Anomaly detection method, apparatus, device, and storage medium

Abstract

Embodiments of the present application provide an anomaly detection method, device, equipment and storage medium. The method comprises: obtaining at least one transaction time of a commodity display place in a target period, an entering time human body image and a leaving time human body image of the same pedestrian and corresponding entering time and leaving time; performing identification processing based on the entering time human body image and the leaving time human body image of the same pedestrian to obtain an identification result for indicating whether the same pedestrian has a purchase behavior; determining a target identification result without corresponding transaction time based on the entering time and the leaving time corresponding to at least one target identification result for indicating the existence of the purchase behavior and at least one transaction time, to obtain the number of target identification results without corresponding transaction time; and obtaining an anomaly detection result corresponding to the target period of the commodity display place based on the number of target identification results without corresponding transaction time. The present application can shorten the time consumption, reduce the labor cost and improve the efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to an exception detection method and device, equipment and a storage medium. BACKGROUND

[0002] In the process of attracting merchants, a shopping mall generally adopts various modes, such as leasing, self-operation, joint operation, etc. The joint operation mode is one of the more common modes.

[0003] In the joint operation mode, the shopping mall generally collects cash and shares risks with merchants. However, in the joint operation mode, there is a situation of single-outlet sales of a joint operation mode merchant, which leads to capital loss on the shopping mall side and damages the trust relationship between the shopping mall and the merchant, and therefore it is necessary to detect single-outlet sales. Generally, in order to find the risk of single-outlet sales, manual investigation is needed from transaction data, inventory data, etc., which is high in labor cost, time-consuming and low in efficiency. SUMMARY

[0004] The embodiments of the present application provide an exception detection method, device, equipment and storage medium to solve the technical problem that manual investigation of data is needed to find the risk of single-outlet sales in the prior art, which is high in labor cost, time-consuming and low in efficiency.

[0005] In a first aspect, the embodiments of the present application provide a method for detecting single-outlet sales, comprising:

[0006] obtaining customer flow data and transaction data of a commodity display place in a target period, the customer flow data comprising a human body image when a same person enters, a human body image when the same person leaves, and corresponding entering time and leaving time, and the transaction data comprising at least one transaction time;

[0007] performing identification processing based on the human body image when the same person enters and the human body image when the same person leaves to obtain an identification result for indicating whether the same person has a purchase behavior;

[0008] based on the entering time and the leaving time corresponding to at least one target identification result for indicating a purchase behavior and the at least one transaction time, determining a target identification result in the at least one target identification result that does not have a corresponding transaction time to obtain a quantity of the target identification result that does not have a corresponding transaction time;

[0009] based on the quantity of the target identification result that does not have a corresponding transaction time, obtaining an exception detection result of the commodity display place corresponding to the target period.

[0010] In a second aspect, the embodiments of the present application provide a device for detecting single-outlet sales, comprising:

[0011] The acquisition module is configured to acquire passenger flow data and transaction data of a commodity display place in a target period, the passenger flow data comprising an entering human body image, a leaving human body image and corresponding entering and leaving times of the same pedestrian, and the transaction data comprising at least one transaction time.

[0012] The identification module is configured to perform identification processing based on the entering human body image and the leaving human body image of the same pedestrian, to obtain an identification result indicating whether the same pedestrian has a purchase behavior.

[0013] The first determination module is configured to determine a target identification result without a corresponding transaction time in the at least one target identification result based on the entering and leaving times corresponding to the at least one target identification result indicating a purchase behavior and the at least one transaction time, to obtain a quantity of the target identification result without the corresponding transaction time.

[0014] The second determination module is configured to obtain an abnormality detection result of the commodity display place corresponding to the target period based on the quantity of the target identification result without the corresponding transaction time.

[0015] In the third aspect, an electronic device is provided, comprising a memory and a processor, wherein the memory stores one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the method in any of the first aspect.

[0016] In the fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program is executed to implement the method in any of the first aspect.

[0017] The embodiments of the present application also provide a computer program, which is executed by a computer to implement the method in any of the first aspect.

[0018] In the embodiments of the present application, the identification processing can be performed based on the entering human body image and the leaving human body image of the same pedestrian, to obtain an identification result indicating whether the same pedestrian has a purchase behavior, the target identification result without the corresponding transaction time in the at least one target identification result can be determined based on the entering and leaving times corresponding to the at least one target identification result indicating a purchase behavior and the at least one transaction time, to obtain a quantity of the target identification result without the corresponding transaction time, and the abnormality detection result of the commodity display place corresponding to the target period can be obtained based on the quantity, so that the commodity display place with a risk of flying single is automatically found based on the passenger flow data and the transaction data of the commodity display place, the time cost for finding the risk of flying single is shortened, the labor cost is reduced, and the efficiency of finding the risk of flying single is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 A schematic diagram of an application scenario of an embodiment of the present application;

[0021] Figure 2 A schematic flow chart of a method for detecting flyers provided in one embodiment of the present application;

[0022] Figure 3 A schematic flow chart of a method for detecting flyers provided in another embodiment of the present application;

[0023] Figure 4 A schematic structural diagram of a flyer detection device provided in one embodiment of the present application;

[0024] Figure 5 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] The terms used in the examples of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a," "the," and "the" used in the examples of this application and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two, but does not exclude the inclusion of at least one.

[0027] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0028] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0029] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or system. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or system comprising the element.

[0030] In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0031] In order to facilitate those skilled in the art to understand the technical solution provided by the embodiments of the present application, the technical environment in which the technical solution is implemented is described below.

