Automatic inventory management method for vending machines based on image recognition

Abstract

The present invention relates to the technical field of inventory management of vending machines, and specifically to an automatic inventory management method for vending machines based on image recognition, which collects internal and external images of a vending machine in real time, extracts internal images of the vending machine to identify and analyze commodity types, generates a commodity display list, extracts sales number evaluation indexes and sales speed evaluation indexes corresponding to various categories of commodities at each monitoring time point, calculates a first sales evaluation value, extracts external images of the vending machine to identify and analyze consumer behavior, calculates a new consumption value, an old consumption value and a target consumption value ratio, determines a second sales evaluation value, combines the first sales evaluation value and the second sales evaluation value through normalization processing to obtain a total sales evaluation value of various categories of commodities, analyzes the change trend of the total sales evaluation value, determines the sales trend value of various categories of commodities, and adjusts inventory management parameters accordingly, thereby realizing real-time monitoring of the inventory status of commodities and improving the operating efficiency and profitability of vending machines.

Description

Technical Field

[0001] The present invention relates to the technical field of vending machine inventory management, and in particular to an automatic inventory management method for vending machines based on image recognition. Background Art

[0002] With the rapid development of science and technology and the continuous improvement of people's living standards, vending machines have gradually become an indispensable part of people's daily lives due to their convenience and efficiency. They are widely used to sell beverages, snacks, fruits and other commodities.

[0003] However, traditional vending machines have many deficiencies in inventory management and are unable to meet the intelligent management needs of the modern retail industry;

[0004] At present, traditional vending machines mainly rely on manual inspections and manual replenishment for inventory management, and lack real-time monitoring of commodity inventory status and sales. This approach has many problems. For example, when a commodity is out of stock, it fails to replenish it in time, resulting in the loss of sales opportunities and reduced consumer satisfaction. Too much replenishment may cause inventory backlogs and increase operating costs. Too little replenishment may lead to frequent stockouts, further affecting consumers' shopping experience and the profitability of the vending machine. In addition, traditional vending machine inventory management methods usually cannot make full use of consumer behavior data to optimize replenishment strategies, limiting the further improvement of vending machine operating efficiency.

[0005] In order to solve the above defects, a technical solution is now provided. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that traditional vending machines mainly rely on manual inspections and manual replenishment in inventory management, are unable to monitor the inventory status and sales of goods in real time, and are unable to make full use of consumer behavior data to optimize replenishment strategies. Instead, an automatic inventory management method for vending machines based on image recognition is proposed.

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

[0008] The method for automatic inventory management of vending machines based on image recognition comprises the following steps:

[0009] Input the monitoring area on the management end, automatically search for vending machines in the monitoring area, complete the monitoring layout of the vending machines, and transmit the collected images to the management end in real time;

[0010] Extract the internal image of the vending machine from the management end, identify and analyze the types of goods placed inside the vending machine, generate a product type display list, extract the sales number evaluation index and sales speed evaluation index corresponding to each type of goods at each monitoring time point from the product type display list, and determine the first sales evaluation value, where:

[0011] The sales number assessment index is calculated based on the combination of sales number and damage number;

[0012] The sales speed evaluation index is calculated based on the combination of sales quantity and initial sales quantity;

[0013] Extract the external image of the vending machine from the management end, identify and analyze the consumers outside the vending machine, generate a set of images of goods purchased by new consumers and a set of images of goods purchased by old consumers, monitor and analyze the external sales status information of the vending machine based on this, and obtain the new consumption value, old consumption value and target consumption value ratio of each type of goods at each monitoring time point, so as to determine the second sales evaluation value;

[0014] Extracting the first sales evaluation value and the second sales evaluation value and performing normalization processing to obtain the total sales evaluation value of each category of goods;

[0015] The total sales evaluation value of various categories of goods in the set monitoring period is analyzed to determine the sales trend value of various categories of goods, and an increase inventory signal or a decrease inventory signal is generated according to the sales trend value, thereby obtaining the inventory management parameters of various categories of goods.

[0016] Furthermore, the specific solution process for the first sales evaluation value is as follows:

[0017] Extract the sales evaluation index SMS of various types of goods at each monitoring time point from the product category display list s i and Selling Speed ​​Assessment Index SMP s i The value of is normalized according to the formula: , and obtain the first sales evaluation value δ1 of each type of commodity at each monitoring time point s i ;

[0018] Among them, various categories of goods specifically refer to different brands corresponding to different types of goods, s represents the label of various categories of goods, i represents the label of each monitoring time point, and i=1, 2, 3…m1, m1 represents the total number of labels of each monitoring time point, e represents a set natural constant, η1 and η2 represent the set weight coefficients respectively, and η1>η2.

