monitoring and control terminal, which is used to handle abnormal situations. By sending the target customer label, the monitoring and control terminal can quickly obtain the video capture results of this customer. When the target customer leaves the unmanned store, the monitoring and control terminal will further obtain a list of items taken and a list of items checked out based on the video capture results. The video capture results record all the behaviors of the customer in the store, including the items taken and the items finally checked out. By comparing and analyzing these video data, the monitoring and control terminal can generate a list of items taken and a list of items checked out, so as to accurately grasp the actual purchase situation of the customer.

[0026] Furthermore, obtaining the list of items taken includes: Analyze the video capture results, identify the items taken by the target customer, and obtain a first list of items taken; Obtain the trigger data of sensors installed on the commodity cabinets, and obtain a second list of items taken based on the trigger data of the sensors; Perform a fusion process on the first list of items taken and the second list of items taken to obtain the list of items taken.

[0027] Preferably, analyze the video capture results to identify the products taken by the target customer in the video, thereby generating a first product-taking list, which contains the product information taken by the target customer in the unmanned store. The sensors installed on the product cabinets can be used to detect whether a product has been taken, and the sensor trigger data records the information on which products have been taken by the customer. According to the sensor trigger data, extract and generate a second product-taking list of the target customer, which contains the product information taken by the target customer recorded by the sensor trigger. Perform a fusion process on the first product-taking list and the second product-taking list, compare and integrate the product-taking information obtained by the two different methods to eliminate possible errors and duplicates, so as to obtain a more accurate and complete product-taking list, which contains all the products taken by the target customer in the unmanned store.

[0028] Determine whether the product-taking list matches the checkout product list. When they do not match, activate the warning device of the unmanned store to give an alarm and start the anti-theft plan.

[0029] Compare the product-taking list with the checkout product list to determine whether there is a mismatch. If it is found that the product-taking list does not match the checkout product list, that is, there are uncheckout products or other abnormal situations, immediately activate the warning device of the unmanned store to give an alarm. The warning device can be a sound alarm, a flash or sending an alarm message to relevant personnel, etc. At the same time, start the anti-theft plan of the unmanned store, such as notifying security personnel to intervene, locking the exit to prevent customers from leaving, and tracking suspicious customers through video surveillance, etc.

[0030] Furthermore, the method further includes: When the suspiciousness of the target customer exceeds the suspiciousness threshold, synchronously collect the eye feature sequence of the target customer and generate a line-of-sight change sequence based on the eye feature sequence; Perform an abnormality judgment on the line-of-sight change sequence based on a preset abnormal line-of-sight model to determine the line-of-sight abnormality coefficient; Obtain a comprehensive suspiciousness based on the line-of-sight abnormality coefficient and the suspiciousness of the target customer. When the comprehensive suspiciousness reaches the warning threshold, activate the warning device of the unmanned store to give an alarm and start the anti-theft plan.

[0031] Preferably, when the suspiciousness of the target customer exceeds the suspiciousness threshold, the eye feature sequence of the target customer is collected, and the line-of-sight change sequence is generated based on these features. The eye features can be obtained through eye recognition technology. The line-of-sight change sequence records the change of the target customer's eye fixation points in the unmanned vending store. Based on a preset abnormal line-of-sight model, the obtained line-of-sight change sequence is judged for abnormality. The abnormal line-of-sight model is constructed based on a large amount of historical data and experience accumulation, and includes a set of various abnormal line-of-sight changes. By comparing the line-of-sight change sequence of the target customer with the model, it can be judged whether the line-of-sight change of the target customer is abnormal, and a line-of-sight abnormality coefficient is determined accordingly. The comprehensive suspiciousness is obtained by combining the line-of-sight abnormality coefficient and the suspiciousness of the target customer. The comprehensive suspiciousness is a comprehensive evaluation of the abnormal degree of the target customer's behavior, and comprehensively considers multiple aspects of information, including the shopping behavior and line-of-sight change of the target customer. When the comprehensive suspiciousness reaches the warning threshold, the warning device of the unmanned vending store will be immediately activated for warning, and the anti-theft plan will be started.