[0032] Figure 1 Schematic diagram of the application scenario of the method for detecting flyers provided in the embodiment of the present application, such as Figure 1 As shown, the application scenario may include a merchandise display place 11, a photographing device 12, and an electronic device 13. The merchandise display place 11 refers to any type of place that can be used for merchandise display. For example, the merchandise display place 11 can be a counter in a shopping mall or a merchant store, and the number of merchandise display places 11 can be one or more. The photographing device 12 can be used to collect data on pedestrians entering and exiting the merchandise display place 11. The photographing device 12 can be, for example, a camera or other device or equipment that can take pictures. The electronic device 13 can identify merchandise display places 11 that are at risk of flying orders based on the data collected by the photographing device 12. The electronic device 13 can be, for example, a personal computer, a server, or other equipment that can process data. Flying orders refer to the situation where marketing personnel do not enter the sales into the system after receiving the order or do not enter the sales for the day into the system, but instead deliver the order privately and profit by withholding sales orders.

[0033] Usually, in order to detect the risk of unauthorized sales, manual investigation is required on transaction data and inventory data at the product display locations. This is labor-intensive, time-consuming, and inefficient.

[0034] In order to solve the technical problems of high labor cost, long time consumption and low efficiency caused by manually checking data to find the fly single risk, in the embodiment of the present application, the entering human body image and the leaving human body image of the same pedestrian are identified to obtain an identification result for indicating whether the same pedestrian has a purchase behavior, and based on the entering time and the leaving time corresponding to at least one target identification result for indicating the existence of the purchase behavior and at least one transaction time, a target identification result without a corresponding transaction time in the at least one target identification result is determined to obtain the number of target identification results without a corresponding transaction time, and based on the number, an abnormal detection result of the commodity display corresponding to the target period is obtained, which realizes automatic discovery of the commodity display with the fly single risk based on the passenger flow data and the transaction data of the commodity display, not only shortens the time consumption of finding the fly single risk, reduces the labor cost, but also improves the discovery efficiency of the fly single risk.

[0035] It should be noted that, Figure 1 In the embodiment, data acquisition by the shooting device 12 is taken as an example, and it can be understood that when the electronic device 13 has an image acquisition function, the electronic device 13 can also be used for acquisition in other embodiments.

[0036] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the case of no conflict, the embodiments described below and the features in the embodiments can be combined with each other.

[0037] Figure 2 The flowchart of the identification method of the thief provided by an embodiment of the present application, the embodiment can be applied to the electronic device 13 in Figure 1 , and specifically can be executed by the processor of the electronic device 13. As shown in Figure 2 , the method of the embodiment can include:

[0038] Step 21, obtaining passenger flow data and transaction data of a commodity display in a target period, the passenger flow data including entering human body images, leaving human body images and corresponding entering time and leaving time of the same pedestrian, and the transaction data including at least one transaction time;

[0039] Step 22, identifying the entering human body images and the leaving human body images of the same pedestrian to obtain an identification result for indicating whether the same pedestrian has a purchase behavior;

[0040] Step 23, based on the entering time and the leaving time corresponding to at least one target identification result for indicating the existence of the purchase behavior and at least one transaction time, determining a target identification result without a corresponding transaction time in the at least one target identification result to obtain the number of target identification results without a corresponding transaction time;

[0041] Step 24, based on the number of target recognition results corresponding to the absence of transaction time, the abnormal detection result of the commodity display corresponding to the target period is obtained.

[0042] In the embodiments of the present application, the target period refers to a period in which it is necessary to determine whether there is a fly single risk. The target period can be flexibly implemented according to the requirements. For example, the length of the target period can be 1 day, so that it can be determined whether there is a fly single risk in the commodity display on any day.

[0043] On the one hand, the passenger flow data of the commodity display in the target period can be obtained. The passenger flow data can include the human body image when the same pedestrian enters the commodity display (hereinafter referred to as the entering human body image) and the human body image when the same pedestrian leaves the commodity display (hereinafter referred to as the leaving human body image). The passenger flow data can also include the time corresponding to the entering human body image (hereinafter referred to as the entering time) and the time corresponding to the leaving human body image (hereinafter referred to as the leaving time). On the other hand, the transaction data of the commodity display in the target period can be obtained. The transaction data can include at least one transaction time. The transaction data can be any type of data that can reflect the transaction of the commodity display. In one embodiment, the transaction data can be order data, and the corresponding transaction time can be order time. In another embodiment, the transaction data can be payment data, and the corresponding transaction time can be payment time.

[0044] For example, the transaction data can be obtained from a cash register system. Taking the commodity display as a counter of a joint operation mode merchant in a shopping mall as an example, the cash register system can be the cash register system of the shopping mall. Unlike the self-operation and rental modes, the joint operation mode is a mode in which the shopping mall and the merchant share the operating risk, and the cash register is unified by the shopping mall. At the same time, the shopping mall has a deeper involvement in brand goods, salespersons, etc. than the rental mode. Of course, in other embodiments, the transaction data can also be obtained by other means, which are not limited in the present application. Optionally, the transaction data can also include the transaction volume.

[0045] For example, the passenger flow data can be obtained by passenger flow analysis based on the data collected by the shooting device. The shooting device can be arranged at any position that can shoot the pedestrian entering and leaving the commodity display, such as the entrance of the commodity display or the corridor of the shopping mall.

[0046] In an embodiment, the photographing device can be an artificial intelligence (AI) camera, the AI camera can include a processor and an image sensor, and the AI camera can capture images continuously in a case where a pedestrian enters a photographing range of the AI camera, and the AI camera can determine whether the pedestrian enters or leaves the commodity display based on the captured images, and in a case where it is determined that the pedestrian enters or leaves the display, the AI camera can report, to the electronic device, a time at which the pedestrian enters or leaves the display and a corresponding human body image, and the human body image can be obtained by performing human body detection on the images.

[0047] In an embodiment, the entering human body image and the leaving human body image of the same pedestrian can be calculated by the electronic device based on the data reported by the AI camera, and the electronic device can extract features of the human body images to obtain human body features, and determine the entering human body image and the leaving human body image of the same pedestrian based on the human body features.

