CN105005578A - Multimedia target information visual analysis system - Google Patents
Multimedia target information visual analysis system Download PDFInfo
- Publication number
- CN105005578A CN105005578A CN201510263196.8A CN201510263196A CN105005578A CN 105005578 A CN105005578 A CN 105005578A CN 201510263196 A CN201510263196 A CN 201510263196A CN 105005578 A CN105005578 A CN 105005578A
- Authority
- CN
- China
- Prior art keywords
- information
- submodule
- video
- data
- retrieval
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/16—File or folder operations, e.g. details of user interfaces specifically adapted to file systems
- G06F16/168—Details of user interfaces specifically adapted to file systems, e.g. browsing and visualisation, 2d or 3d GUIs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/44—Browsing; Visualisation therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/80—Information retrieval; Database structures therefor; File system structures therefor of semi-structured data, e.g. markup language structured data such as SGML, XML or HTML
- G06F16/83—Querying
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a multimedia target information visual analysis system. A multimedia target information acquisition module acquires multimedia data, and stores the acquired multimedia data in a local database; a data preprocessing module performs semantic synthesis on the multimedia target information with a media-cross information integration technology; a multimedia target information retrieval module retrieves target information concerned by a user to form a multimedia file set; a visual graphical interface module performs visual output on the multimedia file set with a computer information visualization algorithm; and a target information interactive analysis module controls the extraction of data and the display of a picture through an interactive means, and performs intelligent screening and anomaly analysis of the multimedia target information in combination with statistics, clustering and linear regression analysis methods in order to ascertain the value of an analysis object. The multimedia target information visual analysis system has universality in multi-information display and a high flexibility degree in information primitive presentation, and can be used for providing feedback support to aid decision making.
Description
Technical field
The present invention relates to target information analysis, for area of computer aided intelligence analysis and decision support, particularly relate to a kind of multi-media objects information visualization analytic system.
Background technology
Along with the widespread use of computer digit technology, the channel of people's obtaining information is widened day by day, pays close attention to target carry out information retrieval for any one, and the result for retrieval caused may be all the data of magnanimity comprising word, image, Audio and Video.The development of computer picture, graph technology, by means of the huge processing power of computing machine, by information calculation visualization technology, magnanimity all kinds data are converted to static state or dynamic image, figure is presented in face of people, for people analyze, understand data, form concept and find out rule and provide strong means.How to extract from result for retrieval and can reflect the information that user needs, become a more and more important goal in research.Because multimedia adopts flat streaming structure usually, be not easy to treatment and analysis, if do not set up suitable structure and model, can only by carrying out Realization analysis to the deciphering one by one of media.This method efficiency when processing and analyze broad medium data is extremely low.
Existing target intelligence analytical approach is mainly based on " people " leading informix with study and judge.Multi-media objects information visualization analysis and utilization visual computer technology, the distribution of the target information of media statement is expressed with graph image, assisted user is carried out by vision, namely " people " processes a large amount of information faster, and assisted user is understood data, finds rule, made a policy faster.Such as: by the multi-media objects acquisition of information of certain star personality and visual analyzing, its hobby, social relationships, famous speech, nearest behavior trend etc. can be learnt, and predict that its lower step may be movable, for relevant Decision provides foundation by analyzing its mechanics.
At present, have not been reported at home for multimedia target information Visualized Analysis System.The target information analysis software that domestic intelligence circle is commonly used is substantially all the media data obtained for singlehanded section, and display mode is based on battle state display, lacks related tool to target information visual analyzing aspect.Therefore, research and development friendly interface, " multi-media objects information visualization analytic system " simple to operate, improve the comprehensive analysis processing ability of multi-media objects information under visualization tool is assisted, for customer analysis, understanding data, form concept and find out rule and provide strong instrument, is very necessary.
Summary of the invention
The present invention is directed to the deficiency that prior art exists, one is provided to utilize Interactive Object information analysis tool, can data be understood, find rule, make a policy by assisted user fast, and analyst's work efficiency and accuracy rate can be improved, there is versatility and adaptability is good, information pel shows degree of flexibility much higher media object information visualization analytic system.
