A method and system for reporting and statistical analysis of consumables

Abstract

The present invention relates to the technical field of electronic digital data processing, and discloses a method and system for reporting statistics of consumables, including: extracting video entry and exit records of consumables from on-site videos, identifying text entry and exit records of consumables from text records, and extracting voice entry and exit records of consumables from voice records; querying display interest points of simplified reports, calculating interest point indicators of simplified reports, and analyzing the statistical balance of simplified reports using interest point indicators; when there are new products of consumables, inserting the entry and exit data of new products into optimized reports to obtain updated reports, extracting user outbound data of historical users of consumables from updated reports; using user outbound data to identify the outbound demand of historical users for consumables, constructing personalized reports of historical users through updated reports, and using personalized reports as statistical results of consumable reports. The present invention can improve the efficiency of reporting statistics for consumables.

Description

Technical Field

[0001] The invention relates to a report form statistics method and system for consumables, belonging to the technical field of electronic digital data processing. Background Art

[0002] Consumables are items that are used frequently in daily production, operations or life, have a relatively short lifespan and need to be replaced frequently. For example, in an office environment, pens, papers, ink cartridges, etc. are all consumables. In industrial production, cutting tools, sandpaper, cleaning supplies, etc. are also consumables. Their characteristics are relatively low value, but fast consumption rate, and the need for continuous replenishment. Compared with fixed assets, consumables do not have long-term use value and the value transfer method is different. Fixed assets usually have a longer service life, and their value is gradually transferred to products or services through depreciation and other methods, while the value of consumables is consumed all at once in a relatively short period of time and is directly included in the current cost or expense.

[0003] Traditional consumables management mainly relies on manual records. In enterprises, warehouse managers may use paper account books to record the entry and exit of consumables, including information such as date, item name, quantity, and user, and conduct manual inventory regularly to check whether the actual inventory is consistent with the account book records. This method is inefficient and prone to human errors, such as unclear handwriting and calculation errors, and it is difficult to grasp the inventory of consumables in real time. The purpose of reporting and statistics on consumables is, on the one hand, to accurately record the entry and exit data of consumables, and on the other hand, to count the items that are easy to consume, popular items, and unpopular items among consumables, so that enterprises can improve their products in the future. In actual scenarios, the popularity of consumables is often identified by the consumption speed of consumables and whether they are selling well. However, due to the different types of consumables, it is not advisable to select popular consumables only by the sales speed. For example, toothbrushes that are used for a long time and paper towels that need to be replaced frequently, the former are purchased less frequently, and the latter are purchased more frequently. Therefore, the existing technology is not efficient enough in reporting and statistics on consumables. Summary of the invention

[0004] The present invention provides a method and system for reporting statistics of consumables, the main purpose of which is to improve the efficiency of reporting statistics for consumables.

[0005] To achieve the above purpose, the present invention provides a method for compiling reports of consumables, comprising:

[0006] Collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records, and voice records;

[0007] Extracting the video entry and exit records of the consumables from the on-site video, identifying the text entry and exit records of the consumables from the text records, extracting the voice entry and exit records of the consumables from the voice records, and generating standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records;

[0008] Identifying the report relevance between each standard report in the standard report, clustering the standard reports according to the report relevance to obtain clustered reports, and simplifying the clustered reports to obtain simplified reports;

[0009] Querying the display interest points of the simplified report, calculating the interest point index of the simplified report according to the display interest points, analyzing the statistical balance of the simplified report by using the interest point index, and optimizing the simplified report by using the statistical balance to obtain an optimized report;

[0010] Determine whether there are new products of the consumables;

[0011] When there is a new product of the consumables, the in-and-out data of the new product is inserted into the optimization report to obtain an update report, and the user out-and-out data of the historical users of the consumables are extracted from the update report;

[0012] The user outbound data is used to identify the outbound demand of the historical user for the consumables, and according to the outbound demand, a personalized report of the historical user is constructed through the update report, and the personalized report is used as the report statistical result of the consumables.

[0013] Optionally, extracting the video entry and exit record of the consumables from the on-site video includes:

[0014] Distinguish between the purchase period and non-purchase period of the consumables;

[0015] During the non-restocking period, collecting the first personnel frequency of the consumables from the on-site video;

[0016] When the first personnel frequency is greater than a first preset frequency, identifying the employee identity corresponding to the consumable;

[0017] When the frequency of the first person is not greater than a first preset frequency, identifying the user identity corresponding to the consumable;

[0018] During the purchase period, collecting the second personnel frequency of the consumables from the on-site video;

[0019] When the second personnel frequency is greater than a second preset frequency, identifying the non-employee identity corresponding to the consumable according to the employee identity;

[0020] Marking the persons in the on-site video by using the employee identity, the user identity and the non-employee identity to obtain marked identity information;

[0021] Extracting voice data of people in the live video;

[0022] Converting the voice data into conversational voice using the marked identity information and the time of the voice data;

[0023] Performing speech preprocessing on the conversation speech to obtain preprocessed speech;

[0024] Converting the preprocessed speech into a feature vector;

[0025] Converting the feature vector into a dialogue text using a preset hidden Markov model;

[0026] Identifying the personnel intention corresponding to the consumables from the conversation text;

[0027] The following formula is used to analyze the probability of the physical behavior of the personnel in the live video:

[0028] ;

[0029] in, represents the probability of physical behavior, represents the joint point confidence map, represents the joint affinity field, represents the convolutional layer, represents a deep convolutional neural network, An image representing a person in a live video feed, represents the activation function;

[0030] identifying the physical behavior of the person using the physical behavior probability;

[0031] Identify the person's behavioral intention from the physical behavior;

[0032] The intention of the person and the intention of the behavior Figure 1 When the person's intention and the behavior's intention are confirmed, the video entry and exit records of the consumables are recorded;

[0033] When the personnel intention is inconsistent with the behavior intention, an intention review is performed on the personnel intention and the behavior intention to obtain an intention review result;

[0034] The intent audit result is used to determine the video access record as the consumable item.