[0048] Optionally, the passenger flow data can further include a motion trajectory of the same pedestrian when entering the commodity display (which can be denoted as an entering trajectory) and a motion trajectory of the same pedestrian when leaving the display (which can be denoted as a leaving trajectory), and / or an entering number. In an embodiment, the entering number can be counted by the electronic device based on the data reported by the AI camera, and the entering trajectory and the leaving trajectory can be obtained by the AI camera, i.e., the AI camera can track the human body.

[0049] In an embodiment, after obtaining the passenger flow data of the commodity display in the target period, the entering human body image and the leaving human body image of the same pedestrian can be recognized to obtain a recognition result indicating whether the same pedestrian has a purchase behavior.

[0050] In an embodiment, an identification model can be used to process the entering human body image and the leaving human body image of the same pedestrian to obtain a recognition result indicating whether the same pedestrian has a purchase behavior. In an embodiment, the identification model can be a model constructed based on a deep learning algorithm.

[0051] Exemplarily, the identification model can be trained in the following manner: constructing an identification model, the identification model being provided with training parameters; inputting pairs of sample human body images into the identification model respectively to generate prediction results; based on the difference between the prediction results and expected results corresponding to the sample labels of each pair of sample human body images, iteratively adjusting the training parameters until the difference meets preset requirements. Each pair of sample human body images includes one entering human body image and one leaving human body image of the same pedestrian entering and leaving the same commodity display. In one embodiment, the sample label of a pair of sample human body images can be used to represent whether the corresponding pedestrian has a purchase behavior.

[0052] In one embodiment, whether a purchase behavior exists can be identified by identifying the type of the bag carried by the pedestrian when entering and leaving and comparing the change in the number of bags. In this case, the identification model can include a type identification submodule, a number comparison submodule, and a result generation submodule. The type identification submodule can be used to process the entering human body image to obtain a first identification result of whether the entering human body image carries a bag of a target type, and to process the leaving human body image to obtain a second identification result of whether the leaving human body image carries a bag of the target type. The number comparison submodule is used to process the entering human body image and the leaving human body image to obtain a comparison result of whether the number of bags carried by the leaving human body image is greater than the number of bags carried by the entering human body image. The result generation submodule is used to obtain an identification result representing whether the same pedestrian has a purchase behavior based on the first identification result, the second identification result, and the comparison result.

[0053] The target type refers to the type of the bag provided by the merchant for customers to carry purchased goods. For example, in the case of a shopping mall, the target type can be a shopping bag. Since the bag provided by the merchant for customers to carry purchased goods is different from the bag brought by the customer in terms of material, shape, color, etc., the identification model can identify whether the bag of the target type exists in the human body image.

[0054] Exemplarily, in the case where the first identification result is that the entering human body image does not carry a bag of the target type, and the second identification result is that the leaving human body image carries a bag of the target type, the result generation submodule can obtain an identification result representing that a purchase behavior exists.

[0055] Exemplarily, in the case where the first identification result is that the entering human body image carries a bag of the target type, the second identification result is that the leaving human body image carries a bag of the target type, and the comparison result is that the number of bags carried by the leaving human body image is greater than the number of bags carried by the entering human body image, the result generation submodule can obtain an identification result representing that a purchase behavior exists.

[0056] For example, when the first identification result is that the target type of package is carried in, the second identification result is that the target type of package is carried out, and the comparison result is that the number of packages carried out is not more than the number of packages carried in, the result generation sub-module can obtain an identification result indicating that there is no purchase behavior.

[0057] It should be noted that when there are multiple pedestrians entering and leaving the commodity display area in the target period, multiple identification results can be obtained; when the same pedestrian enters and leaves the commodity display area multiple times in the target period, multiple identification results can be obtained.

[0058] In the embodiments of the present application, the identification result indicating that there is a purchase behavior can be recorded as a target identification result. After obtaining the identification result indicating whether the same pedestrian has a purchase behavior, the number of target identification results without corresponding transaction time can be determined based on the entering time and the leaving time corresponding to at least one target identification result and the at least one transaction time. The purchase behavior corresponding to the target identification result without corresponding transaction time can be understood as a purchase behavior with abnormal risk.

[0059] In one embodiment, the number of target identification results without corresponding transaction time can be determined by traversing the transaction time. In this case, step 23 can specifically include: sequentially processing each transaction time in the at least one transaction time as follows to obtain the number of target identification results without corresponding transaction time in the at least one target identification result: judging whether there is a target identification result in the target identification results without corresponding transaction time, in which the current transaction time is located between the entering time and the leaving time corresponding to the target identification result; if yes, the current transaction time corresponds to the target identification result; the next transaction time of the current transaction time is taken as the current transaction time and the judgment step is executed until the at least one transaction time is traversed.

[0060] For example, assuming that the at least one transaction time is 11:30:20, 12:20:15 and 14:15:45 respectively, and the entering time and the leaving time corresponding to the plurality of target recognition results are 11:29:15 and 11:32:12, 11:29:18 and 11:35:20, 12:19:20 and 12:30:30, 14:10:10 and 14:16:00, 15:15:20 and 15:20:00 respectively, firstly, the transaction time "11:30:20" can be taken as the current transaction time, at this time, the five target recognition results are all target recognition results which do not correspond to the current transaction time, thus it can be judged whether there is a target recognition result in the five target recognition results in which the current transaction time "11:30:20" is located between the entering time and the leaving time corresponding to the target recognition result. Since there are "11:29:15 and 11:32:12" and "11:29:18 and 11:35:20", the transaction time "11:30:20" can be corresponded to the target recognition result corresponding to any one of the two, for example, "11:29:15 and 11:32:12". Then, the transaction time "12:20:15" can be taken as the current transaction time, at this time, there are still four target recognition results which are target recognition results which do not correspond to the current transaction time, thus it can be judged whether there is a target recognition result in the four target recognition results in which the current transaction time "12:20:15" is located between the entering time and the leaving time corresponding to the target recognition result. Since there are "12:19:20 and 12:30:30", the transaction time "12:20:15" can be corresponded to the target recognition result corresponding to "12:19:20 and 12:30:30". Finally, the transaction time "14:15:45" can be taken as the current transaction time, at this time, there are still three target recognition results which are target recognition results which do not correspond to the current transaction time, thus it can be judged whether there is a target recognition result in the three target recognition results in which the current transaction time "14:15:45" is located between the entering time and the leaving time corresponding to the target recognition result. Since there are "14:10:10 and 14:16:00", the transaction time "14:15:45" can be corresponded to the target recognition result corresponding to "14:10:10 and 14:16:00". Since there are still two target recognition results which do not correspond to the transaction time, it can be obtained that the number of target recognition results which do not correspond to the transaction time is two.

