CN117829714A - Intelligent storage identification analysis system and method thereof - Google Patents

Abstract

The invention discloses an intelligent storage identification analysis system and a method thereof. The intelligent storage identification analysis system comprises a storage data acquisition module, a storage identification analysis module and a storage monitoring center. According to the storage cabinet storage system, storage data of the storage cabinet to be monitored are obtained through the storage data obtaining module and are sent to the storage identification analysis module and the storage monitoring center, then the storage identification analysis module detects the shielding degree of objects in the preset range outside the storage cabinet to be monitored according to the received data, then the characteristics of corresponding storage clients are identified, and when the storage clients take out the storage objects, characteristic comparison is carried out, the corresponding available storage cabinet is analyzed to obtain storage analysis results, finally the storage data and the storage analysis results are stored by the storage monitoring center, and whether the storage cabinet to be monitored is abnormal or not is judged to take processing measures, so that the effect of improving the storage identification analysis accuracy is achieved, and the problem that the storage identification analysis accuracy is low in the prior art is solved.

Description

Intelligent storage identification analysis system and method thereof

Technical Field

The invention relates to the technical fields of supermarket shopping, market shopping, warehouse shipping and the like, in particular to an intelligent storage identification analysis system and a method thereof.

Background

The storage identification analysis refers to a process of identifying and analyzing the storage by a technical means so as to improve the efficiency and accuracy of storage management. This process involves a variety of storage scenarios including warehouse, shelf, warehouse, home, etc. Through automatic and intelligent storing discernment, can improve storing management's efficiency greatly, reduce manual operation and mistake. The storage identification analysis can provide accurate storage information including positions, numbers, states and the like, and is helpful for timely knowing the inventory condition. By analyzing the storage information, the utilization of the storage space can be optimized, and the storage density of the warehouse or the warehouse room can be improved.

In the prior art, the following methods are mainly used for identifying and analyzing the stored objects: the method comprises the steps of scanning a bar code and a two-dimensional code, and reading the bar code or the two-dimensional code on the storage by using scanning equipment, so that information of the storage is identified; radio Frequency Identification (RFID) technology, the position and state of the storage are identified in real time through radio signals, and direct contact is not needed; visual identification, shooting a storage scene through a camera, and then analyzing storage information in an image by utilizing an image identification algorithm; sensor technology, which utilizes various sensors to collect data about the storage.

For example, bulletin numbers: the storage system based on the database for intelligently identifying the articles, disclosed by the invention of CN107752584B, comprises a storage device main body, an intelligent video module, a position locating module, a data transmission module, a central processing unit module, a mobile phone terminal module and the like; the storage device main body is used for storing articles; the intelligent video module is used for shooting the appearance of the article; the position positioning module is used for acquiring the position data of the main body of the storage device; the data transmission module is used for uploading shot video data and position data to the central processing unit module; the central processing unit module is used for processing shot video and position data, identifying the loading and unloading of articles, identifying the types and the quantity of articles put into the storage device main body, and transmitting related data to the mobile phone terminal, the storage device main body and other equipment; the mobile phone terminal module is mainly used for displaying data processing results.

For example, bulletin numbers: an automatic article taking cabinet and an intelligent article storage system of the invention patent publication of CN110033062B, the intelligent article storage system comprises: the system comprises an RFID identification module, an input module, a storage module, an analysis matching module and an output module; the RFID identification module is used for identifying RFID tag information on the articles stored in the storage drawer; the input module is used for inputting accurate name information or fuzzy name information of the searched article when the article is taken out and selecting the taken-out article; the storage module is used for storing RFID tag information on articles stored in the storage drawer and corresponding position information of the storage drawer; the analysis matching module is connected with the input module, the storage module and the output module and is used for comparing and matching the RFID tag information in the storage module with the accurate name information or the fuzzy name information of the searched object, giving out the matched object name for selection, and then conveying the position information of the selected object to the output module; the output module is connected with the automatic object taking cabinet, and the output module enables the object taking device to move to the corresponding position according to the position information of the selected object, enables the corresponding object storing drawer to move to the object taking device, and then enables the object taking device to move to the proper position.

However, in the process of implementing the technical scheme of the invention in the embodiment of the application, the application finds that the above technology has at least the following technical problems:

in the prior art, the accuracy of the identification technology directly influences the stability and accuracy of the storage identification analysis, the selection of the identification data acquisition equipment directly influences the identification analysis result, and the problem of low storage identification analysis accuracy exists.

Disclosure of Invention

According to the intelligent storage identification analysis system and the intelligent storage identification analysis method, the problem of low storage identification analysis accuracy in the prior art is solved, and the improvement of the storage identification analysis accuracy is realized.

The embodiment of the application provides an intelligent storage identification analysis system, which comprises a storage data acquisition module, a storage identification analysis module and a storage monitoring center; the storage data acquisition module is used for acquiring storage data of the storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and the acquired storage data are sent to the storage identification analysis module and the storage monitoring center; the storage identification analysis module is used for analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to the storage monitoring center; the storage monitoring center is used for storing storage data of the storage cabinet to be monitored, the storage data comprise storage data and storage analysis results, and whether the storage cabinet to be monitored has storage abnormality or not is judged according to the storage analysis results so as to take processing measures.