[0019] Furthermore, the specific solution process for the sales volume evaluation index is as follows:

[0020] By obtaining the sales quantity M of various types of goods corresponding to each collection time 售卖 cs and the number of damages M 损坏 cs , and add the sold quantity and damaged quantity together. The specific expression is: , get the total sales quantity M of various types of goods corresponding to each collection time 总售 cs ;

[0021] According to the formula: , get the sales evaluation index SMS of various types of goods s , where c represents the number of each acquisition moment, and c=1, 2, 3…n, n represents the total number of acquisition moment numbers, M 总售 c-1s It represents the total sales quantity of various types of goods corresponding to the c-1th collection time, and a1 and a2 represent the set correction coefficients respectively.

[0022] Furthermore, the specific solution process for the sales quantity and the damaged quantity is as follows:

[0023] Total quantity actually sold + total quantity purchased out of stock = quantity sold, where the total quantity actually sold refers to the number of goods actually sold at a certain time of collection, and the total quantity purchased out of stock refers to the number of goods that users clicked to buy when the goods were out of stock;

[0024] Evenly arrange the detection points on the outer packaging of each commodity of each category, identify the chromaticity of each detection point on the outer packaging of each commodity of each category, and calculate the average value thereof to obtain the average chromaticity of each commodity of each category as the chromaticity average of each commodity of each category;

[0025] At the same time, the reference chromaticity corresponding to each category of goods is extracted from the database as the reference chromaticity value corresponding to each category of goods;

[0026] Determine the chromaticity evaluation index based on the chromaticity mean value and the reference chromaticity value;

[0027] Identify the apparent state parameters of each commodity corresponding to each category of commodities, including the damaged area, damaged length, dented area and number of dents;

[0028] The maximum damaged area, maximum damaged length, maximum dented area and number of dents are extracted from the apparent state parameters as the damaged face value, damaged length value, dented face value and dented value corresponding to each commodity of each category of commodities;

[0029] Extract reference damaged face value, reference damaged length value, reference dent face value and reference dent value value corresponding to various types of commodities from the database respectively;

[0030] Determine the apparent evaluation index based on the damaged face value, damaged length value, concave face value, concave value, reference damaged face value, reference damaged length value, reference concave face value and reference concave value;

[0031] Extract the production date and placement date of each commodity of each category from the database, and extract the date corresponding to the collection time at the same time, and subtract the production date of each commodity of each category from the date corresponding to the collection time to obtain the production time;

[0032] Subtract the placement date of each product of each category from the date corresponding to the collection time to obtain the placement time;

[0033] Extract the shelf life of each product of each category from the database to obtain the shelf life;

[0034] Determine the placement evaluation index based on the production time, storage time and shelf life;

[0035] Extract the values ​​of the color evaluation index, appearance evaluation index and placement evaluation index corresponding to each product of each category, perform normalization processing, and obtain the quality evaluation value;

[0036] Compare and analyze the quality assessment value of each commodity of each category with the preset quality assessment threshold. When the quality assessment value of a commodity of a certain category is greater than or equal to the preset quality assessment threshold, the commodity of this category is judged to be in a damaged state.

[0037] The number of commodities in each category that are judged to be damaged is counted, thereby obtaining the number of damaged commodities in each category.

[0038] Furthermore, the specific solution process for the sales speed evaluation index is as follows:

[0039] By obtaining the sales quantity M of various types of goods corresponding to each collection time 售卖 cs , and extract the sales quantity of each type of goods corresponding to the initial collection time as the initial sales quantity M 售卖 1s ;

[0040] According to the formula: , get the sales speed evaluation index SMP of various types of goods s , where ΔT （c·1） Indicates the duration between each acquisition moment and the initial acquisition moment, ΔT （c·c-1） It indicates the duration between adjacent acquisition moments, and a3 and a4 respectively indicate the set correction coefficients.

[0041] Furthermore, the specific solution process for the second sales evaluation value is as follows:

[0042] Extract the new consumption value xf of various types of goods corresponding to each monitoring time point outside the vending machine s i 、Old consumption value jfs i Ratio of target consumption to mb s i The value of is normalized according to the formula: , and obtain the second sales evaluation value δ2 corresponding to each type of commodity at each monitoring time point s i , where η3, η4 and η5 represent the set weight coefficients respectively, and η4>η3>η5.

[0043] Furthermore, the specific solution process for the new consumption value and the old consumption value is as follows:

[0044] Obtain a set of product images purchased by new consumers and a set of product images purchased by old consumers corresponding to each monitoring time point outside the vending machine;

[0045] By identifying and integrating the image set of goods purchased by new consumers, the sales quantity of various types of goods is obtained, which is used as the new consumption value xf corresponding to various types of goods at each monitoring time point outside the vending machine s i ;

[0046] By identifying and integrating the image collection of goods purchased by old consumers, the sales quantity of various types of goods is obtained, which is used as the old consumption value jf corresponding to various types of goods at each monitoring time point outside the vending machine s i .