[0032] In summary, the embodiments of the present application at least have the following technical effects: An information collection device is arranged in the unmanned vending store. When the target customer enters the unmanned vending store, information of the target customer is collected, and a target customer label is established based on the information collection result. Then, the shopping behavior of the target customer is collected in real time through the distributed camera network in the unmanned vending store. The video collection result will be associated with the target customer label, and action feature extraction is performed to obtain the target customer behavior sequence. Then, based on the target customer label, the target consumption behavior analysis model is retrieved from the preset model library, and the suspiciousness of the target customer is judged for the target customer behavior sequence to obtain the suspiciousness of the target customer. When the suspiciousness of the target customer exceeds the suspiciousness threshold, the target customer label is sent to the monitoring and control terminal. When the target customer leaves the unmanned vending store, the monitoring and control terminal obtains the list of items taken and the list of items checked out based on the video collection result. It is judged whether the list of items taken matches the list of items checked out. When they do not match, the warning device of the unmanned vending store is activated for warning, and the anti-theft plan is started. This solves the technical problem in the prior art that it is difficult for unmanned vending stores to accurately identify theft behaviors, and achieves the technical effect of improving the accuracy of theft behavior identification.

[0033] Embodiment 2 Based on the same inventive concept as an anti-theft monitoring and warning method for an unmanned vending store in the foregoing embodiment, as Figure 2 shown, the present application provides an anti-theft monitoring and warning device for an unmanned vending store. The device in the embodiments of the present application and the method embodiments are based on the same inventive concept. Among them, the device includes: An information collection module 11, which is used to deploy information collection devices in the unmanned vending store. When a target customer enters the unmanned vending store, it collects information about the target customer and establishes a target customer label based on the information collection result; A behavior sequence acquisition module 12, which is used to perform real-time video collection on the shopping behavior of the target customer through the distributed camera network of the unmanned vending store, map and associate the video collection result with the target customer label, and extract action features from the video collection result to obtain the target customer behavior sequence; A suspiciousness judgment module 13, which is used to retrieve a target consumption behavior analysis model from a preset model library based on the target customer label, judge the suspiciousness of the target customer behavior sequence, and obtain the target customer suspiciousness; A monitoring module 14, which is used to send the target customer label to the monitoring control terminal when the target customer suspiciousness exceeds the suspiciousness threshold. When the target customer leaves the unmanned vending store, the monitoring control terminal obtains the list of items taken and the list of items checked out based on the video collection result; An early warning module 15, which is used to judge whether the list of items taken matches the list of items checked out. When they do not match, it activates the warning device of the unmanned vending store to give an alarm and starts an anti-theft plan.

[0034] Furthermore, the information collection module 11 is used to execute the following method: Preset information collection dimensions, form information collection devices based on the information collection dimensions, and deploy them in different areas of the unmanned vending store. Among them, the information collection dimensions at least include facial features, age, gender, and height; Assign a unique tracking identifier to each customer entering the unmanned vending store, and monitor the tracking identifiers of each customer in the unmanned vending store in real time; If it is monitored that there is a customer without an assigned tracking identifier, determine it as the target customer, trigger the information collection device corresponding to the target customer's location information, collect information about the target customer, and obtain the information collection result; Construct a multi-dimensional feature vector of the target customer based on the information collection result, use the multi-dimensional feature vector as the target customer label of the target customer, and assign a tracking identifier to the target customer.

[0035] Furthermore, the suspiciousness judgment module 13 is used to execute the following method: Collect historical customer shopping videos, and obtain a sample customer set and a sample customer behavior sequence set based on the historical customer shopping videos. Each sample customer has a corresponding sample customer label; Identify the suspicious degree of the sample customer behavior sequences in the sample customer behavior sequence set to obtain the sample customer suspicious degree set; Perform clustering analysis on the sample customer labels to obtain multiple label clustering clusters, and extract the centers of the multiple label clustering clusters to generate multiple model identification labels; Divide the sample customer behavior sequence set and the sample customer suspicious degree set based on the multiple label clustering clusters to obtain multiple training sample sets; Perform supervised training on the multiple training sample sets respectively to obtain multiple standard consumption behavior analysis models, and map the multiple standard consumption behavior analysis models one by one with the multiple model identification labels to form a preset model library.