[0061] In the embodiments of the present application, after the number of target recognition results which do not correspond to the transaction time is determined, the abnormality detection result corresponding to the target period of the commodity display place can be obtained based on the number. The abnormality detection result can exist abnormal risk or not exist abnormal risk, and it needs to be noted that the abnormality in the present application specifically refers to the flying single.

[0062] In one embodiment, the abnormality detection result of the commodity display corresponding to the target period can be obtained based on only the number of target recognition results without corresponding transaction time, for example, the abnormality detection result of the commodity display corresponding to the target period can be obtained when the number of target recognition results without corresponding transaction time is greater than a number threshold, and the abnormality detection result of the commodity display corresponding to the target period can be obtained when the number of target recognition results without corresponding transaction time is less than the number threshold.

[0063] In another embodiment, the index data can be determined based on the number of target recognition results without corresponding transaction time, and the abnormality detection result of the commodity display corresponding to the target period can be determined according to the index data. Thus, the data dimension considered in detecting the abnormality can be increased, which is beneficial to improve the accuracy of the detection result. Based on this, step 24 can specifically include: calculating first index data based on the number of target recognition results without corresponding transaction time; and obtaining the abnormality detection result of the commodity display corresponding to the target period based on the obtained index data, the index data including the first index data.

[0064] The first index data is used to represent the recall rate of the purchase behavior. In one embodiment, the first index data can be obtained based on the transaction amount and the number of at least one target recognition result, and based on this, the transaction data can further include the transaction amount, and the calculation of the first index data based on the number of target recognition results without corresponding transaction time can specifically include: summing the transaction amount and the number of target recognition results without corresponding transaction time to obtain a sum result; and taking the ratio of the number of at least one target recognition result to the sum result as the first index data. The transaction amount + the number of target recognition results without corresponding transaction time can be understood as the number of transactions that actually occur, so that the proportion of the number of transactions that can be identified in the number of transactions that actually occur can be considered when identifying the abnormality risk.

[0065] In yet another embodiment, the relationship between the number of at least one target recognition result and / or the transaction amount and the number of entries can also be considered when determining the abnormality risk, and based on this, the passenger flow data can further include the number of entries; the abnormality detection result of the commodity display corresponding to the target period based on the number of target recognition results without corresponding transaction time can further include: taking the ratio of the number of at least one target recognition result to the number of entries as second index data; and / or, taking the ratio of the transaction amount to the number of entries as third index data; and the index data further includes the second index data and / or the third index data. Thus, the conversion rate from entering the store to being identified as a purchase and / or the conversion rate from entering the store to a successful transaction can be considered when identifying the abnormality risk.

[0066] Optionally, in the abnormality detection of the commodity display, in addition to considering the passenger flow data and transaction data of the commodity display in the same period, the change of transaction data between different periods of the commodity display can also be considered. Based on this, in another embodiment, the method provided by the embodiment can further include: obtaining first transaction data of the commodity display in a first period and second transaction data of the commodity display in a second period; the first transaction data and the second transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the total number of transactions, or the number of commission transactions; and based on the first transaction data and the second transaction data, a fourth index data used to represent the change of transaction data is calculated. It should be understood that the number of fourth index data can be one or more.

[0067] The first period and the target period can be the same period or different periods. When they are different periods, the first period can include the target period. The first period and the second period are different periods, and the length of the first period is the same as the length of the second period. Taking order data as an example, the number of refund transactions can be the number of refund orders, the number of return and exchange transactions can be the number of return and exchange orders, the total number of transactions can be the number of transaction orders, and the number of commission transactions can be the number of commission orders.

[0068] Taking the length of the first period and the second period as 1 month and the transaction data as the number of refund orders as an example, the number of refund orders can be compared with the number of refund orders of the last month. A fourth index data can be equal to (current month refund order number / last month refund order number-1), and / or, the number of refund orders can be compared with the number of refund orders of the corresponding month of the last year. Another fourth index data can be equal to (current month refund order number / last year corresponding month refund order number-1). The current month can refer to the month in which the target period is located.

[0069] Optionally, in the abnormality detection of the commodity display, in addition to considering the transaction data of the commodity display, the change of transaction data between the commodity display and other commodity displays can also be considered. Based on this, in another embodiment, the method provided by the embodiment can further include: obtaining first transaction data of the commodity display in a first period and third transaction data of the same industry commodity display in the first period; the first transaction data and the third transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the total number of transactions, or the number of commission transactions; and based on the first transaction data and the third transaction data, a fifth index data used to represent the change of transaction data is calculated. The index data can also include the fifth index data. It should be understood that the number of fifth index data can be one or more.

[0070] The first time period and the target time period can be the same time period or different time periods. When they are different time periods, the first time period can include the target time period. The same format commodity display place refers to a commodity display place of the same format. For example, when the commodity display place is a clothing counter, the same format commodity display place can be one or more clothing counters.