Further, the storage data acquisition module comprises a storage image data acquisition unit, a user characteristic data acquisition unit, a size data acquisition unit and a category data acquisition unit; the storage image data acquisition unit: the storage cabinet monitoring method comprises the steps of acquiring storage image data of a storage cabinet to be monitored, wherein the storage image data are used for recording pictures in a preset range outside the storage cabinet to be monitored in a monitoring time period; the user characteristic data acquisition unit: the method comprises the steps of acquiring user characteristic data of a storage cabinet to be monitored, wherein the user characteristic data are used for describing characteristics of storage clients of the storage cabinet to be monitored in a monitoring time period; the size data acquisition unit: the method comprises the steps of acquiring article size data of a storage cabinet to be monitored, wherein the article size data are used for describing external characteristic conditions of articles stored in the storage cabinet to be monitored in a monitoring time period, and specifically comprise length, width, height and weight; the category data acquisition unit: the storage type of the storage cabinet to be monitored is used for describing the storage type of the storage articles of the storage cabinet to be monitored in the monitoring time period, and the storage type comprises freezing storage and normal-temperature storage.

Further, the storage identification and analysis module comprises a shielding identification unit, a user characteristic identification unit and an available locker analysis unit; the shielding identification unit: the storage cabinet shielding model is used for detecting the article shielding degree in a preset range outside the storage cabinet to be monitored in a monitoring time period by combining the received storage image data with a trained storage cabinet shielding model, and the storage cabinet shielding model is used for detecting a preset shielding object in the input storage image data; the user characteristic recognition unit: the storage client characteristic recognition module is used for recognizing and storing the characteristics of the storage client of the storage cabinet to be monitored in the monitoring time period by combining the received user characteristic data with the trained user characteristic recognition module, and the user characteristic recognition module is used for recognizing and storing the characteristics of the storage client of the storage cabinet to be monitored and comparing the characteristics when the storage client takes out the storage object; the available locker analysis unit: the storage cabinet monitoring system is used for analyzing the corresponding available storage cabinets in the storage cabinets to be monitored in the monitoring time period by combining the received article size data and the article category data, and opening the distributed available storage cabinets for storage of storage clients according to the analysis result, wherein the available storage cabinets are storage cabinets meeting the storage requirements of the storage clients.

Further, the specific training method of the locker shielding model is as follows: constructing a historical storage image data set, wherein the historical storage image data set is a set of historical storage image data, and the historical storage image data is used for recording pictures in a preset range outside a storage cabinet to be monitored in a historical time period; dividing a historical storage image data set according to a preset proportion to obtain a historical storage image training set and a historical storage image verification set, and carrying out preset shielding object labeling on the historical storage image data in the historical storage image training set; training a first locker shielding model, and verifying the first locker shielding model through a historical storage image verification set, wherein the first locker shielding model is used for describing a converged model obtained by training and optimizing a preset detection model through a marked historical storage image training set, and the preset detection model is used for describing a model for realizing target detection based on deep learning; the method comprises the steps of evaluating shielding detection performance of a first locker shielding model, screening the first locker shielding model through shielding detection performance to obtain the locker shielding model, and evaluating shielding detection performance by obtaining shielding detection performance scores of the first locker shielding model, wherein the shielding detection performance scores are used for describing detection accuracy degree and detection real-time degree of the first locker shielding model.

Further, a specific acquisition process of the shielding detection performance score of the shielding model of the first locker is as follows: acquiring historical storage image verification time and historical storage image verification data, wherein the historical storage image verification time is used for describing the time for detecting each historical storage image data in a historical storage image verification set through a first locker shielding model, and the historical storage image verification data is used for describing a result obtained by detecting each historical storage image data in the historical storage image verification set through the first locker shielding model; acquiring reference historical storage image verification data, wherein the reference historical storage image verification data is used for describing results obtained by manually marking preset shielding objects of each historical storage image data in the historical storage image verification set; the method comprises the steps of obtaining a shielding detection weight, wherein the shielding detection weight is used for describing the influence degree of detection accuracy relative deviation and detection real-time relative deviation of a preset shielding object detected through a first locker shielding model on shielding detection performance, and comprises the detection accuracy weight and the detection real-time weight; the method comprises the steps of obtaining a detection accuracy score and a detection instantaneity score of a first locker shielding model, and obtaining a shielding detection performance score of the first locker shielding model by combining shielding detection weights, wherein the detection accuracy score is used for describing the detection accuracy degree of a first locker detection model for detecting a preset shielding object, and the detection instantaneity score is used for describing the detection instantaneity degree of the first locker detection model for detecting the preset shielding object.

Further, the shielding detection performance score of the shielding model of the first locker is calculated by adopting the following formula:in which, in the process,numbering the shielding model of the first locker, +.>，/>The total number of occlusion models for the first locker,is->Shielding detection performance of shielding model of first storage cabinetScore, ->And->Respectively +.>Detecting accuracy score and detecting real-time score of shielding model of first locker, and (E) detecting real-time score of shielding model of first locker>And->A reference detection accuracy score and a reference detection real-time score, respectively, < >>And->Correction factors for the detection accuracy score and the detection real-time score, respectively, < >>And->The detection accuracy weight and the detection real-time weight are respectively.

Further, the specific training method of the user characteristic recognition model is as follows: constructing a reference user characteristic data set, wherein the reference user characteristic data set is a set of reference user characteristic data, the reference user characteristic data comprises historical user characteristic data and generated user characteristic data, the historical user characteristic data is used for describing characteristics of storage clients of a storage cabinet to be monitored in a historical time period, and the generated user characteristic data is used for describing characteristics of the storage clients generated according to the historical user characteristic data; dividing the reference user characteristic data set according to a preset proportion to obtain a reference user characteristic training set and a reference user characteristic verification set, and labeling the reference user characteristic training set; training a first user characteristic recognition model, and verifying the first user characteristic recognition model by referring to a user characteristic verification set, wherein the first user characteristic recognition model is used for describing a converged model obtained by training and optimizing a preset recognition model by referring to a labeled reference user characteristic training set, and the preset recognition model is used for describing a model for realizing user characteristic recognition based on multiple modes; and evaluating the user recognition performance of the first user feature recognition model, screening the first user feature recognition model through the user recognition performance to obtain the user feature recognition model, wherein the user recognition performance is evaluated by obtaining the user recognition performance score of the first user feature recognition model, and the user recognition performance score is used for describing the recognition accuracy degree and the recognition instantaneity degree of the first user feature recognition model.