[0047] Furthermore, the specific solution process for the target consumption ratio is as follows:

[0048] By identifying the ages of new and old consumers corresponding to various categories of goods;

[0049] Set multiple age ranges, match the ages of new and old consumers of various types of goods with the corresponding age ranges, perform statistical calculations on each age range, obtain the number of consumers in each age range, and identify the age range with the largest number of consumers as the target age range for various types of goods;

[0050] Taking the geographical location of the vending machine as the center point and the set radius R, a monitoring range is constructed and delineated. The personnel data within the monitoring range is extracted, and the number of people in the target age range corresponding to each type of commodity and the total number of people within the monitoring range are counted respectively. The number of people in the target age range corresponding to each type of commodity and the total number of people are calculated to calculate the ratio of the target number of people for each type of commodity within the monitoring range, which is used as the target consumption number ratio mb of each type of commodity at each monitoring time point outside the vending machine s i .

[0051] Furthermore, the specific process of solving the sales trend value of various types of goods is as follows:

[0052] Obtain the total sales evaluation value of each category of goods corresponding to each monitoring time point in the set monitoring period of the vending machine, take the total sales evaluation value of each category of goods as the ordinate and the set monitoring period as the abscissa, and establish a two-dimensional dynamic coordinate system of total sales of various categories of goods, and plot the total sales evaluation value of each category of goods corresponding to each monitoring time point on the two-dimensional dynamic coordinate system of total sales of various categories of goods by plotting points, and obtain a total sales line chart of various categories of goods;

[0053] A total sales reference line for various categories of goods parallel to the horizontal axis is set in the two-dimensional dynamic coordinate system of total sales of various categories of goods. The total sales evaluation value of various categories of goods corresponding to each monitoring time point is compared and analyzed with the corresponding total sales reference line. When the total sales evaluation value of a certain category of goods is greater than or equal to the corresponding total sales reference line, the sales status of this type of goods is determined to be a sales growth state. The number of times various categories of goods are determined to be in a sales growth state is counted to obtain the number of sales growth determinations for various categories of goods. The sales growth determination number is calculated as a percentage of the total determination number to obtain the sales trend value of various categories of goods.

[0054] Furthermore, the specific solution process for inventory management parameters of various types of goods is as follows:

[0055] Compare and analyze the sales trend value of each category of goods with the preset sales trend threshold value. When the sales trend value of each category of goods is greater than or equal to the preset sales trend threshold value, an increase inventory signal is generated. Otherwise, a decrease inventory signal is generated.

[0056] If an inventory increase signal is captured, the sales trend values ​​of various categories of goods are retrieved and matched and analyzed with the stored sales trend increase status table to obtain inventory management parameters of various categories of goods;

[0057] If a signal for reducing inventory is captured, the sales trend values ​​of various categories of goods are retrieved, and matched and analyzed with the stored sales trend reduction status table to obtain inventory management parameters for various categories of goods;

[0058] Inventory management parameters include replenishment quantity and replenishment countdown length or reduction quantity and reduction countdown length.

[0059] Compared with the known prior art, the technical solution provided by the present invention has the following beneficial effects:

[0060] 1. The present invention calculates the total sales quantity by obtaining the sales quantity and damaged quantity of various types of goods corresponding to each collection moment, and uses formulas to respectively obtain the sales number evaluation index and the sales speed evaluation index. The former reflects the sales and damage of the goods, and the latter reveals the sales speed trend. Subsequently, by analyzing the color evaluation index, appearance evaluation index and placement evaluation index of the goods, the quality evaluation value is calculated to determine whether the goods are damaged. The two evaluation indexes are combined for normalization processing to obtain the first sales evaluation value, thereby achieving a comprehensive reflection of the overall sales performance of the goods, providing data support for the management end to grasp the sales and inventory dynamics of the vending machine in real time, and helping to optimize inventory management and adjust sales strategies.

[0061] 2. The present invention obtains new consumption values ​​and old consumption values ​​corresponding to various types of commodities at each monitoring time point by identifying and integrating the commodity images purchased by new consumers and old consumers, intuitively reflects the purchase preferences and consumption capabilities of different groups, further analyzes the age characteristics of consumers, statistically distributes each age range, and determines the target age range, so as to accurately locate the target consumer group and provide support for marketing strategies. At the same time, a monitoring range is constructed with the geographical location of the vending machine as the center, and the proportion of people in the target age range is statistically calculated to calculate the target consumption number ratio, and then the new consumption value, the old consumption value and the target consumption number ratio are combined to calculate the second sales evaluation value, thereby achieving data support for optimizing inventory management and precise marketing strategies.

