Intelligent roller shutter door type compact shelf and control system thereof

Abstract

The invention belongs to the technical field of archive management, and provides an intelligent roller shutter door type compact shelf and a control system thereof in order to solve the problem that data management and searching of an existing compact shelf are inconvenient. The control system comprises a data preprocessing module, a classification module, a retrieval preprocessing module, a retrieval module, an execution control module and the like. The data preprocessing module sets a first-level retrieval anchor point and a second-level retrieval anchor point according to the weight of the keyword; clustering the first-level retrieval anchor points to generate a set, establishing association between the second-level retrieval anchor points and the first-level retrieval set, processing the retrieval anchor points by using a cloud transformation algorithm, generating a query cloud based on user query, performing multi-level matching, sorting and cloud overlapping region probability integral retrieval archives, and performing positioning and outputting an archive detection instruction by an execution control module. The intelligent roller shutter door type compact shelf is provided with a user interaction system and a shelf body opening and closing system and works in cooperation with the control system. According to the method, automatic classification and multi-stage retrieval of the archives are realized, the retrieval precision and comprehensiveness are improved, and the archive management efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of file management, and relates to an intelligent roller shutter type compact shelf and a control system thereof. BACKGROUND

[0002] The compact shelf, also known as a compact cabinet, is a kind of compact file storage device, which increases the file storage capacity per unit area of a warehouse and is widely used in government, enterprise and institution libraries, sample rooms and the like for storing books, data, archives, financial vouchers and goods. Compared with traditional book cabinets, book shelves, shelves and file shelves, the compact shelf has a large storage capacity and is more systematic. However, the existing compact shelf has the following problems in use: the collected data are of various types, the number of files is large, and a complex correlation cannot be established in the process of processing the archives, so a large amount of time is wasted in searching for the data, which is not conducive to the quick searching of the files. SUMMARY

[0003] To solve the problems in the background art, the application provides an intelligent roller shutter type compact shelf and a control system thereof.

[0004] To achieve the above object, the application adopts the following technical scheme: On the one hand, the application provides an intelligent roller shutter type compact shelf control system, which comprises: A data preprocessing module sets retrieval anchor points according to keywords contained in the content of archives stored in the intelligent roller shutter type compact shelf, and the retrieval anchor points are divided into first-level retrieval anchor points and second-level retrieval anchor points according to the weight of the keywords in the archives; A first classification module clusters the first-level retrieval anchor points and classifies the first-level retrieval anchor points according to semantic correlation to generate a plurality of first-level retrieval sets; A second classification module determines the correlation degree of the second-level retrieval anchor points and the first-level retrieval sets, and sets the second-level retrieval anchor points and the associated first-level retrieval sets in correlation; A retrieval preprocessing module performs cloud transformation preprocessing on the retrieval anchor points of the archives through a cloud transformation algorithm; A retrieval module generates a query cloud based on user query when the user needs to query, performs multi-level matching according to the query cloud, sorts according to the correlation degree, adds the probability integral of the cloud overlap area in the cloud transformation, realizes the retrieval of the archives, and sends the retrieval result to the user interface; An execution control module positions the archive object according to the retrieval result and outputs an archive detection instruction.

[0005] Further, the specific method for dividing the keywords in the archives into first-level retrieval anchor points and second-level retrieval anchor points is as follows: Removing all non-compileable characters including punctuation marks and special characters in all archives, and performing word segmentation on the text content of the archives; Calculating the frequency of each word appearing in the text content of all archives, the formula is: ; Wherein is the number of times the word appears in the text content of the archives, is the total number of times all words appear in the text content of the archives, is the frequency of the word appearing in the text content of the archives; ; According to the frequency of the word, the word is divided into a first-level retrieval anchor point and a second-level retrieval anchor point; The frequency of the first-level retrieval anchor point is greater than the frequency of the second-level retrieval anchor point.