[0035] Optionally, the identifying the personnel intention corresponding to the consumables from the conversation text includes:

[0036] Performing position encoding on the dialogue text to obtain a position vector;

[0037] splicing the position vector and the dialogue text to obtain a splicing vector;

[0038] The attention feature of the concatenated vector is extracted using the following formula:

[0039] ;

[0040] in, Represents the attention feature, represents the multi-head attention mechanism, , , The query vector, key vector and key vector representing the calculated concatenated vector, , , express The weight of , , express The weight of , , express The weight of represents the total number of dimensions of the concatenated vector, express The splicing weight of

[0041] The attention feature is used to identify the personnel intention corresponding to the consumables.

[0042] Optionally, the identifying the behavioral intention of a person from the body behavior includes:

[0043] Get live video;

[0044] Performing target detection on the live video to obtain a detection target;

[0045] constructing behavior description information of the person using the detection target and the body behavior;

[0046] The behavior description information is classified by intent to obtain the behavior intention.

[0047] Optionally, querying the display interest points of the simplified report includes:

[0048] Querying the scheduled sequence of the simplified report;

[0049] generating a scheduling graph structure of the scheduled sequence;

[0050] According to the scheduled sequence and the scheduling graph structure, the scheduling density of the simplified report is calculated using the following formula:

[0051] ;

[0052] in, Indicates the scheduling density of the u-th report data in the simplified report. Indicates the bit sequence of the u-th report data in the j-th scheduled sequence, Indicates the total order, express The number of neighboring report data of the u-th report data in the scheduling graph structure under the position order, Indicates the number of scheduled sequences;

[0053] When the scheduled density is greater than a preset density, the report data corresponding to the scheduled density is used as a display interest point of a simplified report.

[0054] Optionally, the calculating the interest point index of the simplified report according to the displayed interest points includes:

[0055] Classifying the displayed interest points to obtain classified interest points;

[0056] Collecting the number of points of interest and the number of categories of points of interest corresponding to the classified points of interest;

[0057] The number of points of interest and the number of categories of points of interest are used as the points of interest indicators of the simplified report.

[0058] Optionally, analyzing the statistical balance of the simplified report using the interest point index includes:

[0059] Obtaining the number of points of interest and the number of categories of points of interest in the point of interest index;

[0060] The number balance of the number of interest points is calculated using the following formula:

[0061] ;

[0062] in, Indicates the balance of numbers. represents the number of interest points of the vth classification interest point, express The mean number of interest points of each classification point of interest, Indicates the number of classified interest points;

[0063] Calculating the category balance of the number of categories of the interest points;

[0064] The numerical balance and the category balance are used as statistical balance for simplified reporting.

[0065] Optionally, optimizing the simplified report by using the statistical balance to obtain an optimized report includes:

[0066] Using the statistical balance to construct a number anomaly threshold and a category anomaly threshold of the simplified report;

[0067] According to the number abnormality threshold and the category abnormality threshold, identifying the number of abnormal points of interest and the number of abnormal categories of points of interest in the simplified report;

[0068] Based on the number of abnormal interest points and the number of abnormal interest point categories, identifying abnormal classified interest points of the simplified report;

[0069] The simplified report corresponding to the abnormal classification interest point is increased or decreased to obtain an optimized report.

[0070] Optionally, the using the user delivery data to identify the historical user's delivery demand for the consumables includes:

[0071] splicing the user outbound data with the position vector of the user outbound data to obtain spliced ​​data;

[0072] and extracting the splicing features of the splicing data through a preset attention mechanism;

[0073] Outputting the shipment demand probability of the historical user through a preset feedforward network;

[0074] The outbound demand of the consumables by the historical user is identified by using the outbound demand probability.

[0075] In order to solve the above problems, the present invention further provides a report statistics system for consumables, the system comprising:

[0076] A data collection module is used to collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records and voice records;

[0077] A report generation module, used to extract the video entry and exit records of the consumables from the on-site video, identify the text entry and exit records of the consumables from the text records, extract the voice entry and exit records of the consumables from the voice records, and generate standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records;

[0078] A report simplification module, used for identifying the report relevance between each standard report in the standard report, clustering the standard reports according to the report relevance to obtain clustered reports, and simplifying the clustered reports to obtain simplified reports;

[0079] A report optimization module, used for querying the display interest points of the simplified report, calculating the interest point index of the simplified report according to the display interest points, analyzing the statistical balance of the simplified report by using the interest point index, and optimizing the simplified report by using the statistical balance to obtain an optimized report;

[0080] A product determination module, used to determine whether there are new products of the consumables;

[0081] A data extraction module, used for inserting the in-and-out data of the new product into the optimization report when there is a new product of the consumables, obtaining an updated report, and extracting the user out-and-out data of the historical users of the consumables from the updated report;

[0082] The report statistics module is used to use the user's outbound data to identify the historical user's outbound demand for the consumables, build a personalized report for the historical user through the update report based on the outbound demand, and use the personalized report as the report statistics result of the consumables.