The present invention is achieved through the following technical solutions: a kind of multi-media objects information visualization analytic system, comprise: multi-media objects information acquisition module, data preprocessing module, multi-media objects information searching module, visualized graph interface module and target information interactive analysis module, it is characterized in that: multi-media objects information acquisition module connects different databases, the multimedia information source of the multimedia information source that analyze and alternate manner collection is stored in local data base, data preprocessing module carries out pre-service for multimedia messages, format manipulation is carried out respectively according to video, image, text and audio frequency, semantic processes is carried out to video and audio and image, utilizing is undertaken comprehensively across media information integrated technology by multi-media objects information, make full use of the information entrained by different media object, obtain multimedia and associate content of text of attaching troops to a unit, multi-media objects information searching module is according to user's input information, with reference to ontology model and object knowledge rule, the target information that retrieval user is paid close attention to from index file, forms multimedia file collection, and becomes XML data file to be retained in local data base according to the data structure organization of design, visualized graph interface module reads the multimedia file collection of above-mentioned XML file tissue from local data base, carry out visual to the multi-media objects information of target signature vocabulary representative, export visualized graphs information, and different pattern layout's types is provided: network topology, hierarchical layout, grouping layout, circular layout, tree topology and time series layout, target information interactive analysis module and visualized graph interface module carry out interactive analysis, by the extraction of interactive means control data and the display of picture, obtain the personalized input parameter that user provides, the visualization result of target information is adjusted or revised, in conjunction with statistics, cluster, linear regression analysis method, carry out intelligent screening and the anomaly analysis of multi-media objects information, verify the value of analytic target or propose the demand of further obtaining information, further information refining is carried out by existing iteration and convergence algorithm, until user's end of input information, represent that target information analysis terminates, the decision information obtained has met user's decision requirements.
The present invention's beneficial effect compared with prior art:
1. the present invention utilize across media information integrated technology and graph visualization technology, multi-media objects information have been carried out comprehensive, make full use of the information entrained by different media object, intelligence analysis personnel are enable to obtain different information about data set according to the needs of self from different angles, and check unstructured information intuitively, thus contribute to carrying out of target information analysis.
2. the present invention relies on the powerful data-handling capacity of computing machine and computer picture, graphics rudimentary algorithm and visualized algorithm, a large amount of multi-medium datas is converted to static graphic image to be presented in face of people, and allow by the extraction of interactive means control data and the display of picture, make to be implied in sightless phenomenon among data and become visible, for customer analysis, understand data, form concept and find out rule and provide strong means.
3 in order to see the relation between information from different perspectives, and the present invention also provides different pattern layout's types: network topology, hierarchical layout, grouping layout, circular layout, tree topology, time series layout.Can be used for the analyses such as public security case relevant information, military target information, the intuitive that its visual mode and specific analytical approach are brought and analysis result, improve work efficiency and the accuracy rate of analyst greatly.
4 the present invention utilize Interactive Object information analysis tool, by target information interactive analysis module and the interactive analysis of visualized graph interface module, can understand data, find rule, make a policy by assisted user faster.
5 the present invention are that user provides a kind of instrument directly can understanding target situation development evolvement situation.The present invention is by multi-media objects information acquisition module and local data base interactive interfacing, obtain object information data, after pre-service is carried out to different media information, multi-media objects information searching module is utilized to inquire about multi-media objects information data, subsequently by the attributive character of visualized graph interface modules exhibit target information, and carry out Visual Interactive analysis by target information analysis module, thus have multiple information show versatility and adaptability is good, information pel show degree of flexibility high, can feed back for aid decision making provides advantages such as supporting better.
Accompanying drawing explanation
In order to more clearly understand the present invention, now by embodiment of the present invention, simultaneously with reference to accompanying drawing, the present invention will be described, wherein:
Fig. 1 is multi-media objects information visualization analytic system operation logic schematic diagram of the present invention.
Fig. 2 is the composition structured flowchart of module described in Fig. 1.
Fig. 3 is the operation logic figure of Fig. 2 multi-media objects information searching module.
Fig. 4 is the operation logic figure of Fig. 2 visualized graph interface module.
Fig. 5 is the anomaly analysis algorithm flow chart of Fig. 2 target information interactive analysis module.
Embodiment
For making the object of the application, technical scheme and advantage clearly; below in conjunction with drawings and the specific embodiments; the application is described in further detail: the present embodiment is implemented under premised on technical solution of the present invention; give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to the embodiment of subordinate.
As depicted in figs. 1 and 2.A kind of multi-media objects information visualization analytic system comprises: multi-media objects information acquisition module, data preprocessing module, multi-media objects information searching module, visualized graph interface module and target information interactive analysis module, wherein: multi-media objects information acquisition module, multi-media objects information searching module is connected visualized graph interface module and target information interactive analysis module with data preprocessing module by unification user access interface, semantic indexed database is connected by universal data access interface, object knowledge storehouse, multimedia document storehouse and local target information storehouse.
Multi-media objects information acquisition module comprises data management submodule and data transform subblock.After completing database and network configuration, multi-media objects information acquisition module connects other multimedia databases by network.Data management submodule completes storage and the stock management of data, namely be stored in after obtaining the multimedia data resources of other databases in local data base, data transform subblock utilizes Software tool to convert multimedia file to unified file layout, as: bmp picture is converted to jpg form, or mkv format video is converted to rmvb form.Unified form refers to according to following form preservation multimedia document:
Text document: txt, pdf, doc tri-kinds;
Image: jpg and tiff two kinds;
Video: avi and rmvb two kinds;
Audio frequency: wav and mp3 two kinds;
Form: xls;
Webpage: html and mht two kinds.