[0006] Further, the specific method of clustering according to the first-level retrieval anchor point of each archive is: Collecting the first-level retrieval anchor point information of all archives in the intelligent roller shutter door type dense shelf; Generating a first-level retrieval anchor point pair from any two first-level retrieval anchor points, and calculating the semantic correlation between any two first-level retrieval anchor points using a Word2Vec model; Initializing a clustering set, regarding each first-level retrieval anchor point as a separate clustering block, traversing all first-level retrieval anchor point pairs, and calculating the semantic correlation between any two first-level retrieval anchor points, if the semantic correlation of a first-level retrieval anchor point pair is higher than a first correlation threshold, merging the clustering blocks where the two first-level retrieval anchor points are located; Using the Word2Vec model to calculate the semantic correlation between any two clustering blocks for each clustering block, if the semantic correlation of two clustering blocks is higher than the first correlation threshold, merging them, repeating the above steps until there is no clustering block that can be merged; After clustering is completed, each clustering block obtained is a first-level retrieval set.

[0007] Further, the specific method of setting the association between the second-level retrieval anchor point and the associated first-level retrieval set is: For each second-level retrieval anchor point, using a Word2Vec model to calculate the correlation between the second-level retrieval anchor point and each first-level retrieval set; Traversing all second-level retrieval anchor points, for each second-level retrieval anchor point, establishing an association relationship with the first-level retrieval set whose correlation is greater than a second correlation threshold.

[0008] ​Furthermore, the specific method of performing cloud transformation preprocessing on the retrieval anchor points of the archives through the cloud transformation algorithm is as follows: A cloud model is constructed for each first-level retrieval anchor point. The specific expression is: ; in For the A cloud model of first-level retrieval anchor points, For the The expected value of the cloud model function of the first-level retrieval anchor point; For the The entropy of the cloud model function of the first-level retrieval anchor point; For the The hyperentropy of the cloud model function of the first-level retrieval anchor; Define the cloud membership function for each cloud model of the first-level retrieval anchor , the expression is: ; in is the random entropy, The expected value of the cloud model function for a certain level of retrieval anchor point, Retrieve the anchor point vector for this level.

[0009] Furthermore, the specific method of the user query is: Generate query cloud parameters based on user queries. The expression is: ; in To query the expected value of the cloud parameter; To query the entropy of cloud parameters; To query the hyperentropy of cloud parameters, The code for the user's query content. The total number of parameters in the user query content. For the query The embedding vector of keywords; Calculate the similarity between the query cloud parameters and the first-level retrieval anchor. The expression is: ; in The user's query content Hedi First-level search collection The first-level anchor matching threshold; First-level retrieval anchor Query content with users The cloud membership of First-level retrieval anchor With the a primary search set cloud membership of the file; Based on the similarity results of the query cloud parameters and the primary search anchor points, the secondary search anchor points are expanded, the similarity of the query cloud parameters and the secondary search anchor points is calculated, and the expression is: ; Wherein the secondary search anchor point The first primary search set The secondary anchor point association threshold value, is sigmoid The function, is the weight matrix; Based on the similarity results of the query cloud parameters and the secondary search anchor points, the first preset value of the candidate archives is extracted; Join the probability integral of the cloud overlap area , the expression is: ; Wherein the cloud membership of the file and the primary search anchor point; The first preset value of the candidate archives is sorted by the probability integral of the cloud overlap area, and the file required by the user is filtered out.

[0010] On the other hand, the application also provides an intelligent roller shutter door type dense shelf loaded with the intelligent roller shutter door type dense shelf control system, further comprising: User interaction system, through the input module of the touch screen, receiving the content and the location information of the stored file input by the user when placing the file, and through the touch screen, receiving the query content of the user, and through the user interface on the touch screen, feeding back the query result of the user; Memory, storing the content and the location information of the stored file input by the user when placing the file; The intelligent roller shutter door type dense shelf control system receives the user query command of the user interaction system and executes, and feeds back the search result; Shelf opening and closing system, installed between adjacent two groups of dense shelf bodies, adjusting the positional relationship between the dense shelf bodies through the motor; The execution control module of the intelligent roller shutter door type dense shelf control system controls the shelf opening and closing system according to the file detection instruction.

[0011] Compared with the prior art, the application has the following beneficial effects: The application realizes automatic classification and multi-level search of files through keyword extraction, clustering algorithm and semantic correlation analysis.​

[0012] The application is based on cloud transformation algorithm and cloud membership function, and realizes high-precision file matching by combining query cloud parameters generated by user query.