[0036] Further, the suspicious degree judgment module 13 is used to execute the following method: Obtain multiple model identification labels in the preset model library, calculate the label similarity with the target customer label, and obtain the maximum label similarity; Determine the target model identification label based on the maximum label similarity, and extract the corresponding standard consumption behavior analysis model from the preset model library according to the target model identification label as the target consumption behavior analysis model.

[0037] Further, the suspicious degree judgment module 13 is used to execute the following method: Extract the standard consumption behavior analysis model corresponding to the target customer label from the preset model library according to the target customer label; Based on the standard consumption behavior analysis model, copy and generate the target consumption behavior analysis model of the target customer; After the target customer leaves the unmanned vending store, destroy the target consumption behavior analysis model.

[0038] Further, the monitoring module 14 is used to execute the following method: Analyze the video acquisition result, identify the commodities taken by the target customer, and obtain the first commodity taking list; Obtain the trigger data of the sensors arranged on the commodity cabinet, and obtain the second commodity taking list based on the trigger data of the sensors; Perform fusion processing on the first commodity taking list and the second commodity taking list to obtain the taken commodity list.

[0039] Further, the early warning module 15 is used to execute the following method: When the suspicious degree of the target customer exceeds the suspicious degree threshold, synchronously collect the eye feature sequence of the target customer, and generate a line-of-sight change sequence based on the eye feature sequence; Based on a preset abnormal line-of-sight model, perform abnormal judgment on the line-of-sight change sequence to determine the line-of-sight abnormality coefficient; Obtain a comprehensive suspicious degree according to the line-of-sight abnormality coefficient and the target customer's suspicious degree. When the comprehensive suspicious degree reaches the warning threshold, activate the warning device of the unmanned vending store to give an alarm and start the anti-theft plan.

[0040] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0041] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0042] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A theft prevention monitoring and early warning method for an unmanned store, characterized in that: The method comprises: An information collection device is arranged in the unmanned store, and when a target customer enters the unmanned store, information of the target customer is collected, and a target customer tag is established based on the information collection result; The target customer's shopping behavior is captured in real time through a distributed camera network of the unmanned store, the video capture result is mapped and associated with the target customer label, and action features are extracted from the video capture result to obtain the target customer behavior sequence; Based on the target customer tag, a target consumption behavior analysis model is retrieved from a preset model library, and a suspiciousness judgment is made on the target customer behavior sequence to obtain the target customer suspicion; When the target customer's suspicion exceeds the suspicion threshold, the target customer tag is sent to the monitoring control terminal. When the target customer leaves the unmanned store, the monitoring control terminal obtains a list of taken goods and a list of checked-out goods based on the video acquisition results. It is determined whether the taken commodity list matches the checkout commodity list. If they do not match, an early warning device of the unmanned store is activated to give an alarm and start an anti-theft plan.

2. The method according to claim 1, characterized in that Establish target customer tags, including: Preset information collection dimensions, and establish information collection devices based on the information collection dimensions, and arrange them in the unmanned store in different areas, wherein the information collection dimensions at least include facial features, age, gender, and height; Allocating a unique tracking ID to each customer entering the unmanned store, and monitoring the tracking IDs of each customer in the unmanned store in real time; If a customer who is not assigned a tracking identifier is detected, the customer is determined to be a target customer, and an information collection device corresponding to the location information of the target customer is triggered to collect information about the target customer to obtain an information collection result; A multi-dimensional feature vector of the target customer is constructed based on the information collection result, the multi-dimensional feature vector is used as a target customer label of the target customer, and a tracking identifier is allocated to the target customer.