Data management submodule obtains the document after above-mentioned conversion, there is local data base, according to medium type, sets up different data table items respectively, file storage is entered database.
Data preprocessing module carries out pre-service for multimedia messages, format manipulation is carried out respectively according to video, image, text and audio frequency, and semantic processes is carried out to video and audio and image, obtain multimedia and associate content of text of attaching troops to a unit, finally semantic indexing is set up to each media content.
Described data preprocessing module comprises format analysis submodule, key frame of video extracts submodule, video image semantic processes submodule, audio frequency semantic processes submodule, text semantic process submodule, wherein:
The realization of format analysis submodule is read in the parsing of dissimilar multimedia file format and data, is plain text by the text resolution such as webpage and word.
Key frame of video extracts the video data that submodule obtains for multi-media objects information acquisition module, completes the extraction of key frame representative in video, for representing the main contents of this video after carrying out the structuring process of video content.Video content structuring disposal route is: key frame of video extracts submodule and first video is divided into camera lens, using as basic indexing units, then extracts the key frame of each camera lens to represent the content of whole video.After completing shot boundary detector and key-frame extraction, just can change into the retrieval association of the huge video of data volume and the retrieval of its key frame images is associated.
The implementation method of described shot boundary detector is: shot boundary detector is basic, an important technology in video retrieval technology, and the quality of its Detection results directly affects the performance of subsequent video retrieval.Consider the feature that the video data volume is large, the present invention adopts the method based on border, is namely determined the dislocation of camera lens by the boundary between detector lens.Classical histogram relative method is adopted to obtain the shot boundary of video.When the histogram relative method adopted obtains the shot boundary of video,
First the histogram H [f (x, y, t), k] of two field picture is calculated, wherein k=0,1 ..., K-1, and compare the histogrammic correspondence statistics of front and back two field picture, now luminance difference is
Wherein, f (x, y, t) represents the picture frame of t, x and y represents that the ranks value .K of image is histogrammic dimension respectively, such as, for 8 gray level images, and K=255.
If the D of this two two field picture is greater than certain predetermined threshold value, so usually can think that sequence has interruption.For automatically
T=m+sσ (2)
Determine this threshold value, the distribution of difference between frame and frame and difference will be calculated whole video frequency program.If the average of difference and variance are respectively m and σ, then threshold value is:
Wherein s is weight coefficient, and be generally a less number, usual value is 0.01.
Key-frame extraction: key-frame extraction is the key link of the video features realizing based target.
The key frame of a camera lens is exactly a frame of main contents in this camera lens of reflection or some two field pictures.According to the complexity of camera lens content, one or more key frame can be chosen from a camera lens.The present invention adopts a kind of Key-frame Extraction Algorithm compared based on non-adjacent frame to extract the key frame of each camera lens in infrared video.Algorithm basic thought is: first determine a present frame, for each frame follow-up, all carries out distance with it and calculates, if its distance is greater than setting threshold value δ
i, then think that this frame is a key frame, and this frame carried out follow-up calculating as present frame.Arthmetic statement is as follows:
If camera lens C is by frame set { F
1, F
2... F
i, F
nform.If F
cfor present frame, F
sfor F
csubsequent frame.The element number that key frame integrates is as M, and frame pitch is from being D (F
1, F
2), and set threshold value δ
i, key frame set is K, is initialized as an empty set.
Step 1 establishes F
c=F
1, then K=K ∪ F is had
1, M=1;
Step 2 calculates D (F
c, F
s):
If D is (F
c, F
s) > δ
i, then K=K ∪ F
s, M=M+1, F
c=F
s, return the 2nd step;
If D is (F
c, F
s) < δ
i, F
c=F
c, return the 2nd step;
If step 3 F
c=F
n, exit.Wherein K is required key frame set, δ
ibe a threshold value, what be used for controlling key frame chooses number, and after extraction key frame collection, the number M of the key frame collection in camera lens is the important parameter of a reflection lens features.If M is very little, such as M=1, can think that shot change is very little.If M is very large, then think the altering a great deal or threshold value δ of camera lens
iif must be less, δ can be adjusted
isize recalculate.Threshold value δ
ichoose and rule of thumb obtain with experiment.
The video data that video image semantic processes submodule obtains for multi-media objects information acquisition module, use classical OCR technology and speech recognition technology, obtain the word content in video, in addition, the target visual feature information extraction of image and key frame of video is also saved in database, as the feature of later retrieval by video image semantic processes submodule.Target visual characteristic information mainly comprises the features such as color, texture, shape.