[0013] The application dynamically expands the search range through the association mechanism of the first-level search anchor point and the second-level search anchor point, and improves the comprehensiveness of the query. BRIEF DESCRIPTION OF DRAWINGS

[0014] figure 1 is a structural schematic diagram of the intelligent roller shutter door type compact shelf control system described in the application; figure 2 is a running flow chart of the intelligent roller shutter door type compact shelf control system described in the application; figure 3 is a structural schematic diagram of the intelligent roller shutter door type compact shelf described in the application. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0016] On the one hand, as shown in figure 3 The application further provides an intelligent roller shutter door type compact shelf, which comprises: A user interaction system receives the content of the stored file and the location information of the placed file input by the user when placing the file through an input module of a touch screen, and receives the query content of the user through the touch screen, and feeds back the query result of the user through a user interface on the touch screen; A storage stores the content of the stored file and the location information of the placed file input by the user when placing the file; The user interaction system of the intelligent roller shutter door type compact shelf receives the content of the stored file and the location information of the placed file input by the user when placing the file, and stores them in the storage.

[0017] The intelligent roller shutter door type compact shelf control system receives the user query command of the user interaction system and executes, and feeds back the search result; A shelf opening and closing system is installed between two adjacent groups of compact shelf bodies, and adjusts the positional relationship between the compact shelf bodies through a motor; The execution control module of the intelligent roller shutter door type compact shelf control system controls the shelf opening and closing system according to the file detection instruction.

[0018] On the other hand, as shown in figures 1-2As shown, the present application provides a smart roller shutter type dense shelf control system, comprising: The data preprocessing module sets retrieval anchor points according to keywords contained in the content of the archives stored in the smart roller shutter type dense shelf, and the retrieval anchor points are divided into first-level retrieval anchor points and second-level retrieval anchor points according to the weight of the keywords in the archives. The first classification module clusters according to the first-level retrieval anchor points, classifies the first-level retrieval anchor points according to semantic correlation, and generates a plurality of first-level retrieval sets. The second classification module determines the correlation degree of the second-level retrieval anchor points and the first-level retrieval sets, and sets the association of the second-level retrieval anchor points and the associated first-level retrieval sets. The retrieval preprocessing module performs cloud transformation preprocessing on the retrieval anchor points of the archives through a cloud transformation algorithm. The retrieval module generates a query cloud based on user queries when the user needs to query, performs multi-level matching according to the query cloud, sorts according to the correlation degree, adds the probability integral of the cloud overlap area in the cloud transformation, realizes the retrieval of the archives, and sends the retrieval results to the user interface. The retrieval results include the content information of the archives corresponding to the user query content and the location information of the corresponding archives. The execution control module positions the archive object according to the retrieval results and outputs an archive detection instruction.

[0019] The data preprocessing module first sets retrieval anchor points according to keywords contained in the content of the archives stored in the smart roller shutter type dense shelf in the storage.

[0020] All non-compileable characters including punctuation marks and special characters in all archives are removed, and the text content of the archives is processed by word segmentation. Removing non-compileable characters and word segmentation processing makes the text data more regular, which is convenient for subsequent analysis. The frequency of each word appearing in the text content of all archives is calculated, and the formula is: ; Wherein is the number of times the word appears in the text content of the archives, is the total number of times all words appear in the text content of the archives, is the frequency of the word appearing in the text content of the archives. According to the frequency of the word, the word is divided into a first-level retrieval anchor point and a second-level retrieval anchor point.

[0021] ​​The frequency of the first-level search anchor point is greater than the frequency of the second-level search anchor point. By distinguishing the first-level and second-level search anchor points through frequency, important words can be highlighted, laying the foundation for efficient search, and high-frequency words are more likely to represent the core content of archives, which helps to quickly locate related archives.

[0022] The first classification module clusters according to the first-level search anchor points, classifies the first-level search anchor points according to semantic relevance, and generates a plurality of first-level search sets.

[0023] Collect the first-level search anchor point information of all archives in the intelligent roller shutter door type compact shelf.