[0083] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0084] at least one processor; and,

[0085] a memory communicatively connected to the at least one processor; wherein,

[0086] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to implement the above-mentioned report statistics method for consumables.

[0087] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned report statistics method for consumables.

[0088] Compared with the problems described in the background technology, the embodiment of the present invention extracts the video entry and exit records of the consumables from the on-site video to identify the personnel's entry and exit needs for consumables through the body movements and voice data of the personnel in the video, which can reduce the behavior of manually entering the entry and exit data, thereby improving the efficiency of recording data into reports. The embodiment of the present invention identifies the report correlation between each standard report in the standard report to identify the similarity between different types of consumables. Furthermore, the embodiment of the present invention clusters the standard reports according to the report correlation to centrally count consumables in the same category. Furthermore, the embodiment of the present invention simplifies the clustered reports to simplify a large number of reports when presenting them to the decision-making level of the enterprise, so as to facilitate the decision-making of the decision-making level of the enterprise. The embodiment of the present invention queries the display interest points of the simplified report to query the user's interest points for the consumables in the simplified report, so as to facilitate the subsequent increase of consumables that the user is interested in and the reduction of consumables that the user is not interested in, and when detecting the user's interest points, avoid only considering Considering the situation of frequent purchases by users, and also considering the situation of users continuously purchasing other items after purchasing a certain item, if the user purchases several items in a row before it is his turn to continuously purchase other items after purchasing a certain item, it means that the importance of the last purchased item is relatively low for the user, thereby improving the efficiency of report display. Further, the embodiment of the present invention calculates the interest point index of the simplified report according to the display interest point, so as to count the number and category of interest points under each major category. Further, the embodiment of the present invention analyzes the statistical balance of the simplified report by using the interest point index, so as to identify some abnormal data that are not of interest and abnormal data that are of interest relative to the average data. Further, the embodiment of the present invention optimizes the simplified report by using the statistical balance, so as to filter out the consumables data below the threshold and above the threshold, so as to reduce or increase such consumables data in the actual scenario. The embodiment of the present invention identifies the historical user's outbound demand for the consumables by using the user's outbound data, so as to predict which consumables the user will purchase in the future period based on the user's historical purchase record. Therefore, the consumables report statistics method proposed by the present invention can improve the efficiency of consumables report statistics. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 A schematic diagram of a flow chart of a method for reporting and compiling consumables according to an embodiment of the present invention;

[0090] Figure 2 A schematic diagram of a module for implementing the report statistics method for consumables provided in one embodiment of the present invention.

[0091] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0092] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0093] The embodiment of the present application provides a method for reporting and counting consumables. The execution subject of the method for reporting and counting consumables includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for reporting and counting consumables can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. Example

[0094] Reference Figure 1 FIG. 1 is a flow chart of a method for calculating the report of consumables according to an embodiment of the present invention. In this embodiment, the method for calculating the report of consumables includes:

[0095] S1. Collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records, and voice records.

[0096] In an embodiment of the present invention, the consumables refer to items that are used frequently, have a relatively short lifespan and need to be replaced frequently in daily production, operation or life; the on-site video refers to video data when the manufacturer or merchant of consumables trades the consumables with a wholesaler or user, such as surveillance in a store; the text record refers to text data when the manufacturer or merchant of consumables trades the consumables with a wholesaler or user, such as chat information; and the voice record refers to voice data when the manufacturer or merchant of consumables trades the consumables with a wholesaler or user, such as chat voice.

[0097] S2. Extract the video entry and exit records of the consumables from the on-site video, identify the text entry and exit records of the consumables from the text records, extract the voice entry and exit records of the consumables from the voice records, and generate standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records.

[0098] The embodiment of the present invention extracts the video entry and exit records of the consumables from the on-site video to identify the personnel's entry and exit needs for consumables through the personnel's body movements and voice data in the video. This can reduce the manual entry of entry and exit data, thereby improving the efficiency of recording data into reports.

[0099] The video entry and exit records refer to the entry and exit data of consumables extracted from the on-site video, such as the name of the item, the quantity of the item, and the unit of quantity.