The voice data that audio frequency semantic processes submodule obtains for multi-media objects information acquisition module, by speech recognition technology, transfers main contents to text document;
The content of text that data are converted to by text semantic process submodule carries out word segmentation processing, obtains the semantic classification of these data, and sets up semantic indexing to data, deposit in local data base through semantic analysis.
Need to carry out text semantic process and the document setting up semantic indexing comprises: text document, audio conversion document, video image content mark document, video image convert documents, webpage.
When content of text carries out word segmentation processing in order to carry out follow-up semantic analysis use, be that Chinese text is divided into multiple lexical item, and remove the stop-word of not expressing any semantic information in text, make it to become independently Feature Words, wherein, stop-word refers to preposition, auxiliary words of mood etc.Then carry out stem extraction, remove the affixe of word, obtain the literary style that word is the most general, the different expression-forms of same word can be avoided the impact of text analyzing.
At present, Chinese Word Automatic Segmentation model can be divided three classes substantially: be mechanical Chinese word segmentation method respectively, and also known as dictionary formula syncopation, semantic point morphology and artificial intelligence method, also known as understanding point morphology.Mechanical Chinese word segmentation refers to mate in dictionary; Semantic point morphology introduces semantic analysis, more processes the language message of natural language self; Artificial intelligence is a kind of pattern of information being carried out to intelligent processing method, mainly contains two kinds of processing modes: one is based on psychologic symbol processing method, the function of simulation human brain.Namely be the function of wishing simulation human brain as expert system, constructive inference network, through symbol transition, thus can the process of making an explanation property.One is based on physiological analogy method.Neural network is intended to the operating mechanism of the nervous system mechanism of simulating human brain to realize certain function.
The segmenting method adopted in the present invention is mechanical Chinese word segmentation method, the technology used can be that the FreeICTCLAS Words partition system researched and developed by the Chinese Academy of Sciences carries out word segmentation processing, and calculate the word frequency that isolated Feature Words occurs in the text, then the Feature Words in every section of text and word frequency thereof are kept in text document in vector form.In addition, this module is also responsible for removing stop-word, as " ", " with ", " and about " etc. some prepositions and auxiliary words of mood or very conventional words.
Described semantic indexing, adopts Lucene to carry out the establishment of index.In order to the work efficiency of index can be made to increase, the method taked first will the data setting up index be needed to put into internal memory, after having set up index, the data of write memory is before written in file.Such way by means of the buffer zone of internal memory, avoids disk operating frequently, and what improve index sets up speed.In the process of index creation, the mode that system application merges carries out the generation of index file, improves system performance.Can also be arranged operation index file being carried out to other by the application programming interfaces in Lucence, such as data compression step, the space that the index file generated can be made like this to take is a lot of less.
Content due to reality constantly updates, so the content of index also needs constantly to upgrade.The renewal of index mainly contains incremental update and these two kinds of update modes of batch updating, and wherein batch updating refers to that every day all will upgrade, mainly in order to batch updating is about the text of target information and pictorial information.Incremental update is then can be retrieved in time this function to realize the up-to-date target information collected.The index file of these two kinds of index upgrade modes separates, and they are divided to be placed in different files, and these two index lists all can be used in the process of retrieval.
Multi-media objects information searching module is according to user instruction, from index file, knowledge rule is formed with reference to Ontology Modeling and user individual knowledge, realize retrieval user being paid close attention to target information, form multimedia file collection and characteristic feature word finder, and become XML data file to be retained in local data base according to the data structure organization of design; Data structure contents comprises: title, medium type, content, media time, source, association place, association personage.
System provides character search and Image Retrieval two kinds of retrieval modes.Described character search mode is carried out knowledge augmented to inputted target text information and is retrieved generating expanded text collection together with other words of described Word message synonym, and the visual signature of described Image Retrieval foundation object knowledge image carries out the retrieval of associated images and video.
Multi-media objects information searching module comprises: user inputs submodule, knowledge acquisition submodule, character search submodule, Image Retrieval submodule, result for retrieval encapsulation submodule, wherein:
User inputs submodule: the retrieval request obtaining user's input, resolve target information retrieval of content, range of search and retrieval type, retrieval type comprises exact matching and fuzzy matching, described exact matching, refers to that result for retrieval must comprise all input keywords; Described fuzzy matching, as long as refer to that comprising a certain keyword just counts result for retrieval by the document.
Knowledge acquisition submodule: linking objective knowledge base, obtains relevant knowledge rule according to user's targeting, thus realizes the expansion of query contents.Knowledge rule comprises: be kept at object knowledge image corresponding to the synonym of the target information in object knowledge storehouse, near synonym, conjunctive word, target information and visual signature thereof, such as: the maritime patrol STOWAGE PLAN sheet that maritime patrol ship is corresponding and visual signature information thereof.