[0024] Any two first-level search anchor points generate a first-level search anchor point pair, and the semantic relevance between any two first-level search anchor points is calculated using the Word2Vec model. Using the Word2Vec model to calculate semantic relevance for clustering can classify first-level search anchor points with similar semantics into a category, forming a first-level search set.

[0025] Initialize the cluster set, treat each first-level search anchor point as a separate cluster block, traverse all first-level search anchor point pairs, calculate the semantic relevance between any two first-level search anchor points, and if the semantic relevance of a first-level search anchor point pair is higher than the first correlation threshold, merge the cluster blocks where the two first-level search anchor points are located. Complete the preliminary merging of the first-level search anchor points, and subsequently merge the cluster blocks again.

[0026] For each cluster block, use the Word2Vec model to calculate the semantic relevance between any two cluster blocks, and if the semantic relevance of two cluster blocks is higher than the first correlation threshold, merge them. Repeat the above steps until there are no cluster blocks that can be merged.

[0027] After clustering is completed, each cluster block obtained is a first-level search set. This helps to narrow the scope during search and improve search efficiency, because archives within the same set have similarity in theme or semantics, facilitating quick positioning to related archive groups.

[0028] The second classification module determines the association degree of the first-level search set according to the second-level search anchor point, and sets the association between the second-level search anchor point and the associated first-level search set. The association between the second-level search anchor point and the first-level search set is established, enriching the search system.

[0029] For each second-level search anchor point, the Word2Vec model is used to calculate the relevance of the second-level search anchor point to each first-level search set.

[0030] Traverse all secondary search anchors and establish an association between each secondary search anchor and a primary search set with a relevance greater than the second relevance threshold. Secondary search anchors serve as more detailed search dimensions. When a user query involves these secondary features, the association allows for quick retrieval to the corresponding primary search set, thereby finding relevant archives and improving search accuracy.

[0031] The retrieval preprocessing module performs cloud transformation preprocessing on the retrieval anchor points of the archives through the cloud transformation algorithm. Cloud transformation preprocessing converts the retrieval anchor points into cloud models, taking into account the uncertainty and fuzziness of the data.

[0032] A cloud model is constructed for each first-level retrieval anchor point. The specific expression is: ; in For the A cloud model of first-level retrieval anchor points, For the The expected value of the cloud model function of the first-level retrieval anchor point; For the The entropy of the cloud model function of the first-level retrieval anchor point; For the The hyperentropy of the cloud model function of the first-level retrieval anchor; Define the cloud membership function for each cloud model of the first-level retrieval anchor , the expression is: ; in is the random entropy, The expected value of the cloud model function for a certain level of retrieval anchor point, Retrieve the anchor point vector for this level.

[0033] The cloud membership function can more flexibly measure the degree of association between search terms and archives. Compared with traditional exact matching, it can solve the ambiguity problem of semantic understanding to a certain extent and improve the accuracy and adaptability of retrieval.

[0034] When a user needs to make a query, the retrieval module generates a query cloud based on the user query, performs multi-level matching based on the query cloud, sorts based on relevance, adds the probability integral of the cloud overlapping area in the cloud transformation, realizes the retrieval of archives, and sends the retrieval results to the user interface. The retrieval results include the content information of the archive corresponding to the user query content and the location information of the corresponding archive. The content information of the archive corresponding to the user query content and the location information of the corresponding archive are obtained by reading the memory in the intelligent rolling door type compact shelving.

[0035] Generate query cloud parameters based on user queries. The expression is: ; wherein is the expected value of the query cloud parameter; is the entropy of the query cloud parameter; is the hyper entropy of the query cloud parameter, is the encoding of the query content of the user, is the total number of parameters in the query content of the user, is the embedding vector of the th keyword in the query. Generating the query cloud parameter according to the user query can adapt to the specific retrieval requirements of the user to retrieve the archives, and improve the accuracy in retrieval.

[0036] The similarity between the query cloud parameter and the first-level retrieval anchor point is calculated, and the expression is: ; wherein is the query content of the user and the first first-level retrieval set first-level anchor point matching threshold; is the cloud membership of the first-level retrieval anchor point and the query content of the user , is the cloud membership of the first-level retrieval anchor point and the first first-level retrieval set . By calculating the similarity between the query cloud parameter and the first-level retrieval anchor point based on the user query, the relevant first-level retrieval anchor point is screened out, the subsequent retrieval range is narrowed, and the efficiency of the user in retrieving the archives is improved.