Further, the specific acquisition process of the user identification performance score of the first user characteristic identification model is as follows: acquiring reference user feature verification data and reference user feature verification time, wherein the reference user feature verification data is used for describing a result obtained by identifying each piece of reference user feature data in a reference user feature verification set through a first user feature identification model, and the reference user feature verification time is used for describing time for identifying each piece of reference user feature data in the reference user feature verification set through the first user feature identification model; acquiring user identification weights, wherein the user identification weights are used for describing the influence degree of the identification accuracy absolute deviation and the identification real-time absolute deviation of the user characteristics on the user characteristic performance through a first user characteristic identification model, and the influence degree comprises the identification accuracy weights and the identification real-time weights; acquiring identification accuracy scores of a first user characteristic identification model by combining preset user characteristic verification data, wherein the identification accuracy scores are used for describing the accuracy degree of the first user characteristic identification model for identifying user characteristics; acquiring an identification real-time score of a first user characteristic identification model by combining with a preset verification time, wherein the identification real-time score is used for describing the real-time degree of the user characteristic identification model for identifying the user characteristic; and calculating the user recognition performance score of the first user characteristic recognition model according to the recognition accuracy score and the recognition real-time score and combining the user recognition weight.

Further, the user recognition performance score of the first user feature recognition model is calculated using the following formula:wherein->Identifying the number of the model for the first user feature, +.>，/>Identifying a total number of models for the first user feature, < >>Is->User identification performance score of the first user feature identification model,/for each user feature>And->Respectively +.>Identification accuracy score and identification real-time score of the first user feature identification model,/->And->A reference identification accuracy score and a reference identification real-time score, respectively, < >>And->Correction factors for the recognition accuracy score and the recognition real-time score, respectively, < >>And->Respectively, an identification accuracy weight and an identification real-time weight.

The embodiment of the application provides an intelligent storage identification analysis method, which comprises the following steps: the method comprises the steps of obtaining storage data of a storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and sending the obtained storage data to a storage identification analysis module and a storage monitoring center; analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to a storage monitoring center; and storing storage data of the storage cabinet to be monitored, wherein the storage data comprises storage data and storage analysis results, and judging whether the storage cabinet to be monitored has storage abnormality according to the storage analysis results so as to take processing measures.

One or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:

1. through obtaining the storing data of the locker to be monitored, then detect the article shielding degree of the outside preset scope of the locker to be monitored according to storing data, then discern the characteristic of corresponding storage customer in order to carry out the characteristic comparison when storage customer takes out the storage article, available locker of re-analysis obtains storing analysis result, storing data and storing analysis result and judging whether the locker to be monitored exists the storing unusual, thereby realized the accurate discernment analysis of locker storing process to be monitored, and then realized the improvement of storing discernment analysis accuracy, effectively solved the problem that storing discernment analysis accuracy is low among the prior art.

2. Through constructing the historical storage image dataset, dividing the historical storage image dataset to obtain a historical storage image training set and a historical storage image verification set, training and optimizing a preset detection model through the marked historical storage image training set to obtain a first locker shielding model, verifying through the historical storage image verification set, evaluating shielding detection performance of the first locker shielding model, and finally screening according to the shielding detection performance to obtain a locker shielding model, so that accurate detection of a preset shielding object is achieved, and further the shielding problem of the outside of the locker to be monitored in a preset range is processed more accurately and rapidly.

3. The method comprises the steps of obtaining reference user characteristic verification data and reference user characteristic verification time, calculating the identification accuracy score of a first user characteristic identification model, calculating the corresponding identification real-time score by combining with preset verification time, calculating the user identification performance score of the first user characteristic identification model by combining with user identification weight, and finally screening the user characteristic identification model according to the user identification performance score, so that the user identification performance of the first user characteristic identification model is quantized, and further the characteristics of a storage client of a storage cabinet to be monitored are accurately identified.

Drawings

Fig. 1 is a schematic structural diagram of an intelligent storage identification analysis system according to an embodiment of the present application;

fig. 2 is a schematic structural diagram of a storage data acquisition module in an intelligent storage identification analysis system according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of a storage identification analysis module in an intelligent storage identification analysis system according to an embodiment of the present application.

Detailed Description

According to the storage identification analysis system and the storage identification analysis method, the problem of low storage identification analysis accuracy in the prior art is solved, storage image data, user characteristic data, article size data and article type data of a storage cabinet to be monitored are respectively acquired through a storage image data acquisition unit, a user characteristic data acquisition unit, a size data acquisition unit and a type data acquisition unit in a storage data acquisition module, the acquired data are sent to a storage identification analysis module and a storage monitoring center, then a shielding recognition unit in the storage identification analysis module is combined with received storage image data and a trained storage cabinet shielding model to detect the shielding degree of articles in a storage cabinet outside preset range in a monitoring time period, then the user characteristic recognition unit is combined with the received user characteristic data and the trained user characteristic recognition model to recognize characteristics of a storage customer to be monitored in the monitoring time period, and when the storage customer is taken out of the storage article, the storage cabinet is compared in the storage cabinet analysis time period by the storage cabinet analysis unit, and the storage cabinet is combined with the received article size data and the article type data, and the storage cabinet is corresponding to be monitored in the storage cabinet monitoring time period, and finally the storage cabinet is opened according to the storage cabinet analysis result to the storage cabinet analysis result, whether the storage customer is accurately stored in the storage cabinet storage analysis result is judged, and the storage analysis result is accurately stored, and whether the storage analysis result is accurately is judged by the storage analysis result.