Character search submodule: foundation is according to user's targeting and retrieval request, and synonym, near synonym, the conjunctive word of the target information to obtain from knowledge acquisition submodule, use the semantic retrieval algorithm that Lucence provides, realize the character search in aforementioned semantic indexing, obtain and input with inquiry the document that there is similarity.
When submitting text based query word to, user is usually difficult to provide the complete description of information of will searching.User can only express with one or two word the information that they want usually.These query words do not defined very well can only provide fuzzy description, therefore, there will be information deficiency and cause poor Search Results.In order to address this problem and disclose minority query word beyond information, present invention employs the method for expanding query input word, thus obtain a series of new query word.Here, the expansion of query word is use the correlativity semantically of vocabulary to carry out substantially, and these words organize together with synonym form substantially, are called synonym collection.These synonym collection are organized in together according to the vocabulary correlativity be defined on them.According to the relation of word, be added in newly-generated text set concentrating the word listed at synonym.Except based on except the method for vocabulary, the enquiry expanding method based on certain aspect knowledge can also be used.Such as supplement the word relevant to special scenes by the knowledge base of some specific areas with the method for artificial intelligence.
The text set of this expansion is regarded as a Text eigenvector, in text based retrieval process, just can be carried out the coupling of proper vector by word index.Namely be that all words in superset are delivered in searching system with "or" relation.Like this, in the text degree of correlation, document relatively can be given higher grade with query text collection.According to this grade, the text of some and the multimedia messages of correspondence comprises picture, Voice & Video is out selected, forms the document sets that a text is relevant.
Image Retrieval submodule obtains corresponding object knowledge picture according to text event detection target information, target expansion Word message in knowledge acquisition submodule, similarity according to vision content carries out the retrieval of image and key frame of video, replaces video with the key frame of video retrieved.
Image Retrieval submodule has accumulation histogram feature, direction details histogram and Tamura texture from the visual signature that image data set extracts.Utilize described feature to carry out characteristic matching, the similarity retrieval between can realizing based on image, and utilize the retrieval of image and key frame of video, realize the retrieval from image to video.
Result for retrieval is become XML data file according to the data structure organization of design by result for retrieval encapsulation submodule.Multimedia title, data type, media time, source, content, theme, conjunctive word is defined in XML.
Visualized graph interface module reads multimedia file collection and the characteristic feature word finder of above-mentioned XML file tissue from local data base, carries out visual, export visualized graphs information to the multi-media objects information of target signature vocabulary representative.
Described visualized graph interface module comprises timeline information submodule, calendar information submodule, network information submodule, wherein:
Timeline information submodule take time as coordinate, and all target information key elements are carried out tissue line with seasonal effect in time series order;
Calendar information submodule take time as coordinate, and all target information key elements are carried out tissue line with the order of calendar (year, month, day);
Network information submodule represents incidence relation between target information key element in the mode of network chart;
Graph Control submodule, according to user instruction operation and input parameter, adjusts multi-media objects information graphic visual means.
Target information interactive analysis module and visualized graph interface module are carried out alternately, obtain the personalized input parameter that user provides, and the visualization result of target information is adjusted or revised, and combine statistics, cluster, classification, the analytical approachs such as recurrence, carry out intelligent screening and the anomaly analysis of multi-media objects information, verify the value of analytic target, or the demand of further obtaining information is proposed, and carry out further information refining by existing iteration and convergence algorithm, until user's end of input information, represent that target information analysis terminates, the decision information obtained has met user's decision requirements.
Described target information comprises people information, information of place names, platform target (as: automobile, naval vessel etc.) information, event (as: 911 event) information.
Target information interactive analysis module comprises statistical study submodule, cluster analysis submodule, forecast analysis submodule, anomaly analysis submodule, wherein,
Statistical study submodule uses histogram, cake chart, has retrieved the associated statistical information of media information in broken line graph display systems, as target type with quantity, multimedia messages relate to theme, multimedia messages relates to personage etc.; The multimedia document of cluster analysis submodule to same subject or other attributes is sorted out; Forecast analysis submodule carries out prediction to target development activity trend and judges; The target that provides anomaly analysis submodule develops the abnormal judged result in active procedure.
Cluster analysis submodule comprises document subject matter cluster and document similarity cluster to the mode retrieving media information and carry out cluster.Wherein, the document subject matter of mark that document subject matter cluster obtains according to XML file, utilizes similar theme to carry out target information cluster.The method of document similarity cluster is: obtain image respectively, Word message that the markup information of video, video/audio are converted to and text document, utilizes classical vector space model to carry out document description.