[0100] In one embodiment of the present invention, the extracting the video entry and exit records of the consumables from the on-site video includes: distinguishing between a purchase period and a non-purchase period for the consumables; collecting a first personnel frequency of the consumables from the on-site video during the non-purchase period; identifying the employee identity corresponding to the consumables when the first personnel frequency is greater than a first preset frequency; identifying the user identity corresponding to the consumables when the first personnel frequency is not greater than the first preset frequency; collecting a second personnel frequency of the consumables from the on-site video during the purchase period; identifying the user identity of the consumables according to the employee identity when the second personnel frequency is greater than a second preset frequency. The corresponding non-employee identity; the identity of the personnel in the on-site video is marked by the employee identity, the user identity and the non-employee identity to obtain the marked identity information; the voice data of the personnel in the on-site video is extracted; the voice data is converted into a conversational voice by using the marked identity information and the moment of the voice data; the conversational voice is preprocessed to obtain a preprocessed voice; the preprocessed voice is converted into a feature vector; the feature vector is converted into a conversation text by using a preset hidden Markov model; the intention of the personnel corresponding to the consumables is identified from the conversation text; the probability of the physical behavior of the personnel in the on-site video is analyzed by using the following formula:

[0101] ;

[0102] in, represents the probability of physical behavior, represents the joint point confidence map, represents the joint affinity field, represents the convolutional layer, represents a deep convolutional neural network, An image representing a person in a live video feed, represents the activation function;

[0103] Using the body behavior probability to identify the body behavior of the person; identifying the behavior intention of the person from the body behavior; and comparing the person's intention with the behavior intention. Figure 1 When the personnel intention and the behavior intention are consistent, the personnel intention and the behavior intention are used as the video entry and exit record of the consumable; when the personnel intention and the behavior intention are inconsistent, the personnel intention and the behavior intention are intentionally audited to obtain an intention audit result; the intention audit result is used to determine the video entry and exit record of the consumable.

[0104] Among them, the purchasing period refers to the period when consumables are put into the warehouse, the non-purchasing period refers to the period when consumables are taken out of the warehouse, the first personnel frequency refers to the number of times a person appears in the on-site video during the purchasing period, the first preset frequency refers to the threshold for screening employees and users, which can be set according to the actual scenario, the employee identity refers to the person who appears frequently, the user identity refers to the person who appears occasionally, the second personnel frequency refers to the number of times a person appears in the on-site video during the non-purchasing period, the meaning of the second preset frequency is similar to the meaning of the aforementioned first preset frequency, and no further details are given here, the conversational voice refers to the voice data with the speaker marked for each sentence, and the body behavior refers to the behavior of a person operating consumables, including picking up consumables, pointing to consumables, and putting down consumables.

[0105] Optionally, the process of performing speech preprocessing on the conversational speech to obtain preprocessed speech refers to the process of denoising the conversational speech, the process of converting the preprocessed speech into a feature vector refers to the process of converting speech audio into a mathematical vector, such as spectral features, cepstrum features, etc., the process of converting the feature vector into a conversational text using a preset hidden Markov model refers to using the probability value of a Markov chain to identify text data corresponding to different speech feature data, and the process of performing intention review on the personnel intention and the behavioral intention to obtain the intention review result is achieved through manual review.

[0106] In another embodiment of the present invention, the identifying the personnel intention corresponding to the consumables from the conversation text includes: position encoding the conversation text to obtain a position vector; splicing the position vector and the conversation text to obtain a splicing vector; and extracting the attention feature of the splicing vector using the following formula:

[0107] ;

[0108] in, Represents the attention feature, represents the multi-head attention mechanism, , , The query vector, key vector and key vector representing the calculated concatenated vector, , , express The weight of , , express The weight of , , express The weight of represents the total number of dimensions of the concatenated vector, express The splicing weight of

[0109] The attention feature is used to identify the personnel intention corresponding to the consumables.

[0110] Optionally, the process of using the attention feature to identify the personnel intention corresponding to the consumables refers to the process of using a classifier to receive the attention feature and output a corresponding intention probability.

[0111] In another embodiment of the present invention, identifying a person's behavioral intention from the physical behavior includes: acquiring a live video; performing target detection on the live video to obtain a detection target; using the detection target and the physical behavior to construct behavioral description information of the person; and performing intent classification on the behavioral description information to obtain behavioral intention.

[0112] The behavior description information refers to the combined data obtained by splicing the detection target and the limb behavior.

[0113] Optionally, the process of performing target detection on the live video can be implemented through a YOLOX neural network, and the process of performing intent classification on the behavior description information refers to the process of using a classifier to receive the behavior description information and output a corresponding intent probability.

[0114] Optionally, the principle of identifying the text input and output records of the consumables from the text records is similar to the aforementioned principle of identifying the intention of the person corresponding to the consumables from the conversation text, and is not further elaborated here.

[0115] Optionally, the principle of extracting the voice entry and exit record of the consumables from the voice record is similar to the aforementioned principle of using the marked identity information and the moment of the voice data to convert the voice data into conversational voice; performing voice preprocessing on the conversational voice to obtain preprocessed voice; converting the preprocessed voice into a feature vector; using a preset hidden Markov model to convert the feature vector into a conversation text; and identifying the intention of the person corresponding to the consumables from the conversation text, and will not be further elaborated here.

[0116] Furthermore, in an embodiment of the present invention, the standard report refers to log data containing the time, name, quantity, and quantity unit of the consumables. For example, the report contains log data of toothbrushes, paper towels, and skin care products, and the log data of toothbrushes is that 10 pieces were purchased at 9 o'clock. Then the effect presented in the report should be time: 9 o'clock, name: toothbrush, quantity: 10, quantity unit: piece.

[0117] S3. Identify the report relevance between each standard report in the standard report, cluster the standard reports according to the report relevance to obtain clustered reports, and simplify the clustered reports to obtain simplified reports.

[0118] The embodiment of the present invention is used to identify the similarities between different types of consumables by identifying the report association between each standard report in the standard report.