The technical scheme in this application embodiment is for solving the low problem of above-mentioned storing discernment analysis accuracy, and the overall thinking is as follows:

according to the storage cabinet monitoring method, storage data of the storage cabinet to be monitored are obtained, then article shielding degree in a preset range outside the storage cabinet to be monitored in a monitoring time period is detected according to the storage data, then characteristics of corresponding storage clients are identified, characteristic comparison is carried out when the storage clients take out storage articles, the available storage cabinets are analyzed to obtain storage analysis results, finally the storage data and the storage analysis results are stored, whether storage abnormality exists in the storage cabinet to be monitored is judged, and processing measures are taken, so that the effect of improving storage identification analysis accuracy is achieved.

In order to better understand the above technical solutions, the following detailed description will refer to the accompanying drawings and specific embodiments.

As shown in fig. 1, a schematic structural diagram of an intelligent storage identification analysis system provided in an embodiment of the present application includes a storage data acquisition module, a storage identification analysis module, and a storage monitoring center; the storage data acquisition module is used for acquiring storage data of the storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and the acquired storage data are sent to the storage identification analysis module and the storage monitoring center; the storage identification analysis module is used for analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to the storage monitoring center; the storage monitoring center is used for storing storage data of the storage cabinet to be monitored, wherein the storage data comprises storage data and storage analysis results, and judging whether the storage cabinet to be monitored has storage abnormality according to the storage analysis results so as to take processing measures.

In this embodiment, the storage abnormality includes that the storage space cannot meet the requirement, that there is a large-area shielding in a preset range outside the locker to be monitored, that there is an abnormality in taking out the stored article, and the like, and the corresponding problem takes corresponding processing measures; according to the obtained storage data, the available storage cabinets are distributed and the storage articles are taken out, so that wrong taking is avoided, storage identification and analysis can be more accurately carried out, and the problem that the storage articles cannot be taken out due to forgetting the storage cabinet password or losing the storage cabinet bar code in the prior art is avoided; the storage identification analysis accuracy is improved.

Further, as shown in fig. 2, a schematic structural diagram of a storage data acquisition module in an intelligent storage identification analysis system provided in an embodiment of the present application is provided, where the storage data acquisition module includes a storage image data acquisition unit, a user feature data acquisition unit, a size data acquisition unit, and a category data acquisition unit; a storage image data acquisition unit: the storage image data are used for recording pictures in a preset range outside the storage cabinet to be monitored in a monitoring time period; user feature data acquisition unit: the method comprises the steps of acquiring user characteristic data of a storage cabinet to be monitored, wherein the user characteristic data are used for describing characteristics of storage clients of the storage cabinet to be monitored in a monitoring time period; size data acquisition unit: the method comprises the steps of acquiring article size data of a storage cabinet to be monitored, wherein the article size data are used for describing external characteristic conditions of articles stored in the storage cabinet to be monitored in a monitoring time period, and specifically comprise length, width, height and weight; category data acquisition unit: the storage cabinet storage type is used for acquiring article type data of the storage cabinet to be monitored, wherein the article type data are used for describing storage types of articles stored in the storage cabinet to be monitored in a monitoring time period, and the storage types comprise freezing storage and normal-temperature storage.

In this embodiment, the user feature data specifically includes user face feature data and user audio feature data, where the user face feature data is used to describe a face feature of a storage client of the storage cabinet to be monitored, and the user audio feature data is used to describe an audio feature of the storage client of the storage cabinet to be monitored, and specifically may include other types of data in combination with actual situations; the user characteristic data is used for checking the identity of the storage client when the storage object is taken out, so that the problem of trouble in taking out the storage object caused by the loss of the bar code under the general condition is avoided; the storage image data are used for monitoring that no large-area shielding exists in an external preset range so as to ensure the normal use of the storage cabinet to be monitored; the storage cabinet to be monitored can be provided with storage spaces with different sizes, and a part of storage cabinets with freezing functions are arranged according to requirements, so that the storage customers can use the storage cabinets conveniently, and meanwhile, the utilization efficiency is improved; the storage work of the storage cabinet to be monitored is managed more efficiently.

Further, as shown in fig. 3, a schematic structural diagram of a storage identification analysis module in an intelligent storage identification analysis system provided in an embodiment of the present application is provided, where the storage identification analysis module includes a shielding identification unit, a user feature identification unit and an available locker analysis unit; occlusion recognition unit: the storage cabinet shielding model is used for detecting the article shielding degree in the preset range outside the storage cabinet to be monitored in the monitoring time period by combining the received storage image data with the trained storage cabinet shielding model, and the storage cabinet shielding model is used for detecting the preset shielding object in the input storage image data; user feature recognition unit: the storage client characteristic recognition module is used for recognizing and storing the characteristics of the storage client of the storage cabinet to be monitored in the monitoring time period by combining the received user characteristic data with the trained user characteristic recognition module, and carrying out characteristic comparison when the storage client takes out the storage object; available locker analysis unit: the storage cabinet monitoring system is used for analyzing the corresponding available storage cabinets in the storage cabinets to be monitored in the monitoring time period by combining the received article size data and the article category data, and opening the distributed available storage cabinets for storage of storage clients according to the analysis result, wherein the available storage cabinets are storage cabinets meeting the storage requirements of the storage clients.

In this embodiment, the judging of the shielding degree considers from two aspects, firstly, whether a preset shielding object exists or not, and secondly, the influence of the existing preset shielding object on the storage client to use the storage cabinet to be monitored; the preset shielding object is other objects of non-storage clients; the storage work of identifying and analyzing the storage cabinet to be monitored more accurately and efficiently is realized by detecting and identifying the corresponding information through the storage cabinet shielding model and the user characteristic identification model.