[0062] 3. The present invention obtains sales trend values ​​of various categories of goods in vending machines by analyzing the sales trends of various categories of goods in the vending machines, and compares them with preset sales trend thresholds to generate corresponding inventory signals. Corresponding inventory management parameters are matched according to these signals, thereby realizing accurate monitoring of sales trends of goods in vending machines and intelligent inventory management, which helps to improve the operating efficiency of vending machines and enhance their profitability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0064] Figure 1 The figure is a flow chart of the overall method of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] like Figure 1 As shown, the automatic inventory management method for vending machines based on image recognition includes:

[0067] Step 1: Analyze the monitoring layout of the vending machine. The specific analysis process is as follows:

[0068] Enter the monitoring area on the management end, automatically search for vending machines in the monitoring area, and transmit the geographical location information of the vending machines in the monitoring area to the management end. Then, the management end numbers the vending machines according to the entry time. The numbering rule is: entry date - annual sequence number. For example, if a vending machine is entered on December 11, 2024 and is the ninth vending machine entered in that year, its number is 20241211-009;

[0069] Install a high-resolution camera on each vending machine in the monitoring area to ensure that the camera's field of view covers the internal and external areas of the vending machine, collect internal and external images of the vending machine, and transmit the collected images to the management end in real time;

[0070] It should be pointed out that, for internal image acquisition, the camera regularly captures images inside the vending machine at a preset acquisition frequency T, and for external image acquisition, an event trigger mechanism is used. For example, when the vending machine door is opened or the user stays in front of the vending machine door for more than t time, the camera will automatically capture external images.

[0071] Step 2: Identify and analyze the types of goods placed inside the vending machine. The specific analysis process is as follows:

[0072] The internal image of the vending machine is extracted from the management end, and the corresponding product image is extracted from each internal image. Each product image is subjected to feature matching analysis with a pre-stored product type library to generate a display list of product types in each internal image of the vending machine, which also includes but is not limited to the product name, category, quantity, distribution location, and placement date.

[0073] Step 3: Monitor and analyze the internal sales status information of the vending machine. The specific analysis process is as follows:

[0074] By obtaining the commodity display list corresponding to each monitoring time point in the vending machine, and extracting the sales evaluation index SMS of each type of commodity corresponding to each monitoring time point s i and Selling Speed ​​Assessment Index SMP s i The value of is normalized according to the formula: , and obtain the first sales evaluation value δ1 of each type of commodity at each monitoring time point s i ;

[0075] Among them, various types of goods specifically refer to different brands corresponding to different types of goods, s represents the number of various types of goods, i represents the number of each monitoring time point, and i=1, 2, 3…m1, m1 represents the total number of numbers of each monitoring time point, e represents a set natural constant, η1, η2 represent set weight coefficients respectively, and η1>η2;

[0076] It should be noted that the commodity type display list table refers to the table obtained by comprehensively sorting out the commodity type display information in all internal images obtained at each monitoring time point of the vending machine. Specifically, this table is obtained by carefully processing the internal images collected at each monitoring time point, extracting the commodity information in each image, and generating a corresponding commodity type display list based on this. Afterwards, these commodity type display lists are summarized according to the collection time to form a commodity type display list table corresponding to each monitoring time point in the vending machine, which reflects the commodity display situation of the vending machine at different monitoring time points and is the basis for subsequent analysis.

[0077] In the above, the specific solution and analysis process for the sales volume evaluation index of various types of goods is as follows:

[0078] By obtaining the sales quantity and damage quantity of various types of goods corresponding to each collection time, the sales quantity M of various types of goods corresponding to each collection time is obtained. 售卖 cs and the number of damages M 损坏 cs , where sales quantity = actual total sales quantity + total purchase out-of-stock quantity, and the sales quantity and damaged quantity are added together. The specific expression is: , get the total sales quantity M of various types of goods corresponding to each collection time 总售 cs ;

[0079] It should be noted that the actual total quantity sold refers to the number of goods actually sold at a certain time of collection, and the total number of out-of-stock purchases refers to the number of goods that users click to buy when the goods are out of stock (even though there is no stock, users still choose to buy), which also needs to be included in the sales quantity;

[0080] According to the formula: , get the sales evaluation index SMS of various types of goods s , where c represents the number of each acquisition moment, and c=1, 2, 3…n, n represents the total number of acquisition moment numbers, M 总售 c-1s It represents the total sales quantity of various types of goods corresponding to the c-1th collection time, and a1 and a2 represent the set correction coefficients respectively.