[0119] Optionally, the process of identifying the report association between each standard report in the standard report refers to the process of identifying whether the types of consumables in the report belong to the same category. For example, toothbrushes and paper towels belong to the same category of consumables, so these two consumables are filled in the same category of reports.

[0120] Furthermore, the embodiment of the present invention clusters the standard reports according to the report relevance, so as to centrally count the consumables in the same category.

[0121] Furthermore, the embodiment of the present invention simplifies the clustered reports so as to simplify a large number of reports when presenting the reports to the decision-making layer of the enterprise, thereby facilitating the decision-making layer of the enterprise to make decisions.

[0122] Optionally, the process of simplifying the cluster report to obtain a simplified report is a process of merging duplicate data, retaining only consumables that have been put in and out of the warehouse, and removing consumables that have not been put in and out of the warehouse.

[0123] S4. Query the display interest points of the simplified report, calculate the interest point index of the simplified report based on the display interest points, analyze the statistical balance of the simplified report using the interest point index, and optimize the simplified report through the statistical balance to obtain an optimized report.

[0124] The embodiment of the present invention queries the display interest points of the simplified report to query the user's interest points for consumables in the simplified report, which is convenient for subsequently adding consumables that the user is interested in and reducing consumables that the user is not interested in. When detecting the user's interest points, it avoids only considering the situation where the user frequently purchases, and also considers the situation where the user continuously purchases other items after purchasing a certain item. If the user purchases several items in a row before it is his turn to continuously purchase other items after purchasing a certain item, it means that the importance of the last purchased item is lower for the user, thereby improving the efficiency of report display.

[0125] The display interest points refer to the consumables data that the user is interested in.

[0126] In one embodiment of the present invention, the querying of the display interest points of the simplified report includes: querying the scheduled sequence of the simplified report; generating a scheduling graph structure of the scheduled sequence; and calculating the scheduled density of the simplified report according to the scheduled sequence and the scheduling graph structure using the following formula:

[0127] ;

[0128] in, Indicates the scheduling density of the u-th report data in the simplified report. Indicates the bit sequence of the u-th report data in the j-th scheduled sequence, Indicates the total order, express The number of neighboring report data of the u-th report data in the scheduling graph structure under the position order, Indicates the number of scheduled sequences;

[0129] When the scheduled density is greater than a preset density, the report data corresponding to the scheduled density is used as a display interest point of a simplified report.

[0130] The scheduled sequence refers to a continuous outbound data of consumables presented in the simplified report, for example, at 9 o'clock, a user buys consumable No. A once, consumable No. B once, and consumable No. C once. The scheduling graph structure refers to a directed graph obtained by arranging the consumables scheduled in chronological order. Other nodes represent the scheduled consumables, and pointers represent the order. The neighborhood report data refers to the consumables in the scheduling graph structure. The predecessor and successor of the u-th report data in the order.

[0131] Furthermore, the embodiment of the present invention calculates the interest point index of the simplified report based on the displayed interest points, so as to count the number and category of interest points under each major category.

[0132] In one embodiment of the present invention, the calculation of the interest point index of the simplified report based on the displayed interest points includes: classifying the interest points of the displayed interest points to obtain classified interest points; collecting the number of interest points and the number of interest point categories corresponding to the classified interest points; and using the number of interest points and the number of interest point categories as the interest point index of the simplified report.

[0133] The classified interest points refer to the displayed interest points under each major category, and the number of interest point categories refers to the number of interest point categories included in the classified interest points.

[0134] Furthermore, the embodiment of the present invention analyzes the statistical balance of the simplified report by utilizing the interest point index to identify certain abnormal data that are not of interest and abnormal data that are of interest relative to the average data.

[0135] The statistical balance refers to the variance corresponding to the classified interest points.

[0136] In one embodiment of the present invention, the analyzing the statistical balance of the simplified report using the interest point index includes: obtaining the number of interest points and the number of interest point categories in the interest point index; and calculating the number balance of the number of interest points using the following formula:

[0137] ;

[0138] in, Indicates the balance of numbers. represents the number of interest points of the vth classification interest point, express The mean number of interest points of each classification point of interest, Indicates the number of classified interest points;

[0139] Calculate the category balance of the number of interest point categories; and use the number balance and the category balance as the statistical balance of the simplified report.

[0140] Optionally, the principle of calculating the category balance of the number of interest point categories is similar to the principle of calculating the number balance of the number of interest points mentioned above. It should be noted that when calculating the category balance of the number of interest point categories, it is necessary to use the one-hot encoding method to convert the categories into numbers.

[0141] Furthermore, the embodiment of the present invention optimizes the simplified report through the statistical balance to filter out the consumables data below a threshold or above a threshold, thereby reducing or increasing such consumables data in actual scenarios.

[0142] In one embodiment of the present invention, the simplified report is optimized by the statistical balance to obtain an optimized report, including: using the statistical balance to construct a number abnormality threshold and a category abnormality threshold for the simplified report; identifying the number of abnormal interest points and the number of abnormal interest point categories of the simplified report based on the number abnormality threshold and the category abnormality threshold; identifying abnormal classification interest points of the simplified report based on the number of abnormal interest points and the number of abnormal interest point categories; and performing addition and subtraction processing on the simplified report corresponding to the abnormal classification interest points to obtain an optimized report.