[0071] For example, the fifth index data can be calculated by calculating the proportion of the transaction data of the commodity display place in the average transaction data of the same format commodity display place. Taking the first time period as one month and the transaction data as the number of transaction orders as an example, a corresponding fifth index data can be equal to (the number of transaction orders in the current month / the average number of transaction orders of the same format counter-1). The current month can refer to the month in which the target time period is located.

[0072] In the embodiments of the present application, the number of index data of the commodity display place corresponding to the target time period can be one or more.

[0073] In one embodiment, when the number of index data is multiple, each index data can have a corresponding detection threshold. The aforementioned obtaining of the abnormal detection result of the commodity display place corresponding to the target time period based on the obtained index data can include: determining a comparison value corresponding to each index data based on the size relationship between each index data and its corresponding detection threshold; calculating a total comparison value based on the comparison value corresponding to each index data; and obtaining the abnormal detection result of the commodity display place corresponding to the target time period based on the size relationship between the total comparison value and the total threshold.

[0074] For example, when the index data is less than its detection threshold, the comparison value obtained can be 0, and when the index data is greater than its detection threshold, the comparison value obtained can be 1. For example, when the total comparison value is less than the total threshold, the abnormal detection result of the commodity display place corresponding to the target time period can be that there is no abnormal risk, and when the total comparison value is greater than the total threshold, the abnormal detection result of the commodity display place corresponding to the target time period can be that there is an abnormal risk.

[0075] In one embodiment, the importance of different index data can be distinguished when calculating the total comparison value. Based on this, each index data can have a corresponding weight. The calculation of the total comparison value based on the comparison value corresponding to each index data can include: calculating the total comparison value by weighted summation based on the comparison value and the weight corresponding to each index data. For example, the weight corresponding to the first index data, the second index data and the third index data can be greater than the weight corresponding to the fourth index data and the fifth index data, and the weight corresponding to the fourth index data and the fifth index data can be the same.

[0076] Optionally, if there is an abnormal risk at the merchandise display location, a corresponding prompt message can be output so that staff can be promptly notified and conduct manual review. Based on this, in one embodiment, the method provided by the embodiment of the present application can also include: outputting an alarm prompt message, where the alarm prompt message is used to indicate that there is an abnormal risk at the merchandise display location within a target time period. For example, using the electronic device as a server, the alarm prompt message can be sent to a terminal used by staff, and the terminal can then prompt the staff with the alarm prompt message.

[0077] Among them, the alarm prompt information can be flexibly implemented according to the prompt needs. For example, the alarm prompt information can also be used to prompt the indicator data corresponding to the commodity display location with abnormal risks, the human body image at the time of entry, the human body image at the time of exit, the trajectory at the time of entry and the trajectory at the time of exit, etc. corresponding to the target recognition result that does not have the corresponding transaction time.

[0078] In one embodiment, when there are multiple product display locations and target time periods, the alarm prompt information can be used to specifically prompt the existence of abnormal risks at the product display location within the target time period from the time dimension and the product display location dimension, for example, it can be displayed in the form of a chart.

[0079] And / or in another embodiment, if the product display location includes a counter in a shopping mall, the alarm information can also be used to indicate the number of people entering the mall during a target time period. For example, the number of people entering can be obtained by electronic devices based on data reported by AI cameras installed at the mall entrance. This allows staff to review abnormal risks based on the number of people entering the mall during the target time period.

[0080] The anomaly detection method provided in this embodiment obtains an identification result indicating whether the same pedestrian has engaged in purchasing behavior by performing recognition processing based on the human body images of the same pedestrian at the time of entry and the human body images of the same pedestrian at the time of exit. Based on the entry time and exit time and at least one transaction time corresponding to at least one target recognition result indicating the existence of purchasing behavior, the method determines the target recognition result for which no corresponding transaction time exists in the at least one target recognition result, so as to obtain the number of target recognition results for which no corresponding transaction time exists, and obtains the anomaly detection result corresponding to the target time period at the commodity display location based on the number. This method realizes automatic discovery of commodity display locations with risk of flying orders based on the passenger flow data and transaction data at the commodity display locations, which not only shortens the time consumption in discovering the risk of flying orders and reduces labor costs, but also improves the efficiency in discovering the risk of flying orders.

[0081] In one embodiment, taking the counter in a shopping mall in a joint venture model (hereinafter referred to as a joint venture counter) as an example, the overall processing flow of the method for detecting fly-by-line can be as follows: Figure 3As shown, it can be mainly divided into five stages of data preparation, data aggregation, data mining, data presentation and manual review.

[0082] 1. Data preparation stage

[0083] (1) Passenger flow data preparation: To identify the passenger flow of the joint store of the mall, it is necessary to establish the passenger flow identification capability of the mall and the passenger flow identification capability of the joint store, so as to identify the passenger flow of the mall and the passenger flow of the joint store. The passenger flow identification capability can be established by building a passenger flow identification system and deploying AI cameras online, and corresponding passenger flow data can be generated. The input of the passenger flow identification system can be the data reported by the AI camera, and the output of the passenger flow identification system can be the passenger flow data. The passenger flow data of the mall may, for example, include the number of entries, and the passenger flow data of the joint store may, for example, include the body image of the same person when entering and leaving, as well as the corresponding entering time and leaving time.

[0084] (2) Bag data preparation: To identify whether the customer entering and leaving the joint store has a purchase behavior in the joint store, the bag action and handbag can be identified based on the body image when entering and leaving, and the bag data can be differentiated. The bag data can be understood as the identification result described above.

[0085] 2. Data aggregation stage

[0086] In the data aggregation stage, the passenger flow data, bag data and transaction data of the joint store obtained in the data preparation stage can be aggregated, so as to be used in the data mining stage. In the data aggregation stage, data monitoring can also be performed to timely find the abnormal changes of the data, so as to find the cause of the abnormal changes of the data as soon as possible, such as camera abnormality, etc.