[0081] Preferably, the specific determination and analysis of the damaged quantity of each type of goods is as follows:

[0082] The detection points are evenly arranged on the outer packaging of each commodity of each category, and the chromaticity of each detection point on the outer packaging of each commodity of each category is detected through image recognition to obtain the chromaticity of each detection point on the outer packaging of each commodity of each category, and the average value is calculated to obtain the average chromaticity of each commodity of each category as the chromaticity mean value sd of each commodity of each category s g , where g represents the label of each commodity;

[0083] At the same time, the reference chromaticity corresponding to each category of goods is extracted from the database as the reference chromaticity value sd corresponding to each category of goods s g* ;

[0084] According to the formula: , and obtain the chromaticity evaluation index θ1 of each product of each category s g , where Δsd s It is represented by the allowable color difference corresponding to the s-th type of product stored in the database;

[0085] Extract the appearance image of each commodity corresponding to each category of commodities, extract the appearance state parameters of each commodity corresponding to each category of commodities through image recognition, and obtain the appearance state parameters of each commodity corresponding to each category of commodities, the appearance state parameters including the damaged area, damaged length, dent area and dent quantity;

[0086] Extract the maximum damaged area from the apparent state parameters of each commodity in each category as the damaged face value pm of each commodity in each category. s g ;

[0087] Extract the maximum damaged length from the apparent state parameters of each commodity in each category as the damaged length value pc of each commodity in each category s g ;

[0088] Extract the maximum concave area from the apparent state parameters of each commodity in each category as the concave face value am of each commodity in each category s g ;

[0089] Extract the number of depressions from the apparent state parameters of each commodity in each category as the depression value of each commodity in each category as s g ;

[0090] Extract the reference damaged face value, reference damaged length value, reference concave face value and reference concave value value of each category of goods from the database respectively, and use them as the reference damaged face value pm of each category of goods corresponding to each product. s g* , refer to the damaged length value pc s g* , reference concave surface value am s g* and the reference depression value as s g* ;

[0091] According to the formula: , and obtain the apparent evaluation index θ2 of each commodity corresponding to each category of commodities s g , where Δpm s , Δpc s , Δam s and Δas s They respectively represent the allowable damaged face difference, allowable damaged length difference, allowable concave face difference and allowable concave number difference corresponding to the s-th type of product stored in the database, and p1, p2, p3 and p4 represent the set weight factors respectively;

[0092] Extract the production date and placement date of each commodity from the database, and extract the date corresponding to the collection time. Subtract the production date of each commodity from the collection time to get the output time yc. s g ; Subtract the placement date of each product from the date corresponding to the collection time to obtain the placement time fc s g ;

[0093] Extract the shelf life of each product from the database and get the shelf life duration bc s g ;

[0094] According to the formula: , and obtain the placement evaluation index θ3 of each type of product corresponding to each product s g , where p5 and p6 represent the set weight factors respectively;

[0095] Extract the chromaticity evaluation index θ1 corresponding to each product of each category s g , Apparent evaluation index θ2 s g and placement evaluation index θ3 s g The value of is normalized according to the formula: , and obtain the quality assessment value ZLP s g , where λ1, λ2 and λ3 represent the weight coefficients of the chromaticity evaluation index, the appearance evaluation index and the placement evaluation index, respectively, and λ1>λ2>λ3;

[0096] Compare and analyze the quality assessment value of each commodity corresponding to each category of commodities with the preset quality assessment threshold. When the quality assessment value of a commodity corresponding to a certain category of commodities is greater than or equal to the preset quality assessment threshold, the commodity corresponding to the category of commodities is judged to be in a damaged state. Otherwise, the commodity corresponding to the category of commodities is judged to be in a normal state.

[0097] The number of commodities in each category that are judged to be damaged is counted, thereby obtaining the number of damaged commodities in each category.

[0098] In the above, the specific solution and analysis process for the sales speed evaluation index of various types of goods is as follows:

[0099] By obtaining the sales quantity M of various types of goods corresponding to each collection time 售卖 cs , and extract the sales quantity of each type of goods corresponding to the initial collection time as the initial sales quantity M 售卖 1s ;

[0100] According to the formula: , get the sales speed evaluation index SMP of various types of goods s , where ΔT （c·1） Indicates the duration between each acquisition moment and the initial acquisition moment, ΔT （c·c-1） It indicates the duration between adjacent acquisition moments, and a3 and a4 respectively indicate the set correction coefficients.

[0101] Step 4: Identify and analyze consumers outside the vending machine. The specific analysis process is as follows:

[0102] Extract the external image of the vending machine from the management end, and extract the facial image corresponding to each consumer from each external image, perform feature matching analysis on the facial image of each consumer and the pre-stored facial information database, obtain the facial image of each new consumer and the facial image of each old consumer, and then extract the image of the goods purchased by each new consumer and the image of the goods purchased by each old consumer, respectively integrate the image of the goods purchased by each new consumer and the image of the goods purchased by each old consumer, and obtain the image set of the goods purchased by the new consumer and the image set of the goods purchased by the old consumer;

[0103] Among them, the facial information database contains the facial feature data of all consumers who have made purchases at the vending machine. These facial feature data are stored by the facial recognition system when the consumer makes a purchase for the first time and are used for subsequent identity recognition and behavior analysis.