Further, the specific training method of the locker shielding model is as follows: constructing a historical storage image data set, wherein the historical storage image data set is a set of historical storage image data, and the historical storage image data is used for recording pictures in a preset range outside a storage cabinet to be monitored in a historical time period; dividing a historical storage image data set according to a preset proportion to obtain a historical storage image training set and a historical storage image verification set, and carrying out preset shielding object labeling on the historical storage image data in the historical storage image training set; training a first locker shielding model, verifying the first locker shielding model through a historical storage image verification set, wherein the first locker shielding model is used for describing a converged model obtained by training and optimizing a preset detection model through a marked historical storage image training set, and the preset detection model is used for describing a model for realizing target detection based on deep learning; the method comprises the steps of evaluating shielding detection performance of a first locker shielding model, screening the first locker shielding model through shielding detection performance to obtain the locker shielding model, and evaluating shielding detection performance by obtaining shielding detection performance scores of the first locker shielding model, wherein the shielding detection performance scores are used for describing detection accuracy degree and detection real-time degree of the first locker shielding model.

In this embodiment, the preset detection model is a model with excellent performance commonly used in the target detection model; the screening process for the first locker shielding model is approximately as follows: comparing the shielding detection performance of the shielding model of the first locker with the reference shielding detection performance to obtain a model which is better than the reference shielding detection performance, selecting a model with the best shielding detection performance from the models, if only one model is used, selecting the model as the shielding model of the locker, otherwise, continuing to screen from the angles of model size, training time and the like until a unique model is obtained, and if the shielding detection performance of the shielding model of the first locker obtained through training is worse than the reference shielding detection performance, reselecting a preset detection model for training, thereby realizing more accurate detection of shielding conditions in a preset range outside the locker to be monitored.

Further, the specific acquisition process of the shielding detection performance score of the shielding model of the first locker is as follows: acquiring historical storage image verification time and historical storage image verification data, wherein the historical storage image verification time is used for describing the time for detecting each historical storage image data in a historical storage image verification set through a first locker shielding model, and the historical storage image verification data is used for describing a result obtained by detecting each historical storage image data in the historical storage image verification set through the first locker shielding model; acquiring reference historical storage image verification data, wherein the reference historical storage image verification data is used for describing a result obtained by manually marking preset shielding objects of each historical storage image data in the historical storage image verification set; the method comprises the steps of obtaining shielding detection weights, wherein the shielding detection weights are used for describing the influence degree of detection accuracy relative deviation and detection real-time relative deviation of a preset shielding object detected through a first locker shielding model on shielding detection performance, and the influence degree comprises detection accuracy weights and detection real-time weights; the method comprises the steps of obtaining a detection accuracy score and a detection instantaneity score of a first locker shielding model, and obtaining the shielding detection performance score of the first locker shielding model by combining shielding detection weights, wherein the detection accuracy score is used for describing the detection accuracy degree of a first locker detection model for detecting a preset shielding object, and the detection instantaneity score is used for describing the detection instantaneity degree of the first locker detection model for detecting the preset shielding object.

In this embodiment, the occlusion detection weight may be obtained according to an objective weighting method, or subjective weighting may be performed by a related person; the detection accuracy and the detection instantaneity are important for the detection of the preset shielding object, the management workload is increased if the detection is inaccurate, and the shielding object can not be processed in time if the detection is not timely; the shielding detection performance of the shielding model of the first locker is evaluated more accurately.

Further, the occlusion detection performance score of the first locker occlusion model is calculated using the following formula:in which, in the process,numbering the shielding model of the first locker, +.>，/>The total number of occlusion models for the first locker,is->Occlusion detection performance score of the first locker occlusion model,/->And->Respectively +.>Detecting accuracy score and detecting real-time score of shielding model of first locker, and (E) detecting real-time score of shielding model of first locker>And->A reference detection accuracy score and a reference detection real-time score, respectively, < >>And->Correction factors for the detection accuracy score and the detection real-time score, respectively, < >>And->The detection accuracy weight and the detection real-time weight are respectively.

In this embodiment, the detection accuracy degree and the detection real-time degree can be obtained by evaluating by a professional in combination with a confusion matrix, accuracy and recall rate of the first locker shielding model, and a more accurate result can be obtained by calculation through a formula; the detection accuracy score was calculated using the following formula: Wherein->Is->F1 fraction of the first locker shielding model,>centralizing historic stores for historic store image verificationNumbering of object image data->，/>Validating the total amount of historical storage image data in the set for the historical storage image, +.>Is->Historical storage image verification set detected by first locker shielding model +.>Cross-reference data of historical storage image data,/-data of the historical storage image data>And->The weights of the F1 score and the average cross ratio data relative to the detection accuracy score are respectively set; the detection real-time score is calculated by adopting the following formula:wherein->Is->Historical storage image verification set detected by first locker shielding model +.>Time of verification of historical storage image, +.>Verifying time for the reference historical storage image; when the shielding detection performance only considers the detection accuracy and the detection real-time performance, the method is suitable forThe sum of the corresponding detection accuracy weight and the detection real-time weight is 1, when the actual detection accuracy score and the detection real-time score are equal to the corresponding reference values, namely +.>，/>The corresponding shielding detection performance score is 1, and the shielding detection performance of the corresponding first locker shielding model is optimal; the detection accuracy score is in direct proportion to the F1 score and the average cross ratio data; detecting that the real-time score is inversely proportional to a minimum value of the historical storage image verification time; the shielding detection performance of the shielding model of the first locker is evaluated more accurately.