[0087] 3. Data mining stage

[0088] In the data mining stage, data mining can be performed on the data in the data aggregation stage to obtain index data, which can include the first index data, the second index data, etc. described above. In addition, in the data mining stage, the fly single algorithm can also be improved. The fly single algorithm refers to an algorithm for determining whether there is a fly single risk based on the index data. The index data used for fly single detection and the feedback results of manual review can be improved by the developers based on the fly single algorithm to improve the accuracy of fly single detection.

[0089] 4. Data presentation stage

[0090] In the data presentation stage, based on the index data obtained in the data mining stage, the single flight risk existing counter can be determined through the single flight algorithm, and the alarm message can be displayed in the form of a chart and timely pushed to the mall operator. For example, the index data, the human body image at the time of entering, and the human body image at the time of leaving can be displayed in the chart.

[0091] 5. Artificial auditing stage

[0092] In the artificial auditing stage, the situation of the counter existing single flight risk obtained in the data presentation stage can be audited offline, the evidence can be collected, and the single flight system based on the single flight algorithm can be fed back to mark the accuracy of the single flight record, which can provide data support for the iteration of the single flight algorithm. The single flight risk existing in a certain target period in a certain commodity display place can be understood as a single flight record.

[0093] The embodiment can combine the joint business transaction data, the customer flow data of the joint counter obtained through customer flow identification, and the bag data obtained through bag identification, analyze and mine index data from the data dimension, identify the joint counter existing single flight risk based on the index data, provide single flight alarm and data support for the mall operator, and the whole process can be automatically completed, thereby saving manpower, improving the inspection efficiency of the single flight, and as the single flight algorithm is strengthened, a higher accuracy can be obtained.

[0094] Figure 4 The structure diagram of the single flight detection device provided by an embodiment of the present application is shown in FIG. 1. As shown in FIG. 1, the embodiment provides a single flight detection device, which can execute the single flight detection method provided by the above-mentioned embodiments. Specifically, the device can include: Figure 4

[0095] The acquisition module 41 is configured to acquire customer flow data and transaction data of a commodity display place in a target period. The customer flow data includes human body images at the time of entering and leaving of the same person and corresponding entering and leaving times, and the transaction data includes at least one transaction time.

[0096] The identification module 42 is configured to perform identification processing based on the human body images at the time of entering and leaving of the same person to obtain an identification result for indicating whether the same person has a purchase behavior.

[0097] The first determination module 43 is configured to determine a target identification result without a corresponding transaction time in at least one target identification result based on the entering and leaving times corresponding to the target identification result with a purchase behavior and the at least one transaction time, to obtain the number of the target identification result without the corresponding transaction time.

[0098] ​The second determining module 44 is configured to obtain an abnormality detection result of the commodity display corresponding to the target time period based on the number of target recognition results without corresponding transaction time.

[0099] In a possible implementation, the identifying module 42 can be specifically configured to input the entering human body image and the leaving human body image of the same pedestrian into an identifying model for processing to obtain an identification result for indicating whether the same pedestrian has a purchase behavior.

[0100] In a possible implementation, the identifying model comprises a type identifying sub-module, a quantity comparing sub-module and a result generating sub-module. The type identifying sub-module is configured to process the entering human body image to obtain a first identification result of whether a bag of a target type is carried at the entering time, and process the leaving human body image to obtain a second identification result of whether the bag of the target type is carried at the leaving time, the target type being a type of a bag provided by a merchant for containing goods purchased by a customer. The quantity comparing sub-module is configured to process the entering human body image and the leaving human body image to obtain a comparison result of whether the number of bags carried at the leaving time is more than the number of bags carried at the entering time. The result generating sub-module is configured to obtain an identification result for indicating whether the same pedestrian has a purchase behavior based on the first identification result, the second identification result and the comparison result.

[0101] In a possible implementation, the identifying model is trained in the following manner: the identifying model is constructed, and the training parameters are set in the identifying model; a plurality of pairs of sample human body images are input into the identifying model respectively to generate prediction results; the training parameters are iteratively adjusted based on a difference between the prediction results and expected results corresponding to sample labels of each pair of sample human body images until the difference meets a preset requirement.

[0102] In a possible implementation, the first determining module 43 can be specifically configured to process each transaction time in the at least one transaction time in sequence to obtain the number of target recognition results without corresponding transaction time in the at least one target recognition result in the following manner: determining whether there is a target recognition result in the target recognition result without corresponding transaction time, in which the current transaction time is located between the corresponding entering time and leaving time of the target recognition result; if yes, the current transaction time is corresponding to the target recognition result; a next transaction time of the current transaction time is taken as the current transaction time, and the determining step is returned to be executed until the at least one transaction time is traversed.

[0103] In a possible implementation, the second determining module 44 can be specifically configured to: calculate first index data based on the number of target recognition results without corresponding transaction time, the first index data being used to represent the recall rate of recognizing the purchase behavior; and obtain the anomaly detection result of the commodity display place corresponding to the target time period based on the obtained index data, the index data including the first index data.

[0104] In a possible implementation, the transaction data further includes transaction volume; and the second determining module 44 calculates the first index data based on the number of target recognition results without corresponding transaction time, including: summing the transaction volume and the number of target recognition results without corresponding transaction time to obtain a summation result; and taking the ratio of the number of at least one target recognition result to the summation result as the first index data.

[0105] In a possible implementation, the passenger flow data further includes the number of entries; and the second determining module 44 determines whether the commodity display place has a fly single risk in the target time period based on the number of target recognition results without corresponding transaction time, further including: taking the ratio of the number of at least one recognition result to the number of entries as the second index data; and / or taking the ratio of the transaction volume to the number of entries as the third index data; and the index data further includes the second index data and / or the third index data.

[0106] In a possible implementation, the obtaining module 41 is further configured to: obtain first transaction data of the commodity display place in a first time period and second transaction data of the commodity display place in a second time period; and the first transaction data and the second transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the number of total transactions, or the number of commission transactions.