[0104] Step 5: Monitor and analyze the external sales status information of the vending machine. The specific analysis process is as follows:

[0105] Obtain a set of product images purchased by new consumers and a set of product images purchased by old consumers corresponding to each monitoring time point outside the vending machine;

[0106] By identifying and integrating the image set of goods purchased by new consumers, the sales quantity of various types of goods is obtained, which is used as the new consumption value xf corresponding to various types of goods at each monitoring time point outside the vending machine s i ;

[0107] By identifying and integrating the image collection of goods purchased by old consumers, the sales quantity of various types of goods is obtained, which is used as the old consumption value jf corresponding to various types of goods at each monitoring time point outside the vending machine s i ;

[0108] By identifying the facial images of new consumers corresponding to various categories of goods, and extracting the ages of new consumers corresponding to various categories of goods;

[0109] By identifying the facial images of old consumers corresponding to various types of goods, and extracting the ages of old consumers corresponding to various types of goods;

[0110] Set multiple age ranges, match the ages of new and old consumers of various types of goods with the corresponding age ranges, perform statistical calculations on each age range, obtain the number of consumers in each age range, and identify the age range with the largest number of consumers as the target age range for various types of goods;

[0111] Taking the geographical location of the vending machine as the center point and the set radius R, a monitoring range is constructed and delineated. The personnel data within the monitoring range is extracted, and the number of people in the target age range corresponding to each type of commodity and the total number of people within the monitoring range are counted respectively. The number of people in the target age range corresponding to each type of commodity and the total number of people are calculated to calculate the ratio of the target number of people for each type of commodity within the monitoring range, which is used as the target consumption number ratio mb of each type of commodity at each monitoring time point outside the vending machine s i ;

[0112] Extract the new consumption value xf of various types of goods corresponding to each monitoring time point outside the vending machine s i 、Old consumption value jf s i Ratio of target consumption to mb s i The value of is normalized according to the formula: , and obtain the second sales evaluation value δ2 corresponding to each type of commodity at each monitoring time point s i , where η3, η4 and η5 represent the set weight coefficients respectively, and η4>η3>η5.

[0113] Step 6: Analyze the sales trend of various types of goods in the vending machine. The specific analysis process is as follows:

[0114] By extracting the first sales evaluation value δ1 corresponding to each type of commodity at each monitoring time point s i and the second sales evaluation value δ2 s i The value of is normalized according to the formula: , and obtain the total sales evaluation value XSZ of various types of goods at each monitoring time point s i , where η6 and η7 represent the set weight coefficients respectively, and η6>η7;

[0115] By acquiring the total sales evaluation values ​​of various categories of goods corresponding to each monitoring time point of the vending machine in the set monitoring period, the total sales evaluation values ​​of various categories of goods corresponding to each monitoring time point of the vending machine in the set monitoring period are obtained, and the total sales evaluation values ​​of various categories of goods corresponding to each monitoring time point in the set monitoring period are used as the vertical coordinates, and the set monitoring period is used as the horizontal coordinates, and a two-dimensional dynamic coordinate system of total sales of various categories of goods is established accordingly, and the total sales evaluation values ​​of various categories of goods corresponding to each monitoring time point in the set monitoring period are plotted on the two-dimensional dynamic coordinate system of total sales of various categories of goods by plotting points, so as to obtain a total sales line chart of various categories of goods;

[0116] A total sales reference line of various categories of goods parallel to the horizontal axis is set on the two-dimensional total sales dynamic coordinate system of various categories of goods, and the total sales evaluation value of various categories of goods corresponding to each monitoring time point in the set monitoring period is compared and analyzed with the corresponding total sales reference line. When the total sales evaluation value of a certain category of goods is greater than or equal to the corresponding total sales reference line, the sales status of the category of goods is determined to be a sales growth state, otherwise, the sales status of the category of goods is determined to be a sales reduction state, and the number of times that various categories of goods are determined to be in a sales growth state is counted to obtain the number of sales growth determinations of various categories of goods, and the sales growth determination number is calculated as a proportion of the total determination number to obtain the sales trend value of various categories of goods;

[0117] The sales trend values ​​of various categories of goods are compared and analyzed with the preset sales trend thresholds. When the sales trend values ​​of various categories of goods are greater than or equal to the preset sales trend thresholds, an increase inventory signal is generated. When the sales trend values ​​of various categories of goods are less than the preset sales trend thresholds, a decrease inventory signal is generated.

[0118] Step 7: Manage the inventory of various types of goods in the vending machine. The specific process is as follows:

[0119] If an inventory increase signal is captured, the sales trend values ​​of various categories of goods are retrieved and matched with the sales trend increase status table stored in the cloud database to obtain the sales trend increase level of various categories of goods. Each sales trend value has a sales trend increase level corresponding to it. At the same time, it is matched with the inventory management parameters corresponding to the sales trend increase level to obtain the inventory management parameters of various categories of goods. The inventory management parameters include the replenishment quantity and the replenishment countdown length.