Further, the specific training method of the user characteristic recognition model is as follows: constructing a reference user characteristic data set, wherein the reference user characteristic data set is a set of reference user characteristic data, the reference user characteristic data comprises historical user characteristic data and generated user characteristic data, the historical user characteristic data is used for describing characteristics of storage clients of a storage cabinet to be monitored in a historical time period, and the generated user characteristic data is used for describing characteristics of the storage clients generated according to the historical user characteristic data; dividing the reference user characteristic data set according to a preset proportion to obtain a reference user characteristic training set and a reference user characteristic verification set, and labeling the reference user characteristic training set; training a first user characteristic recognition model, verifying the first user characteristic recognition model by referring to a user characteristic verification set, wherein the first user characteristic recognition model is used for describing a converged model obtained by training and optimizing a preset recognition model by referring to a user characteristic training set after marking, and the preset recognition model is used for describing a model for realizing user characteristic recognition based on multiple modes; and evaluating the user recognition performance of the first user feature recognition model, screening the first user feature recognition model through the user recognition performance to obtain the user feature recognition model, wherein the user recognition performance is evaluated by obtaining the user recognition performance score of the first user feature recognition model, and the user recognition performance score is used for describing the recognition accuracy degree and the recognition instantaneity degree of the first user feature recognition model.

In this embodiment, the generation of the user feature data may imitate the features of the corresponding storage client in different environments, such as wearing mask, wearing cap, wearing glasses or light and shade change, etc., and the increase of the data set also improves the robustness of the model; screening the first user characteristic recognition model basically the same as the screening method of the first locker shielding model, and screening a unique model by combining the reference user recognition performance, the model size, the training time and the like; the storage client of the storage cabinet to be monitored is identified more accurately.

Further, the specific acquisition process of the user identification performance score of the first user feature identification model is as follows: acquiring reference user feature verification data and reference user feature verification time, wherein the reference user feature verification data is used for describing a result obtained by identifying each reference user feature data in a reference user feature verification set through a first user feature identification model, and the reference user feature verification time is used for describing time for identifying each reference user feature data in the reference user feature verification set through the first user feature identification model; acquiring user identification weights, wherein the user identification weights are used for describing the influence degree of the absolute deviation of identification accuracy and the absolute deviation of identification instantaneity of the user characteristics on the performance of the user characteristics, including the identification accuracy weights and the identification instantaneity weights, of the user characteristics through a first user characteristic identification model; acquiring identification accuracy scores of the first user feature identification model by combining preset user feature verification data, wherein the identification accuracy scores are used for describing the accuracy degree of the first user feature identification model for identifying the user features; acquiring an identification real-time score of the first user feature identification model according to a preset verification time, wherein the identification real-time score is used for describing the real-time degree of the first user feature identification model for identifying the user features; and calculating the user recognition performance score of the first user characteristic recognition model according to the recognition accuracy score and the recognition real-time score and combining the user recognition weight.

In this embodiment, the evaluation of the user recognition performance of the user recognition weight corresponding to the first user feature recognition model is important, and may be obtained by an objective weighting method or a subjective weighting method; the identification accuracy ensures that the stored articles cannot be taken by mistake, and the real-time identification ensures that the stored articles can be taken out in time; the method and the device realize more efficient identification and analysis of the characteristics of the storage clients of the storage cabinet to be monitored.

Further, the user recognition performance score of the first user feature recognition model is calculated using the following formula:wherein->Identifying the number of the model for the first user feature, +.>，/>The total number of models is identified for the first user feature,is->User identification performance score of the first user feature identification model,/for each user feature>And->Respectively +.>Identification accuracy score and identification real-time score of the first user feature identification model,/->And->A reference identification accuracy score and a reference identification real-time score, respectively, < >>And->Correction factors for the recognition accuracy score and the recognition real-time score, respectively, < >>And->Respectively, an identification accuracy weight and an identification real-time weight.

In this embodiment, the recognition accuracy degree and the recognition real-time degree can be obtained by the professional evaluating the accuracy, throughput and delay of the first user feature recognition model, and can also obtain a more accurate result by calculating through a formula; the recognition accuracy score is calculated using the following formula: Wherein->Is the firstArea under ROC curve data of first user characteristic recognition model +.>And->Area data under the first reference ROC curve and area data under the second ROC curve respectively, and +.>The method comprises the steps of carrying out a first treatment on the surface of the The first reference ROC curve area under data and the second ROC curve area under data are used to describe the ROC curve area under data of the first user identification modelAccording to the reference range; the identification real-time score is calculated by the following formula:wherein->Reference user feature data number in reference user feature verification set,/-for reference user feature verification set>，/>Total number of reference user feature data in the reference user feature verification set, +.>Is->First user feature identification model identified reference user feature verification set +.>Reference user characteristic verification time corresponding to the reference user characteristic data,/->And->The first preset verification time and the second preset verification time are respectively, and +.>The method comprises the steps of carrying out a first treatment on the surface of the The first preset verification time and the second preset verification time are used for describing a reference range of the preset verification time; the user identification performance score is in direct proportion to the identification accuracy score and the identification real-time score, and when only the identification accuracy and the identification real-time are considered, the sum of the corresponding identification accuracy weight and the identification real-time weight is 1; the recognition accuracy score is proportional to the area data under the ROC curve, when the corresponding data is 1, namely The corresponding recognition accuracy score is 1, when the corresponding data is equal to the area data under the second ROC curve, namely +.>The corresponding recognition accuracy score is +.>The method comprises the steps of carrying out a first treatment on the surface of the The user recognition performance of the first user feature recognition model is evaluated more accurately.

The intelligent storage identification analysis method provided by the embodiment of the application comprises the following steps: acquiring storage data of a storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and transmitting the acquired storage data to a storage identification analysis module and a storage monitoring center; analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to a storage monitoring center; storing storage data of the storage cabinet to be monitored, wherein the storage data comprises storage data and storage analysis results, and judging whether the storage cabinet to be monitored has storage abnormality according to the storage analysis results so as to take processing measures.