As Fig. 3, multi-media objects information searching module, according to user instruction, from index file, realizes retrieval user being paid close attention to target information, form multimedia file collection and characteristic feature word finder, and become XML data file to be retained in local data base according to the data structure organization of design.Wherein, multi-media objects information visualization analytic system provides character search and Image Retrieval two kinds of retrieval modes.Described character search mode is carried out knowledge augmented to inputted target text information and is retrieved generating expanded text collection together with other words of described Word message synonym, and the visual signature of described Image Retrieval foundation object knowledge image carries out the retrieval of associated images and video.The text set of this expansion is regarded as a Text eigenvector, in text based retrieval process, just can be carried out the coupling of proper vector by word index.Namely be that all words in superset are delivered in searching system with "or" relation.Like this, in the text degree of correlation, document relatively can be given higher grade with query text collection.According to this grade, the text of some and the multimedia messages of correspondence comprises picture, Voice & Video is out selected, forms the document sets that a text is relevant.
As Fig. 4, visualized graph interface module reads the data acquisition XML file from this locality, obtains the data relevant with input parameter, converts data to graph-based and be presented on panel.In display mode, comprise time shaft display mode, calendar display mode and network display mode, realized by timeline information submodule, calendar information submodule and network information submodule respectively.
The implementation represented the time axis graphical of multi-media objects information in visualized graph interface module graphical interfaces is as follows:
Introduce time dimension, shown by the line chart of time data as two dimension, x-axis represents the time, and y-axis may be selected to be in XML file the attribute obtaining multimedia file, such as: subject information, thus sets up formation two dimensional surface rectangular coordinate system time layout;
Media time and the theme of multimedia file is obtained from XML file, multimedia file is carried out x-axis alignment according to media time, y-axis alignment is carried out according to theme, multimedia file represents in the mode of loose point, the label of loose point is the title of multimedia file, and the color of loose point is media data type;
Allow user to select different y-axis information to obtain different time layout arrangement modes, so as to the time series relation of different media files under different attribute is described, target information development course in time can be understood; Smash a label by point, browsing multimedia file can be realized, intuitively check the information content.
The implementation represented the calendar graphical of multi-media objects information in visualized graph interface module graphical interfaces is as follows:
By date independently dimension, set up the calendar layout forming two dimensional surface; Media time and the theme of multimedia file is obtained from XML file, multimedia file is carried out the alignment on date on calendar according to media time, multimedia file represents in the mode of loose point, and the label of loose point is the theme of multimedia file, and the color of loose point is media data type; User is allowed to select different calendar time attributes (as: year, month, day) to obtain different calendar layout arrangement modes, so as to the time series relation of different media files under different attribute is described, target information development course in time can be understood; Smash a label by point, browsing multimedia file can be realized, intuitively check the information content.
The implementation represented the network graphic of multi-media objects information in visualized graph interface module graphical interfaces is as follows:
Adopt node link method define grid: the object encoding in network and node are expressed as an independently multimedia document, namely two internodal relations connect and represent two multimedia documents and have identical attribute, as theme.Node adopts circular expression, limit adopts the straight line of connection two nodes to represent, then network formalization can be expressed as an ordered pair <V, E>, i.e. G=<V, E>, wherein V is a nonempty set, and E is that the orderly occasionally unordered idol of element in V obtains geometry.V and E is called the set of the node of G and the set on limit; Employing power guidance mode carries out network topology: power guiding layout is also known as elasticity layout (Spring-embeded Lay-out), it is simulation mechanics equilibrium principle, figure interior joint is modeled as steel loop, connecting analog is spring, make physical system reach mechanical balance by constantly adjusting steel loop position, thus realize the method for layout.Object is the position of laying the node of figure in two dimension or three dimensions, makes the length on all limits as far as possible equal, and makes the intersection between limit few as far as possible.Divide the dynamics between cobordant and set of node according to the relative position of node, the interaction force of then simulating limit and node according to internodal dynamics size reduces the energy of their entirety as far as possible.Generally, the limit in figure can be imagined as a Ge Dan Praise.The repulsion of spring-like and gravitation, according to Hooke's law, will produce gravitation when the node at two ends, limit is stretched time, on the contrary when two nodes from too close to will produce repulsion.Under the state that the energy of this system is in balance, while tend to the length (because the interaction force between node) with equal Uniform, simultaneously isolated node then tends to away from the part be connected (because electric repulsion); Smash a label by point, browsing multimedia file can be realized, intuitively check the information content.Generally speaking, the quantity of information of the data set of network is all very large, and then the quantity of information of each Information Level obtained according to the data set of network is also all very large, like this, if all information of each Information Level be all directly presented on view, then VC can be caused.So, in this step, before generation view, the data set of the main Information Level of the layout sample of the view as each Information Level is sampled.Certainly, the sample data collection after sampling, should by can embody the typical data of profile data set of former main Information Level, namely important node and between link form.