[0037] Based on the similarity result of the query cloud parameter and the first-level retrieval anchor point, the second-level retrieval anchor point is expanded, the similarity between the query cloud parameter and the second-level retrieval anchor point is calculated, and the expression is: ; wherein is the second-level retrieval anchor point about and the second-level anchor point associated threshold of the first first-level retrieval set , is a sigmoid function, is a weight matrix. Based on the retrieval result of the first-level retrieval anchor point, the similarity analysis of the second-level retrieval anchor point is added, and the retrieval range of the archives is further narrowed.

[0038] Based on the similarity results of the query cloud parameters and the secondary retrieval anchor points, the first preset number of candidate archives are extracted. The similarity between the query cloud and the retrieval anchor points is calculated through the above formula, the primary and secondary retrieval anchor points are comprehensively considered, the characteristics of the cloud model are utilized, and the semantic association between the user query and the archives is comprehensively measured.

[0039] Join the probability integral of the cloud overlap area , the expression is: ; Among them is the cloud membership degree of the primary retrieval anchor point and the archive .

[0040] The first preset number of candidate archives are sorted by the probability integral of the cloud overlap area, and the archives required by the user are screened out. The cloud overlap area probability integral further optimizes the sorting, makes the retrieval result more meet the user's demand, improves the accuracy and efficiency of the retrieval, and provides more accurate archive retrieval service for the user.

[0041] The execution control module positions the archive object according to the retrieval result and outputs the archive retrieval instruction. The archive is quickly and accurately positioned and the retrieval instruction is output, realizing seamless connection from retrieval to actual operation, improving the efficiency of archive management, and facilitating the user to obtain the required archives.

[0042] The technical scheme provided by the application aims at efficiently managing and querying the archives stored in the dense shelf. The retrieval anchor points are set through the data preprocessing module, then clustered and associated through the classification module, cloud transformation preprocessing is performed through the retrieval preprocessing module, finally accurate retrieval is realized based on the user query through the retrieval module, and the archive retrieval instruction is positioned and output through the execution control module.

[0043] Although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to part of the technical features, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. An intelligent rolling door type compact shelving control system, characterized in that: Includes: A data preprocessing module sets search anchors based on keywords contained in the contents of the archives stored in the intelligent rolling door type compact shelving, wherein the search anchors are divided into primary search anchors and secondary search anchors based on the weights of the keywords in the archives; The first classification module clusters the first-level search anchors and classifies the first-level search anchors according to semantic relevance to generate multiple first-level search sets; The second classification module determines the degree of association between the secondary search anchor and the primary search set according to the secondary search anchor, and associates the secondary search anchor with the associated primary search set; The retrieval preprocessing module performs cloud transformation preprocessing on the retrieval anchor points of the archives through the cloud transformation algorithm; The retrieval module generates a query cloud based on the user's query when the user needs to query, performs multi-level matching based on the query cloud, sorts the results based on relevance, incorporates the probability integral of the cloud overlap area in the cloud transformation, realizes the retrieval of the archives, and sends the retrieval results to the user interface; The execution control module locates the file object according to the search results and outputs the file checkout instruction.

2. The intelligent rolling door type compact shelving control system according to claim 1, characterized in that: The specific method of dividing the keywords into primary search anchors and secondary search anchors according to their weight in the archives is as follows: Remove all non-compilable characters including punctuation marks and special characters from all files, and perform word segmentation on the text content of the files; Calculate the frequency of each word appearing in the text content of all archives. The formula is: ； in For vocabulary Text content in the archive The number of times it appears in is the total number of times all words appear in the text content of the archive, For vocabulary Text content in the archive Frequency of occurrence in The words are divided into primary search anchors and secondary search anchors according to their frequency of occurrence; The frequency of the first-level search anchor is greater than the frequency of the second-level search anchor.