In the embodiment, the performance of the storage cabinet shielding model and the user characteristic recognition model by means of storage recognition analysis is improved along with the increase of corresponding training sets; when a storage customer takes out a storage article, the corresponding storage data can be deleted so as to reduce the data storage pressure; the storage work of the storage cabinet to be monitored is identified and analyzed more efficiently.

The technical scheme in the embodiment of the application at least has the following technical effects or advantages: relative to the bulletin number: according to the storage system based on the intelligent identification of the articles of the database, disclosed by the CN107752584B, a historical storage image data set is constructed and divided to obtain a historical storage image training set and a historical storage image verification set, then a preset detection model is trained and optimized through the marked historical storage image training set to obtain a first storage cabinet shielding model, verification is carried out through the historical storage image verification set, shielding detection performance of the first storage cabinet shielding model is estimated, and finally a storage cabinet shielding model is obtained through screening according to the shielding detection performance, so that accurate detection of preset shielding objects is achieved, and further the shielding problem in the preset range outside the storage cabinet to be monitored is processed more accurately and rapidly; relative to the bulletin number: according to the embodiment of the application, the reference user characteristic verification data and the reference user characteristic verification time are acquired, then the identification accuracy score of the first user characteristic identification model is calculated, the corresponding identification real-time score is calculated by combining with the preset verification time, the user identification performance score of the first user characteristic identification model is calculated by combining with the user identification weight, and finally the user characteristic identification model is screened according to the user identification performance score, so that the user identification performance of the first user characteristic identification model is digitized, and further the characteristics of a storage client of the storage cabinet to be monitored are accurately identified.

It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (9)

1. An intelligent storage identification analysis system is characterized by comprising a storage data acquisition module, a storage identification analysis module and a storage monitoring center;

the storage data acquisition module is used for acquiring storage data of the storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and the acquired storage data are sent to the storage identification analysis module and the storage monitoring center;

the storage identification analysis module is used for analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to the storage monitoring center;

the storage monitoring center is used for storing storage data of the storage cabinet to be monitored, wherein the storage data comprises storage data and storage analysis results, and judging whether the storage cabinet to be monitored has storage abnormality or not according to the storage analysis results so as to take processing measures;

The storage data acquisition module comprises a storage image data acquisition unit, a user characteristic data acquisition unit, a size data acquisition unit and a category data acquisition unit;

the storage image data acquisition unit: the storage cabinet monitoring method comprises the steps of acquiring storage image data of a storage cabinet to be monitored, wherein the storage image data are used for recording pictures in a preset range outside the storage cabinet to be monitored in a monitoring time period;

the user characteristic data acquisition unit: the method comprises the steps of acquiring user characteristic data of a storage cabinet to be monitored, wherein the user characteristic data are used for describing characteristics of storage clients of the storage cabinet to be monitored in a monitoring time period;

the size data acquisition unit: the method comprises the steps of acquiring article size data of a storage cabinet to be monitored, wherein the article size data are used for describing external characteristic conditions of articles stored in the storage cabinet to be monitored in a monitoring time period, and specifically comprise length, width, height and weight;

the category data acquisition unit: the storage type of the storage cabinet to be monitored is used for describing the storage type of the storage articles of the storage cabinet to be monitored in the monitoring time period, and the storage type comprises freezing storage and normal-temperature storage.

2. The intelligent storage identification analysis system of claim 1, wherein: the storage identification and analysis module comprises a shielding identification unit, a user characteristic identification unit and an available locker analysis unit;

the shielding identification unit: the storage cabinet shielding model is used for detecting the article shielding degree in a preset range outside the storage cabinet to be monitored in a monitoring time period by combining the received storage image data with a trained storage cabinet shielding model, and the storage cabinet shielding model is used for detecting a preset shielding object in the input storage image data;

the user characteristic recognition unit: the storage client characteristic recognition module is used for recognizing and storing the characteristics of the storage client of the storage cabinet to be monitored in the monitoring time period by combining the received user characteristic data with the trained user characteristic recognition module, and the user characteristic recognition module is used for recognizing and storing the characteristics of the storage client of the storage cabinet to be monitored and comparing the characteristics when the storage client takes out the storage object;

the available locker analysis unit: the storage cabinet monitoring system is used for analyzing the corresponding available storage cabinets in the storage cabinets to be monitored in the monitoring time period by combining the received article size data and the article category data, and opening the distributed available storage cabinets for storage of storage clients according to the analysis result, wherein the available storage cabinets are storage cabinets meeting the storage requirements of the storage clients.

3. The intelligent storage identification analysis system of claim 2, wherein the specific training method of the locker shielding model is as follows:

constructing a historical storage image data set, wherein the historical storage image data set is a set of historical storage image data, and the historical storage image data is used for recording pictures in a preset range outside a storage cabinet to be monitored in a historical time period;

dividing a historical storage image data set according to a preset proportion to obtain a historical storage image training set and a historical storage image verification set, and carrying out preset shielding object labeling on the historical storage image data in the historical storage image training set;

training a first locker shielding model, and verifying the first locker shielding model through a historical storage image verification set, wherein the first locker shielding model is used for describing a converged model obtained by training and optimizing a preset detection model through a marked historical storage image training set, and the preset detection model is used for describing a model for realizing target detection based on deep learning;

the method comprises the steps of evaluating shielding detection performance of a first locker shielding model, screening the first locker shielding model through shielding detection performance to obtain the locker shielding model, and evaluating shielding detection performance by obtaining shielding detection performance scores of the first locker shielding model, wherein the shielding detection performance scores are used for describing detection accuracy degree and detection real-time degree of the first locker shielding model.