[0143] Among them, the number anomaly threshold refers to the product of the square root of the number balance and the multiple. The meaning of the category anomaly threshold is similar to that of the number anomaly threshold, and no further elaboration is given here. The number of abnormal interest points and the number of abnormal interest point categories refer to data greater than the threshold or lower than the threshold. It should be noted that there are two number anomaly thresholds, one for filtering data greater than the threshold and the other for filtering data lower than the threshold. Similarly, there are two category anomaly thresholds.

[0144] Optionally, the process of performing addition and subtraction processing on the simplified report corresponding to the abnormal classification interest point to obtain the optimized report refers to the process of reducing the report data corresponding to the abnormal classification interest point or increasing the report data corresponding to the abnormal classification interest point.

[0145] S5. Determine whether there are new products of the consumables.

[0146] In the embodiment of the present invention, the new product refers to a newly developed product whose name has never been recorded in the historical report data.

[0147] S6. When there is a new product of the consumables, the in-and-out data of the new product is inserted into the optimization report to obtain an updated report, and the user out-and-out data of the historical users of the consumables are extracted from the updated report.

[0148] In the embodiment of the present invention, the user outbound data refers to the historical data of the user's purchase of consumables.

[0149] S7. Utilize the user's outbound delivery data to identify the historical user's outbound delivery demand for the consumables, build a personalized report for the historical user through the update report based on the outbound delivery demand, and use the personalized report as the report statistical result of the consumables.

[0150] The embodiment of the present invention uses the user's outbound delivery data to identify the historical user's outbound delivery demand for the consumables, so as to predict which consumables the user will purchase in a future period based on the user's historical purchase record.

[0151] The outbound demand refers to the user's demand for purchasing consumables in a future period.

[0152] In one embodiment of the present invention, the use of the user's out-of-stock data to identify the historical user's out-of-stock demand for the consumables includes: splicing the user's out-of-stock data with the position vector of the user's out-of-stock data to obtain spliced ​​data; and extracting the splicing features of the spliced ​​data through a preset attention mechanism; outputting the probability of the historical user's out-of-stock demand through a preset feedforward network; and using the out-of-stock demand probability to identify the historical user's out-of-stock demand for the consumables.

[0153] Among them, the preset attention mechanism refers to a multi-head attention mechanism.

[0154] Furthermore, in an embodiment of the present invention, the personalized report refers to a report obtained by adding the shipment demand to an update report or by adding a record of the shipment demand report to an original update report.

[0155] Compared with the problems described in the background technology, the embodiment of the present invention extracts the video entry and exit records of the consumables from the on-site video to identify the personnel's entry and exit needs for consumables through the body movements and voice data of the personnel in the video, which can reduce the behavior of manually entering the entry and exit data, thereby improving the efficiency of recording data into reports. The embodiment of the present invention identifies the report correlation between each standard report in the standard report to identify the similarity between different types of consumables. Furthermore, the embodiment of the present invention clusters the standard reports according to the report correlation to centrally count consumables in the same category. Furthermore, the embodiment of the present invention simplifies the clustered reports to simplify a large number of reports when presenting them to the decision-making level of the enterprise, so as to facilitate the decision-making of the decision-making level of the enterprise. The embodiment of the present invention queries the display interest points of the simplified report to query the user's interest points for the consumables in the simplified report, so as to facilitate the subsequent increase of consumables that the user is interested in and the reduction of consumables that the user is not interested in, and when detecting the user's interest points, avoid only considering Considering the situation of frequent purchases by users, and also considering the situation of users continuously purchasing other items after purchasing a certain item, if the user purchases several items in a row before it is his turn to continuously purchase other items after purchasing a certain item, it means that the importance of the last purchased item is relatively low for the user, thereby improving the efficiency of report display. Further, the embodiment of the present invention calculates the interest point index of the simplified report according to the display interest point, so as to count the number and category of interest points under each major category. Further, the embodiment of the present invention analyzes the statistical balance of the simplified report by using the interest point index, so as to identify some abnormal data that are not of interest and abnormal data that are of interest relative to the average data. Further, the embodiment of the present invention optimizes the simplified report by using the statistical balance, so as to filter out the consumables data below the threshold and above the threshold, so as to reduce or increase such consumables data in the actual scenario. The embodiment of the present invention identifies the historical user's outbound demand for the consumables by using the user's outbound data, so as to predict which consumables the user will purchase in the future period based on the user's historical purchase record. Therefore, the consumables report statistics method proposed by the present invention can improve the efficiency of consumables report statistics. Example

[0156] like Figure 2 The figure shows a functional module diagram of a report statistical system for consumables of the present invention.