3. The intelligent rolling door type compact shelving control system according to claim 1, characterized in that: The specific method for clustering the first-level retrieval anchor points is: Collect the primary retrieval anchor information of all files in the intelligent rolling door compact shelving; Any two first-level retrieval anchors generate a first-level retrieval anchor pair, and the Word2Vec model is used to calculate the semantic relevance between any two first-level retrieval anchors; Initialize the cluster set, treat each first-level retrieval anchor point as a separate cluster block, traverse all first-level retrieval anchor point pairs, calculate the semantic relevance between any two first-level retrieval anchor points, and if the semantic relevance of a first-level retrieval anchor point pair is higher than a first relevance threshold, merge the cluster blocks where the two first-level retrieval anchor points are located; The Word2Vec model is used to calculate the semantic relevance between any two cluster blocks for each cluster block. If the semantic relevance of two cluster blocks is higher than the first relevance threshold, they are merged and the above steps are repeated until there are no more cluster blocks that can be merged. After clustering is completed, each cluster block obtained is a first-level retrieval set.

4. The intelligent rolling door type compact shelving control system according to claim 1, characterized in that: The specific method of associating the secondary search anchor with the associated primary search set is: For each secondary retrieval anchor point, the Word2Vec model is used to calculate the relevance between the secondary retrieval anchor point and each primary retrieval set; All secondary search anchor points are traversed, and for each secondary search anchor point, an association relationship is established between it and a primary search set whose relevance is greater than a second relevance threshold.

5. The intelligent rolling door type compact shelving control system according to claim 1, characterized in that: The specific method of performing cloud transformation preprocessing on the retrieval anchor points of archives through the cloud transformation algorithm is as follows: A cloud model is constructed for each first-level retrieval anchor point. The specific expression is: ； in For the A cloud model of first-level retrieval anchor points, For the The expected value of the cloud model function of the first-level retrieval anchor point; For the The entropy of the cloud model function of the first-level retrieval anchor point; For the The hyperentropy of the cloud model function of the first-level retrieval anchor; Define the cloud membership function for each cloud model of the first-level retrieval anchor , the expression is: ； in is the random entropy, The expected value of the cloud model function for a certain level of retrieval anchor point, Retrieve the anchor point vector for this level.

6. The intelligent rolling door type compact shelving control system according to claim 5, characterized in that: The specific method of the user query is: Generate query cloud parameters based on user queries. The expression is: ； in To query the expected value of the cloud parameter; To query the entropy of cloud parameters; To query the hyperentropy of cloud parameters, The code for the user's query content. The total number of parameters in the user query content. For the query The embedding vector of keywords; Calculate the similarity between the query cloud parameters and the first-level retrieval anchor. The expression is: ； in The user's query content Hedi First-level search collection The first-level anchor matching threshold; First-level retrieval anchor Query content with users The cloud membership of First-level retrieval anchor With the The cloud membership of the first-level search set; Based on the similarity between the query cloud parameters and the first-level search anchor, the second-level search anchor is expanded, and the similarity between the query cloud parameters and the second-level search anchor is calculated. The expression is: ； in Secondary search anchor About Hedi First-level search collection The secondary anchor point association threshold, for sigmoid function, is the weight matrix; Based on the similarity results of the query cloud parameters and the secondary search anchor points, extract a first preset number of candidate archives; Probability integral including cloud overlap area , the expression is: ； in First-level retrieval anchor With archives The cloud membership of The first preset number of candidate archives are sorted by the probability integral of the cloud overlapping area to filter out the archives that the user needs to query.

7. An intelligent rolling door type compact shelving, characterized in that: The intelligent rolling door type compact shelving control system according to claim 1 includes: The user interaction system receives the content of the stored file and the location information of the stored file input by the user when placing the file through the input module of the touch screen, receives the user's query content through the touch screen, and feeds back the user's query results through the user interface on the touch screen; A memory for storing the content of the stored file and the location information of the stored file input by the user when placing the file; The intelligent rolling shutter door type compact shelving control system receives and executes the user query command of the user interaction system and feeds back the search results; The rack opening and closing system is installed between two adjacent sets of compact shelving units and uses a motor to adjust the positional relationship between the units. The execution control module of the intelligent rolling door type compact shelving control system controls the shelf opening and closing system according to the file detection instructions.

Images