[0120] If a signal for reducing inventory is captured, the sales trend values ​​of various types of goods are retrieved and matched with the sales trend reduction status table stored in the cloud database to obtain the sales trend reduction level of various types of goods. Each sales trend value has a corresponding sales trend reduction level. At the same time, the sales trend value is matched with the inventory management parameters corresponding to the sales trend reduction level to obtain the inventory management parameters of various types of goods. The inventory management parameters include the quantity of goods reduced and the length of the countdown time for goods reduced.

[0121] Thus, the inventory management parameters of various types of goods in the vending machine and the number of the vending machine are sent to the display terminal of the manager for display reminder.

[0122] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An automatic inventory management method for vending machines based on image recognition, characterized in that: The following steps are involved: Input the monitoring area on the management end, automatically search for vending machines in the monitoring area, complete the monitoring layout of the vending machines, and transmit the collected images to the management end in real time; Extract the internal image of the vending machine from the management end, identify and analyze the types of goods placed inside the vending machine, generate a product type display list, extract the sales number evaluation index and sales speed evaluation index corresponding to each type of goods at each monitoring time point from the product type display list, and determine the first sales evaluation value, where: The sales number assessment index is calculated based on the combination of sales number and damage number; The specific process of solving the sales quantity and damaged quantity is as follows: Total quantity actually sold + total quantity purchased out of stock = quantity sold, where the total quantity actually sold refers to the number of goods actually sold at a certain time of collection, and the total quantity purchased out of stock refers to the number of goods that users clicked to buy when the goods were out of stock; Evenly arrange the detection points on the outer packaging of each commodity of each category, identify the chromaticity of each detection point, and calculate the average value to obtain the average chromaticity as the chromaticity average of each commodity of each category, and determine the chromaticity evaluation index by combining the chromaticity average and the reference chromaticity value; Identify the apparent state parameters of each commodity of each category, and extract the maximum damaged area, damaged length, dented area and dented number as the damaged face value, damaged length value, dented face value and dented number, and determine the apparent evaluation index in combination with the corresponding reference value; Extract the production date, placement date and shelf life of each product of each category, and extract the date corresponding to the collection time at the same time. Subtract the production date and placement date from the date corresponding to the collection time, respectively, to obtain the production time and placement time, and determine the placement evaluation index accordingly. Determine the quality assessment value according to the color assessment index, appearance assessment index and placement assessment index of each product of each category of products, and determine the damage status based on the quality assessment value; Count the number of goods in each category that are judged to be damaged, thereby obtaining the number of damaged goods in each category; Extract the external image of the vending machine from the management end, identify and analyze the consumers outside the vending machine, generate a set of images of goods purchased by new consumers and a set of images of goods purchased by old consumers, monitor and analyze the external sales status information of the vending machine based on this, and obtain the new consumption value, old consumption value and target consumption value ratio of each type of goods at each monitoring time point, so as to determine the second sales evaluation value; Extracting the first sales evaluation value and the second sales evaluation value and performing normalization processing to obtain the total sales evaluation value of each category of goods; The total sales evaluation value of various categories of goods in the set monitoring period is analyzed to determine the sales trend value of various categories of goods, and an increase inventory signal or a decrease inventory signal is generated according to the sales trend value, thereby obtaining the inventory management parameters of various categories of goods.

2. The method for automatic inventory management of vending machines based on image recognition according to claim 1, characterized in that: The specific solution process for the first sales evaluation value is as follows: Extract the sales evaluation index SMS of various types of goods at each monitoring time point from the product category display list s i and Selling Speed ​​Assessment Index SMP s i The value of is normalized according to the formula: , and obtain the first sales evaluation value δ1 of each type of commodity at each monitoring time point s i ; Among them, various categories of goods specifically refer to different brands corresponding to different types of goods, s represents the label of various categories of goods, i represents the label of each monitoring time point, and i=1, 2, 3…m1, m1 represents the total number of labels of each monitoring time point, e represents a set natural constant, η1 and η2 represent the set weight coefficients respectively, and η1>η2.

3. The method for automatic inventory management of vending machines based on image recognition according to claim 2, characterized in that: The specific solution process for the sales volume evaluation index is as follows: By obtaining the sales quantity M of various types of goods corresponding to each collection time 售卖 cs and the number of damages M 损坏 cs , and add the sold quantity and damaged quantity together. The specific expression is: , get the total sales quantity M of various types of goods corresponding to each collection time 总售 cs ; According to the formula: , get the sales evaluation index SMS of various types of goods s , where c represents the number of each acquisition moment, and c=1, 2, 3…n, n represents the total number of acquisition moment numbers, M 总售 c-1s It represents the total sales quantity of various types of goods corresponding to the c-1th collection time, and a1 and a2 represent the set correction coefficients respectively.