[0157] The report statistics system 200 of consumables described in the present invention can be installed in an electronic device. According to the functions to be implemented, the report statistics system of consumables can include a data acquisition module 201, a report generation module 202, a report simplification module 203, a report optimization module 204, a product judgment module 205, a data extraction module 206 and a report statistics module 207. The module described in the present invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0158] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0159] The data collection module 201 is used to collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records and voice records;

[0160] The report generation module 202 is used to extract the video entry and exit records of the consumables from the on-site video, identify the text entry and exit records of the consumables from the text records, extract the voice entry and exit records of the consumables from the voice records, and generate standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records;

[0161] The report simplification module 203 is used to identify the report relevance between each standard report in the standard report, cluster the standard reports according to the report relevance to obtain clustered reports, and simplify the clustered reports to obtain simplified reports;

[0162] The report optimization module 204 is used to query the display interest points of the simplified report, calculate the interest point index of the simplified report according to the display interest points, analyze the statistical balance of the simplified report using the interest point index, and optimize the simplified report through the statistical balance to obtain an optimized report;

[0163] The product determination module 205 is used to determine whether there is a new product of the consumables;

[0164] The data extraction module 206 is used to insert the in-and-out data of the new product into the optimization report when there is a new product of the consumables, obtain an updated report, and extract the user out-and-out data of the historical users of the consumables from the updated report;

[0165] The report statistics module 207 is used to use the user's outbound delivery data to identify the historical user's outbound delivery needs for the consumables, and build a personalized report for the historical user through the updated report based on the outbound delivery needs, and use the personalized report as the report statistics result of the consumables.

[0166] In detail, each module in the consumables report statistics system 200 in the embodiment of the present invention adopts the same method as above when used. Figure 1 The same technical means are used as the statistical method for the report of consumables described in , and can produce the same technical effects, so I will not go into details here.

[0167] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for reporting and counting consumables, characterized in that: The method comprises: Collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records, and voice records; Extracting the video entry and exit records of the consumables from the on-site video, identifying the text entry and exit records of the consumables from the text records, extracting the voice entry and exit records of the consumables from the voice records, and generating standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records; Identifying the report relevance between each standard report in the standard report, clustering the standard reports according to the report relevance to obtain clustered reports, and simplifying the clustered reports to obtain simplified reports; Querying the display interest points of the simplified report, calculating the interest point index of the simplified report according to the display interest points, analyzing the statistical balance of the simplified report using the interest point index, and optimizing the simplified report through the statistical balance to obtain an optimized report, wherein querying the display interest points of the simplified report includes: Querying the scheduled sequence of the simplified report; generating a scheduling graph structure of the scheduled sequence; According to the scheduled sequence and the scheduling graph structure, the scheduling density of the simplified report is calculated using the following formula: ； in, Indicates the scheduling density of the u-th report data in the simplified report. Indicates the bit sequence of the u-th report data in the j-th scheduled sequence, Indicates the total order, express The number of neighboring report data of the u-th report data in the scheduling graph structure under the position order, Indicates the number of scheduled sequences; When the scheduled density is greater than the preset density, the report data corresponding to the scheduled density is used as a display interest point of a simplified report; The step of optimizing the simplified report by using the statistical balance to obtain an optimized report includes: Using the statistical balance to construct a number anomaly threshold and a category anomaly threshold of the simplified report; According to the number abnormality threshold and the category abnormality threshold, identifying the number of abnormal points of interest and the number of abnormal categories of points of interest in the simplified report; Based on the number of abnormal interest points and the number of abnormal interest point categories, identifying abnormal classified interest points of the simplified report; Performing addition and subtraction processing on the simplified report corresponding to the abnormal classification interest point to obtain an optimized report; Determine whether there are new products of the consumables; When there is a new product of the consumables, the in-and-out data of the new product is inserted into the optimization report to obtain an update report, and the user out-and-out data of the historical users of the consumables are extracted from the update report; The user's outbound delivery data is utilized to identify the historical user's outbound delivery demand for the consumables, and based on the outbound delivery demand, a personalized report of the historical user is constructed through the updated report, and the personalized report is used as the report statistical result of the consumables, wherein the outbound delivery demand refers to the user's purchase demand for consumables in a future period, and the personalized report refers to a report obtained by adding the outbound delivery demand to the updated report and adding a record of the report of the outbound delivery demand on the basis of the original updated report.

2. The method for reporting and counting consumables according to claim 1, characterized in that: The extracting the video entry and exit record of the consumables from the on-site video includes: Distinguish between the purchase period and non-purchase period of the consumables; During the non-restocking period, collecting the first personnel frequency of the consumables from the on-site video; When the first personnel frequency is greater than a first preset frequency, identifying the employee identity corresponding to the consumable; When the frequency of the first person is not greater than a first preset frequency, identifying the user identity corresponding to the consumable; During the purchase period, collecting the second personnel frequency of the consumables from the on-site video; When the second personnel frequency is greater than a second preset frequency, identifying the non-employee identity corresponding to the consumable according to the employee identity; Marking the persons in the on-site video by using the employee identity, the user identity and the non-employee identity to obtain marked identity information; Extracting voice data of people in the live video; Converting the voice data into conversational voice using the marked identity information and the time of the voice data; Performing speech preprocessing on the conversation speech to obtain preprocessed speech; Converting the preprocessed speech into a feature vector; Converting the feature vector into a dialogue text using a preset hidden Markov model; Identifying the personnel intention corresponding to the consumables from the conversation text; The following formula is used to analyze the probability of the physical behavior of the personnel in the live video: ； in, represents the probability of physical behavior, represents the joint point confidence map, represents the joint affinity field, represents the convolutional layer, represents a deep convolutional neural network, An image representing a person in a live video feed, represents the activation function; identifying the physical behavior of the person using the physical behavior probability; Identify the person's behavioral intention from the physical behavior; When the personnel intention is consistent with the behavior intention, the personnel intention and the behavior intention are used as the video entry and exit records of the consumables; When the personnel intention is inconsistent with the behavior intention, an intention review is performed on the personnel intention and the behavior intention to obtain an intention review result; The intent audit result is used to determine the video access record as the consumable item.