In this step, adopt described layout method, for each of multiple Information Level, respectively according to the data set that this Information Level comprises, generate view, and then go out to obtain the Information Level of view of optimal layout's effect according to generated views selection, as main Information Level.Particularly, the layout effect of view can be weighed according to following condition:
There is preferably topological structure, can clearly be divided into several part;
There is good symmetrical structure, so-called good symmetrical structure, evaluate like this: select the central point of view (to the node that the surrounding distance of view is all identical or approximate), a cross is drawn centered by this central point, view is divided into four parts, if the quantity of every a interior joint is all identical, so view just has good symmetrical structure;
Average path length is short, so-called average path length, calculate like this: select any two nodes to form a node pair in the view, calculate the bee-line between them, and then calculate the mean value of the right bee-line of all nodes existing in view; The scale of view is less, and the number of the node namely comprised in view is less.
Graphic operation: in order to make some target entity or associate eye-catching, system allows user to selected target entity or is associated in figure aspect and modifies, as amendment represents font and the size of the color of certain entity icon, size and icon word, or amendment represents the color of the line of certain association and thickness etc.After graph of a relation is drawn, user may require that a certain section layout is relatively sparse and another part layout is relatively intensive, and therefore system allows user to carry out layout again to the icon in graph of a relation and line.Under certain conditions, user can the layout of other parts be dissatisfied to the layout of certain part is satisfied, and therefore, system is after the part that user is satisfied with relatively to layout is selected, keep this part icon and line relative position constant, layout is again carried out to other parts.Otherwise system also allows the part to user selects to carry out layout again, and the relative position of the icon of other parts and line remains unchanged.
A point in utilize vector space model (Vector Space Model) to be all mapped as by each document in target information interactive analysis module vector space that one group of normalized orthogonal entry vector opens.Represent that the word of document and the weight of correspondence thereof just constitute a vector (W in " space "
1j, W
2j..., W
nj), wherein, W
kjfor document D
jthe weight of middle entry k.
Weight is defined as: W
kj=TF
kj* IDF
k
TF
kjfor word frequency, IDF
kfor inverse text frequency, j is text sequence number, and k is entry sequence number.
The similarity of document uses the included angle cosine between vector to weigh:
Sim
ij=cosine(D
i,D
j)
The cluster of document adopts the method for arest neighbors, is implemented as follows:
1. choose a document D
i, centered by this sample, set up a bunch of C
1={ D
i, i is document sequence number;
2., in the samples of text of non-cluster, choose and C
1distance sample farthest, and by centered by a newly-built bunch of C
2, recording this maximum distance is m_dis;
3. determine that distance (similarity) threshold value T, general m_dis are the multiple of T;
4., in the paper sample of non-cluster, calculate the minor increment d of this sample and existing all bunches
1;
5. make d=max{d
i, if d>T, newly-built one centered by this object bunch, return step 1; If at newly-built bunch, step 2 cannot be forwarded to;
6. remaining samples of text be assigned to from belonging to its that nearest center bunch.
Forecast analysis submodule carries out prediction to the target development activity trend that user pays close attention to and judges; Time series data law forecasting is on the basis of time series data analysis, and by sequential forecasting models Time Created, reaching to find from mass data needs information, and the object of estimating timing development trend.In the present invention, this technology is mainly used in the prediction of moving-target mechanics.
The method of moving average is adopted to predict moving-target activity schedule data.The method of moving average is the improvement to segmental averaging method, is a kind of Time series analysis method being eliminated the exception interference in data by smoothing effect that is average and movement.The use flow process of the method for moving average is: first by the data of observation phase by far away and be closely averaged by certain leap phase, get its arithmetic average, then along with the passing of observation period, move forward by the observation period data of certain leap phase are also corresponding, often move forward a step, remove data above, increase a new data after the original observation period, and try to achieve moving average one by one, finally predict that last moving average of observation period is as the foundation determining predicted value using close.
The method of moving average is not only the mean value that move some data in seeking time sequence and is formed new time series, and can smooth time series preferably, eliminates erratic variation or seasonal move in time series.This method is usually used in smooth historical data, discloses alteration trend, is therefore widely used in the prediction in social public administration field.The conventional method of moving average once moving average and Double moving average.
Double moving average is on the basis of Single moving average value, and carry out moving average again to the time series data of linear alteration trend, method and Single moving average are substantially identical, but the value of twice average middle n must be consistent.The computing formula of Double moving average value is:
In formula:
it is the Single moving average value in t period;
be the Double moving average value in t period, t is number of times in period, y
tbe the data of t variable in period, n is every section of epoch number or segment data number crossed over.