4. The method for automatic inventory management of vending machines based on image recognition according to claim 3, characterized in that: The specific solution process for the sales speed evaluation index is as follows: By obtaining the sales quantity M of various types of goods corresponding to each collection time 售卖 cs , and extract the sales quantity of each type of goods corresponding to the initial collection time as the initial sales quantity M 售卖 1s ; According to the formula: , get the sales speed evaluation index SMP of various types of goods s , where ΔT （c·1） Indicates the duration between each acquisition moment and the initial acquisition moment, ΔT （c·c-1） It indicates the duration between adjacent acquisition moments, and a3 and a4 respectively indicate the set correction coefficients.

5. The method for automatic inventory management of vending machines based on image recognition according to claim 1, characterized in that: The specific solution process for the second sales evaluation value is as follows: Extract the new consumption value xf of various types of goods corresponding to each monitoring time point outside the vending machine s i 、Old consumption value jf s i Ratio of target consumption to mb s i The value of is normalized according to the formula: , and obtain the second sales evaluation value δ2 corresponding to each type of commodity at each monitoring time point s i , where η3, η4 and η5 represent the set weight coefficients respectively, and η4>η3>η5.

6. The method for automatic inventory management of vending machines based on image recognition according to claim 5, characterized in that: The specific solution process for the new consumption value and the old consumption value is as follows: Obtain a set of product images purchased by new consumers and a set of product images purchased by old consumers corresponding to each monitoring time point outside the vending machine; By identifying and integrating the image set of goods purchased by new consumers, the sales quantity of various types of goods is obtained, which is used as the new consumption value xf corresponding to various types of goods at each monitoring time point outside the vending machine s i ; By identifying and integrating the image collection of goods purchased by old consumers, the sales quantity of various types of goods is obtained, which is used as the old consumption value jf corresponding to various types of goods at each monitoring time point outside the vending machine s i .

7. The method for automatic inventory management of vending machines based on image recognition according to claim 5, characterized in that: The specific solution process for the target consumption ratio is as follows: By identifying the ages of new and old consumers corresponding to various categories of goods; Set multiple age ranges, match the ages of new and old consumers of various types of goods with the corresponding age ranges, perform statistical calculations on each age range, obtain the number of consumers in each age range, and identify the age range with the largest number of consumers as the target age range for various types of goods; Taking the geographical location of the vending machine as the center point and the set radius R, a monitoring range is constructed and delineated. The personnel data within the monitoring range is extracted, and the number of people in the target age range corresponding to each type of commodity and the total number of people within the monitoring range are counted respectively. The number of people in the target age range corresponding to each type of commodity and the total number of people are calculated to calculate the ratio of the target number of people for each type of commodity within the monitoring range, which is used as the target consumption number ratio mb of each type of commodity at each monitoring time point outside the vending machine s i .

8. The method for automatic inventory management of vending machines based on image recognition according to claim 1, characterized in that: The specific process of solving the sales trend value of various types of commodities is as follows: Obtain the total sales evaluation value of each category of goods corresponding to each monitoring time point in the set monitoring period of the vending machine, take the total sales evaluation value of each category of goods as the ordinate and the set monitoring period as the abscissa, and establish a two-dimensional dynamic coordinate system of total sales of various categories of goods, and plot the total sales evaluation value of each category of goods corresponding to each monitoring time point on the two-dimensional dynamic coordinate system of total sales of various categories of goods by plotting points, and obtain a total sales line chart of various categories of goods; A total sales reference line for various categories of goods parallel to the horizontal axis is set in the two-dimensional dynamic coordinate system of total sales of various categories of goods. The total sales evaluation value of various categories of goods corresponding to each monitoring time point is compared and analyzed with the corresponding total sales reference line. When the total sales evaluation value of a certain category of goods is greater than or equal to the corresponding total sales reference line, the sales status of this type of goods is determined to be a sales growth state. The number of times various categories of goods are determined to be in a sales growth state is counted to obtain the number of sales growth determinations for various categories of goods. The sales growth determination number is calculated as a percentage of the total determination number to obtain the sales trend value of various categories of goods.

9. The method for automatic inventory management of vending machines based on image recognition according to claim 1, characterized in that: The specific solution process for inventory management parameters of various types of goods is as follows: Compare and analyze the sales trend value of each category of goods with the preset sales trend threshold value. When the sales trend value of each category of goods is greater than or equal to the preset sales trend threshold value, an increase inventory signal is generated. Otherwise, a decrease inventory signal is generated. If an inventory increase signal is captured, the sales trend values ​​of various categories of goods are retrieved and matched and analyzed with the stored sales trend increase status table to obtain inventory management parameters of various categories of goods; If a signal for reducing inventory is captured, the sales trend values ​​of various categories of goods are retrieved, and matched and analyzed with the stored sales trend reduction status table to obtain inventory management parameters for various categories of goods; Inventory management parameters include replenishment quantity and replenishment countdown length or reduction quantity and reduction countdown length.

Images