3. The method according to claim 1, characterized in that The method further comprises: Collect historical customer shopping videos, and obtain a sample customer set and a sample customer behavior sequence set based on the historical customer shopping videos, wherein each sample customer has a corresponding sample customer label; Marking the suspiciousness of the sample customer behavior sequences in the sample customer behavior sequence set to obtain a sample customer suspiciousness set; Performing cluster analysis on the sample customer labels to obtain multiple label clusters, extracting the centers of the multiple label clusters to generate multiple model identification labels; Dividing the sample customer behavior sequence set and the sample customer suspicion degree set based on the multiple label clusters to obtain multiple training sample sets; The multiple training sample sets are respectively subjected to supervised training to obtain multiple standard consumer behavior analysis models, and the multiple standard consumer behavior analysis models are mapped one by one with multiple model identification labels to form a preset model library.

4. The method according to claim 3, characterized in that Retrieving a target consumer behavior analysis model from a preset model library based on the target customer tag includes: Obtain multiple model identification tags in a preset model library, calculate the tag similarity with the target customer tag, and obtain the maximum tag similarity; A target model identification tag is determined based on the maximum tag similarity, and a corresponding standard consumer line analysis model is extracted from a preset model library according to the target model identification tag as a target consumer line analysis model.

5. The method according to claim 4, characterized in that The method further comprises: According to the target customer tag, extracting a standard consumer behavior analysis model corresponding to the target customer tag from a preset model library; Based on the standard consumer behavior analysis model, a target consumer behavior analysis model of the target customer is copied and generated; When the target customer leaves the unmanned store, the target consumer behavior analysis model is destroyed.

6. The method according to claim 1, characterized in that Get a list of items to pick up, including: Analyze the video acquisition results, identify the commodities picked up by the target customer, and obtain a first commodity picking list; Acquire sensor trigger data arranged on the commodity cabinet, and acquire a second commodity picking list based on the sensor trigger data; The first commodity picking list and the second commodity picking list are merged to obtain the picking commodity list.

7. The method according to claim 1, characterized in that The method further comprises: When the suspicion level of the target customer exceeds a suspicion level threshold, synchronously collecting an eye feature sequence of the target customer, and generating a sight line change sequence based on the eye feature sequence; Based on a preset abnormal sight line model, the sight line change sequence is judged to be abnormal, and a sight line abnormality coefficient is determined; A comprehensive suspicion degree is obtained according to the sight line abnormality coefficient and the target customer suspicion degree. When the comprehensive suspicion degree reaches a warning threshold, the warning device of the unmanned store is activated to give an alarm and start an anti-theft plan.

8. An anti-theft monitoring and early warning device for an unmanned store, characterized in that: The device is used to implement the anti-theft monitoring and early warning method of an unmanned store as described in any one of claims 1 to 7, comprising: An information collection module, wherein the information collection module is used to deploy an information collection device in the unmanned store, collect information about the target customer when the target customer enters the unmanned store, and establish a target customer tag based on the information collection result; A behavior sequence acquisition module, which is used to perform real-time video acquisition of the shopping behavior of the target customer through the distributed camera network of the unmanned store, map and associate the video acquisition result with the target customer label, and extract action features from the video acquisition result to obtain the target customer behavior sequence; A suspicion judgment module, the suspicion judgment module is used to retrieve a target consumption behavior analysis model from a preset model library based on the target customer tag, perform suspicion judgment on the target customer behavior sequence, and obtain the target customer suspicion; A monitoring module, wherein the monitoring module is used to send the target customer tag to the monitoring control terminal when the target customer's suspicion exceeds a suspicion threshold, and when the target customer leaves the unmanned store, the monitoring control terminal obtains a list of taken goods and a list of checked-out goods based on the video acquisition result; The early warning module is used to determine whether the list of taken goods matches the list of checked-out goods. When they do not match, the early warning device of the unmanned store is activated to issue an alarm and start an anti-theft plan.

Images