[0107] The second determining module 44 is further configured to: calculate fourth index data used to represent the change of transaction data based on the first transaction data and the second transaction data, the index data further including the fourth index data.

[0108] In a possible implementation, the obtaining module 41 is further configured to: obtain first transaction data of the commodity display place in a first time period and third transaction data of a same-industry commodity display place in the first time period; and the first transaction data and the third transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the number of total transactions, or the number of commission transactions.

[0109] The second determining module 44 is further configured to calculate, based on the first transaction data and the third transaction data, fifth index data for representing a change in transaction data, and the index data further includes the fifth index data.

[0110] In a possible implementation, the number of index data is a plurality, and each index data has a corresponding detection threshold. The second determining module 44 obtains the abnormality detection result of the product display place corresponding to the target period based on the obtained index data, including: determining a comparison value corresponding to each index data based on the size relationship between each index data and the corresponding detection threshold; calculating a total comparison value based on the comparison value corresponding to each index data; and obtaining the abnormality detection result of the product display place corresponding to the target period based on the size relationship between the total comparison value and a total threshold.

[0111] In a possible implementation, the apparatus of the embodiment can further include an alarm module configured to output an alarm prompt information, and the alarm prompt information is used to prompt that the product display place has an abnormal risk in the target period.

[0112] In a possible implementation, the number of product display places and the number of target periods are both a plurality, and the alarm prompt information is specifically used to prompt the abnormal risk of the product display place in the target period from the time dimension and the product display place dimension; and / or, the product display place includes a store counter, and the alarm prompt information is further used to prompt the number of people entering the mall where the product display place is located in the target period.

[0113] Figure 4 The apparatus can perform the method provided by the embodiment. Figure 2 The method provided by the embodiment does not describe the parts not described in detail, and the related descriptions of the Figure 2 embodiment can be referred to. The execution process and technical effects of the technical solution are described in the Figure 2 embodiment and will not be described here.

[0114] In a possible implementation, Figure 4 The structure of the apparatus can be implemented as an electronic device. As Figure 5 shown, the electronic device can include a processor 51 and a memory 52. The memory 52 stores a program supporting the controller to perform the method provided by the Figure 2 embodiment, and the processor 51 is configured to execute the program stored in the memory 52.

[0115] The program includes one or more computer instructions, and the one or more computer instructions can implement the following steps when executed by the processor 51:

[0116] Obtain customer flow data and transaction data of the commodity display at a target period, the customer flow data comprising an entering human body image, a leaving human body image and corresponding entering time and leaving time of the same pedestrian, and the transaction data comprising at least one transaction time;

[0117] Perform identification processing based on the entering human body image and the leaving human body image of the same pedestrian, to obtain an identification result for indicating whether the same pedestrian has a purchase behavior;

[0118] Based on the entering time and the leaving time corresponding to at least one target identification result for indicating a purchase behavior and the at least one transaction time, determine a target identification result in the at least one target identification result that does not have a corresponding transaction time, to obtain a quantity of the target identification result that does not have a corresponding transaction time;

[0119] Based on the quantity of the target identification result that does not have a corresponding transaction time, obtain an anomaly detection result of the commodity display corresponding to the target period.

[0120] Optionally, the processor 51 is further configured to perform all or part of the steps of the foregoing method. Figure 2

[0121] The electronic device further includes a communication interface 53 for communication between the electronic device and other devices or communication networks.

[0122] The embodiments of the present application also provide a computer readable storage medium having a computer program stored thereon, when the computer program is executed, the method according to the embodiments shown in the above method. Figure 2

[0123] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments. Those skilled in the art can understand and implement without creative labor.

[0124] ​​Those skilled in the art can clearly understand that each embodiment can be implemented by means of an additional universal hardware platform, and of course can also be implemented by means of a combination of hardware and software. Based on such an understanding, the above technical solutions can be embodied in the form of a computer product, and the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0125] The present application is described with reference to flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in the flowcharts and / or block diagrams.

[0126] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in the flowcharts and / or block diagrams.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in the flowcharts and / or block diagrams.

[0128] In a typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memories.

[0129] Memory can include non-persistent memory in computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory. Memory is an example of computer-readable media.

[0130] Computer-readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, link lists, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carriers.

[0131] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting anomalies, characterized in that: include: Obtaining customer flow data and transaction data for a product display during a target time period, wherein the customer flow data includes images of a pedestrian entering and leaving the display, as well as corresponding entry and exit times, and the transaction data includes at least one transaction time; Performing recognition processing based on the human body images of the same pedestrian when entering and when leaving, to obtain a recognition result indicating whether the same pedestrian has engaged in purchasing behavior; Based on the entry time and the exit time corresponding to at least one target recognition result indicating the existence of a purchase behavior and the at least one transaction time, determining a target recognition result in the at least one target recognition result that does not have a corresponding transaction time, so as to obtain the number of target recognition results that do not have a corresponding transaction time; Based on the number of target recognition results that do not have corresponding transaction times, an abnormality detection result corresponding to the target time period at the commodity display location is obtained.

2. The method according to claim 1, characterized in that The identification processing is performed based on the human body image of the same pedestrian when entering and the human body image of the same pedestrian when leaving, to obtain an identification result indicating whether the same pedestrian has engaged in purchasing behavior, including: The human body images of the same pedestrian when entering and when leaving are input into the recognition model for processing to obtain a recognition result indicating whether the same pedestrian has engaged in purchasing behavior.