4. The intelligent storage identification analysis system of claim 3, wherein the specific acquisition process of the occlusion detection performance score of the first locker occlusion model is as follows:

acquiring historical storage image verification time and historical storage image verification data, wherein the historical storage image verification time is used for describing the time for detecting each historical storage image data in a historical storage image verification set through a first locker shielding model, and the historical storage image verification data is used for describing a result obtained by detecting each historical storage image data in the historical storage image verification set through the first locker shielding model;

acquiring reference historical storage image verification data, wherein the reference historical storage image verification data is used for describing results obtained by manually marking preset shielding objects of each historical storage image data in the historical storage image verification set;

the method comprises the steps of obtaining a shielding detection weight, wherein the shielding detection weight is used for describing the influence degree of detection accuracy relative deviation and detection real-time relative deviation of a preset shielding object detected through a first locker shielding model on shielding detection performance, and comprises the detection accuracy weight and the detection real-time weight;

The method comprises the steps of obtaining a detection accuracy score and a detection instantaneity score of a first locker shielding model, and obtaining a shielding detection performance score of the first locker shielding model by combining shielding detection weights, wherein the detection accuracy score is used for describing the detection accuracy degree of a first locker detection model for detecting a preset shielding object, and the detection instantaneity score is used for describing the detection instantaneity degree of the first locker detection model for detecting the preset shielding object.

5. The intelligent storage identification analysis system of claim 4, wherein the occlusion detection performance score of the first locker occlusion model is calculated using the formula:

，

in the method, in the process of the invention,numbering the shielding model of the first locker, +.>，/>For the total number of first locker occlusion models, +.>Is->Occlusion detection performance score of the first locker occlusion model,/->And->Respectively +.>Detecting accuracy score and detecting real-time score of shielding model of first locker, and (E) detecting real-time score of shielding model of first locker>And->A reference detection accuracy score and a reference detection real-time score, respectively, < >>And->Correction factors for the detection accuracy score and the detection real-time score, respectively, < >>And->The detection accuracy weight and the detection real-time weight are respectively.

6. The intelligent storage identification analysis system of claim 2, wherein the specific training method of the user feature identification model is as follows:

constructing a reference user characteristic data set, wherein the reference user characteristic data set is a set of reference user characteristic data, the reference user characteristic data comprises historical user characteristic data and generated user characteristic data, the historical user characteristic data is used for describing characteristics of storage clients of a storage cabinet to be monitored in a historical time period, and the generated user characteristic data is used for describing characteristics of the storage clients generated according to the historical user characteristic data;

dividing the reference user characteristic data set according to a preset proportion to obtain a reference user characteristic training set and a reference user characteristic verification set, and labeling the reference user characteristic training set;

training a first user characteristic recognition model, and verifying the first user characteristic recognition model by referring to a user characteristic verification set, wherein the first user characteristic recognition model is used for describing a converged model obtained by training and optimizing a preset recognition model by referring to a labeled reference user characteristic training set, and the preset recognition model is used for describing a model for realizing user characteristic recognition based on multiple modes;

And evaluating the user recognition performance of the first user feature recognition model, screening the first user feature recognition model through the user recognition performance to obtain the user feature recognition model, wherein the user recognition performance is evaluated by obtaining the user recognition performance score of the first user feature recognition model, and the user recognition performance score is used for describing the recognition accuracy degree and the recognition instantaneity degree of the first user feature recognition model.

7. The intelligent storage identification analysis system of claim 6, wherein the specific acquisition process of the user identification performance score of the first user characteristic identification model is as follows:

acquiring reference user feature verification data and reference user feature verification time, wherein the reference user feature verification data is used for describing a result obtained by identifying each piece of reference user feature data in a reference user feature verification set through a first user feature identification model, and the reference user feature verification time is used for describing time for identifying each piece of reference user feature data in the reference user feature verification set through the first user feature identification model;

acquiring user identification weights, wherein the user identification weights are used for describing the influence degree of the identification accuracy absolute deviation and the identification real-time absolute deviation of the user characteristics on the user characteristic performance through a first user characteristic identification model, and the influence degree comprises the identification accuracy weights and the identification real-time weights;

Acquiring identification accuracy scores of a first user characteristic identification model by combining preset user characteristic verification data, wherein the identification accuracy scores are used for describing the accuracy degree of the first user characteristic identification model for identifying user characteristics;

acquiring an identification real-time score of a first user characteristic identification model by combining with a preset verification time, wherein the identification real-time score is used for describing the real-time degree of the user characteristic identification model for identifying the user characteristic;

and calculating the user recognition performance score of the first user characteristic recognition model according to the recognition accuracy score and the recognition real-time score and combining the user recognition weight.

8. The intelligent storage identification analysis system of claim 7 wherein the user identification performance score of the first user characteristic identification model is calculated using the formula:

，

in the method, in the process of the invention,identifying the number of the model for the first user feature, +.>，/>Identifying a total number of models for the first user feature, < >>Is->User identification performance score of the first user feature identification model,/for each user feature>Andrespectively +.>Identification accuracy score and identification real-time score of the first user feature identification model,/->Anda reference identification accuracy score and a reference identification real-time score, respectively, < > >And->Correction factors for the recognition accuracy score and the recognition real-time score, respectively, < >>And->Respectively, an identification accuracy weight and an identification real-time weight.

9. A method for use in an intelligent storage identification analysis system according to any one of claims 1 to 8, comprising the steps of:

the method comprises the steps of obtaining storage data of a storage cabinet to be monitored, wherein the storage data are used for describing information of articles stored in the storage cabinet to be monitored in a monitoring time period, and sending the obtained storage data to a storage identification analysis module and a storage monitoring center;

analyzing the storage condition of the storage cabinet to be monitored in the monitoring time period according to the received storage data to obtain a storage analysis result, and sending the obtained storage analysis result to a storage monitoring center;

and storing storage data of the storage cabinet to be monitored, wherein the storage data comprises storage data and storage analysis results, and judging whether the storage cabinet to be monitored has storage abnormality according to the storage analysis results so as to take processing measures.