3. The method for reporting and counting consumables according to claim 2, characterized in that: The identifying the personnel intention corresponding to the consumable item from the conversation text includes: Performing position encoding on the dialogue text to obtain a position vector; splicing the position vector and the dialogue text to obtain a splicing vector; The attention feature of the concatenated vector is extracted using the following formula: ； in, Represents the attention feature, represents the multi-head attention mechanism, , , The query vector, key vector and key vector representing the calculated concatenated vector, , , express The weight of , , express The weight of , , express The weight of represents the total number of dimensions of the concatenated vector, express The splicing weight of The attention feature is used to identify the personnel intention corresponding to the consumables.

4. The method for reporting and counting consumables according to claim 2, characterized in that: The identifying the behavioral intention of a person from the body behavior includes: Get live video; Performing target detection on the live video to obtain a detection target; constructing behavior description information of the person using the detection target and the body behavior; The behavior description information is classified by intent to obtain the behavior intention.

5. The method for reporting and counting consumables according to claim 1, characterized in that: The step of calculating the interest point index of the simplified report according to the displayed interest points includes: Classifying the displayed interest points to obtain classified interest points; Collecting the number of points of interest and the number of categories of points of interest corresponding to the classified points of interest; The number of points of interest and the number of categories of points of interest are used as the points of interest indicators of the simplified report.

6. The method for reporting and counting consumables according to claim 1, characterized in that: The analyzing the statistical balance of the simplified report using the interest point index includes: Obtaining the number of points of interest and the number of categories of points of interest in the point of interest index; The number balance of the number of interest points is calculated using the following formula: ； in, Indicates the balance of numbers, represents the number of interest points of the vth classification interest point, express The mean number of points of interest for each category of points of interest, Indicates the number of classified interest points; Calculating the category balance of the number of categories of the interest points; The numerical balance and the category balance are used as statistical balance for simplified reporting.

7. The method for reporting and counting consumables according to claim 1, characterized in that: The using the user delivery data to identify the historical user's delivery demand for the consumables includes: splicing the user outbound data with the position vector of the user outbound data to obtain spliced ​​data; and extracting the splicing features of the splicing data through a preset attention mechanism; Outputting the shipment demand probability of the historical user through a preset feedforward network; The outbound demand of the consumables by the historical user is identified by using the outbound demand probability.

8. A report statistics system for consumables, characterized in that: The system comprises: A data collection module is used to collect historical in-and-out data of consumables, wherein the historical in-and-out data includes on-site video, text records and voice records; A report generation module, used to extract the video entry and exit records of the consumables from the on-site video, identify the text entry and exit records of the consumables from the text records, extract the voice entry and exit records of the consumables from the voice records, and generate standard reports of the video entry and exit records, the text entry and exit records, and the voice entry and exit records; A report simplification module, used for identifying the report relevance between each standard report in the standard report, clustering the standard reports according to the report relevance to obtain clustered reports, and simplifying the clustered reports to obtain simplified reports; The report optimization module is used to query the display interest points of the simplified report, calculate the interest point index of the simplified report according to the display interest points, analyze the statistical balance of the simplified report using the interest point index, and optimize the simplified report through the statistical balance to obtain an optimized report, wherein the querying of the display interest points of the simplified report includes: Querying the scheduled sequence of the simplified report; generating a scheduling graph structure of the scheduled sequence; According to the scheduled sequence and the scheduling graph structure, the scheduling density of the simplified report is calculated using the following formula: ； in, Indicates the scheduling density of the u-th report data in the simplified report. Indicates the bit sequence of the u-th report data in the j-th scheduled sequence, Indicates the total order, express The number of neighboring report data of the u-th report data in the scheduling graph structure under the position order, Indicates the number of scheduled sequences; When the scheduled density is greater than the preset density, the report data corresponding to the scheduled density is used as a display interest point of a simplified report; The step of optimizing the simplified report by using the statistical balance to obtain an optimized report includes: Using the statistical balance to construct a number anomaly threshold and a category anomaly threshold of the simplified report; According to the number abnormality threshold and the category abnormality threshold, identifying the number of abnormal points of interest and the number of abnormal categories of points of interest in the simplified report; Based on the number of abnormal interest points and the number of abnormal interest point categories, identifying abnormal classified interest points of the simplified report; Performing addition and subtraction processing on the simplified report corresponding to the abnormal classification interest point to obtain an optimized report; A product determination module, used to determine whether there are new products of the consumables; A data extraction module, used for inserting the in-and-out data of the new product into the optimization report when there is a new product of the consumables, obtaining an updated report, and extracting the user out-and-out data of the historical users of the consumables from the updated report; A report statistics module is used to use the user's outbound delivery data to identify the historical user's outbound delivery demand for the consumables, and based on the outbound delivery demand, build a personalized report for the historical user through the updated report, and use the personalized report as the report statistics result of the consumables, wherein the outbound delivery demand refers to the user's purchase demand for consumables in a future period, and the personalized report refers to a report obtained by adding the outbound delivery demand to the updated report and adding a record of the report of the outbound delivery demand on the basis of the original updated report.

Images