3. The method according to claim 2, characterized in that The recognition model includes a type recognition submodule, a quantity comparison submodule and a result generation submodule; The type recognition submodule is configured to process the image of the human body upon entry to obtain a first recognition result of whether the person is carrying a bag of a target type upon entry, and to process the image of the human body upon exit to obtain a second recognition result of whether the person is carrying a bag of the target type upon exit, where the target type is used to describe the type of bag provided by the merchant to carry the goods purchased by the customer; The quantity comparison submodule is used to process the human body image at the time of entry and the human body image at the time of exit to obtain a comparison result of whether the number of bags carried when leaving is more than the number of bags carried when entering; The result generation submodule is used to obtain an identification result indicating whether the same pedestrian has a purchasing behavior based on the first identification result, the second identification result and the comparison result.

4. The method according to claim 2, characterized in that The recognition model is trained in the following way: Constructing the recognition model, wherein the recognition model is provided with training parameters; Inputting multiple pairs of sample human body images into the recognition model respectively to generate prediction results; Based on the difference between the predicted result and the expected result corresponding to the sample label of each pair of sample human images, the training parameters are iteratively adjusted until the difference meets the preset requirement.

5. The method according to claim 1, wherein The determining, based on the corresponding entry time and exit time and the at least one transaction time, the number of target recognition results that do not have a corresponding transaction time in the at least one target recognition result includes: The following processing is performed on each transaction time in the at least one transaction time in order to obtain the number of target recognition results for which there is no corresponding transaction time in the at least one target recognition result: Determine whether there is a target recognition result whose current transaction time is between its corresponding entry time and exit time among the target recognition results that do not currently have a corresponding transaction time; If yes, then matching the current transaction time with the target recognition result; The next transaction time after the current transaction time is used as the current transaction time and the process returns to the judgment step until at least one transaction time is traversed.

6. The method according to claim 1, characterized in that The obtaining of the abnormality detection result corresponding to the target time period at the commodity display location based on the number of target recognition results for which no corresponding transaction time exists includes: Calculating first indicator data based on the number of target recognition results for which no corresponding transaction time exists, wherein the first indicator data is used to represent a recall rate of the recognized purchase behavior; Based on the obtained indicator data, an abnormality detection result of the commodity display location corresponding to the target time period is obtained, and the indicator data includes the first indicator data.

7. The method according to claim 6, characterized in that The transaction data also includes transaction volume; The calculating of the first indicator data based on the number of target recognition results that do not have corresponding transaction times includes: summing the transaction volume and the number of target recognition results for which no corresponding transaction time exists to obtain a summation result; The ratio of the number of the at least one target recognition result to the sum result is used as the first indicator data.

8. The method according to claim 7, characterized in that The customer flow data also includes the number of people entering; the number of target recognition results for which no corresponding transaction time exists, obtaining an abnormality detection result corresponding to the target time period at the commodity display location, further comprising: The ratio of the number of the at least one target recognition result to the number of people entering the system is used as the second indicator data; and / or the ratio of the transaction volume to the number of people entering the system is used as the third indicator data; The indicator data also includes the second indicator data and / or the third indicator data.

9. The method according to claim 6, characterized in that The method further comprises: Obtaining first transaction data of the product display location during a first time period and second transaction data of the product display location during a second time period; the first transaction data and the second transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the total number of transactions, or the number of commission-based transactions; Based on the first transaction data and the second transaction data, fourth indicator data for characterizing changes in the transaction data is calculated, and the indicator data also includes the fourth indicator data.

10. The method according to claim 6, characterized in that The method further comprises: Obtaining first transaction data of the product display location during a first time period and third transaction data of a product display location of the same business format during the first time period; the first transaction data and the third transaction data include one or more of the following: the number of refund transactions, the number of return and exchange transactions, the total number of transactions, or the number of commission-based transactions; Based on the first transaction data and the third transaction data, fifth indicator data for characterizing changes in transaction data is calculated, and the indicator data also includes the fifth indicator data.

11. The method according to any one of claims 6 to 10, characterized in that There are multiple indicator data, each indicator data has a corresponding detection threshold, and obtaining an abnormality detection result corresponding to the target time period at the product display location based on the obtained indicator data includes: Determine the comparison value corresponding to each indicator data based on the size relationship between each indicator data and its corresponding detection threshold; Calculate the total comparison value based on the comparison value corresponding to each indicator data; Based on the magnitude relationship between the total comparison value and the total threshold, an abnormality detection result of the commodity display location corresponding to the target time period is obtained.

12. The method according to any one of claims 1 to 10, characterized in that The method further includes: outputting warning prompt information, where the warning prompt information is used to prompt that there is an abnormal risk in the display of the product within the target time period.

13. The method according to claim 12, characterized in that There are multiple product display locations and multiple target time periods, and the alarm prompt information is specifically used to prompt the product display location of abnormal risks within the target time period from the time dimension and the product display location dimension; and / or, the product display location includes a shopping mall counter, and the alarm prompt information is also used to prompt the number of people entering the shopping mall where the product display location is located during the target time period.

14. An abnormality detection device, characterized in that: include: an acquisition module, configured to acquire passenger flow data and transaction data of a product display during a target period of time, wherein the passenger flow data includes a human body image of a pedestrian upon entering and leaving the product display, and corresponding entry and exit times, and the transaction data includes at least one transaction time; a recognition module configured to perform recognition processing based on the human body images of the same pedestrian when entering and when leaving, and obtain a recognition result indicating whether the same pedestrian has engaged in purchasing behavior; a first determining module configured to determine, based on an entry time and an exit time corresponding to at least one target recognition result indicating a purchase behavior and the at least one transaction time, a target recognition result for which no corresponding transaction time exists in the at least one target recognition result, to obtain a number of target recognition results for which no corresponding transaction time exists; The second determining module is configured to obtain an abnormality detection result corresponding to the target time period at the commodity display location based on the number of target recognition results that do not have a corresponding transaction time.

15. An electronic device, characterized in that: include: A memory, a processor; wherein the memory stores one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the method according to any one of claims 1 to 13 is implemented.

16. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, the method according to any one of claims 1 to 13 is implemented.

Images