After completing Double moving average value, need to set up moving average model(MA model).The development law that the core of the method for moving average is to utilize once, Double moving average value has carrys out the development trend of the former time series data of matching, and founding mathematical models provides foundation for predicting.The general type of its model is:
y
t+1=a
t+b
t×l
In formula, y
t+1it is the predicted value in t+l period;
B
tslope, i.e. the variable quantity of unit interval y;
A
tintercept, i.e. the data level of t y;
L be by current period t to prediction time during number in period.
Coefficient a in model
t, b
tcomputing formula as follows:
The target that provides anomaly analysis submodule develops the abnormal judged result in active procedure.
In object information data, carry out abnormality detection, mainly solve following two subproblems:
(1) in the data acquisition in given field, define which type of data and can be considered to abnormal;
(2) design effective method and excavate exception.
For first problem, extremely can be defined as: abnormal is the data departing from most of data in data centralization, departing from is produced by different mechanism, but not random deviation.According to this definition, the exception in a time series can bring adverse influence to Model Distinguish and parameter estimation, is therefore harmful.In statistics, abnormity point typically refers to significantly away from other points, the data point disobeying sequence distribution, in regression model, abnormity point refers to and departs from very large data point with setting models, be offset to which kind of degree actually and just calculate exception, need to make certain hypothesis to the distribution of model error item.
For Second Problem, current Outlier Detection Algorithm roughly can be divided into five classes: Corpus--based Method distribution method, the method based on deviation, the method based on cluster, based on the method for distance and the method etc. of density based.
In several algorithms of abnormality detection, the singular value mining algorithm of density based can detect the singular value of time series local existence and accuracy of detection is higher.But this algorithm cannot directly apply to orderly time series.And target information time series is an orderly data set, the important feature of its one has time attribute, has strict order between sequential value.Therefore, to adjust the singular values standard form algorithm of density based, make it to be adapted to target intelligence Information abnormity and detect.
The local singular value that present invention employs based on moving window differentiates (LOF) method.Concrete grammar is:
First, the moving window of target information time series regular length is split, be the time series of m for length, be after the moving window segmentation of ι by length, obtaining m-ι+1 length is the subsequence of ι, each subsequence is regarded as a sample, so only need to detect singular point in m-ι+1 sample, thus mistake when orderly time series being detected when eliminating local density's Outlier Detection Algorithm.Then the inflection point detection algorithm of classical density based is adopted to detect each sample.Because the singular value in target information time series refers to fluctuate in semantic content compared with front several target information differ larger point.So the up-to-date LOF value entering window object only need be calculated when detecting.Namely for moving window [t
i, t
i+1, t
i+2... t
i+ ι] in only need calculate LOF when detecting
minPis(t
i+ ι).Work as LOF
minPis(t
i+ ι) value thinks object t when being greater than threshold value d
i+ ιfor singular value.
Claims (10)
1. a multi-media objects information visualization analytic system, comprise: multi-media objects information acquisition module, data preprocessing module, multi-media objects information searching module, visualized graph interface module and target information interactive analysis module, it is characterized in that: multi-media objects information acquisition module connects different databases, the multimedia information source of the multimedia information source that analyze and alternate manner collection is stored in local data base, data preprocessing module carries out pre-service for multimedia messages, format manipulation is carried out respectively according to video, image, text and audio frequency, semantic processes is carried out to video and audio and image, utilizing is undertaken comprehensively across media information integrated technology by multi-media objects information, make full use of the information entrained by different media object, obtain multimedia and associate content of text of attaching troops to a unit, multi-media objects information searching module is according to user's input information, with reference to ontology model and object knowledge rule, the target information that retrieval user is paid close attention to from index file, forms multimedia file collection, and becomes XML data file to be retained in local data base according to the data structure organization of design, visualized graph interface module reads the multimedia file collection of above-mentioned XML file tissue from local data base, carry out visual to the multi-media objects information of target signature vocabulary representative, export visualized graphs information, and different pattern layout's types is provided: network topology, hierarchical layout, grouping layout, circular layout, tree topology and time series layout, target information interactive analysis module and visualized graph interface module carry out interactive analysis, by the extraction of interactive means control data and the display of picture, obtain the personalized input parameter that user provides, the visualization result of target information is adjusted or revised, in conjunction with statistics, cluster, linear regression analysis method, carry out intelligent screening and the anomaly analysis of multi-media objects information, verify the value of analytic target or propose the demand of further obtaining information, further information refining is carried out by existing iteration and convergence algorithm, until user's end of input information, represent that target information analysis terminates, the decision information obtained has met user's decision requirements.
2. multi-media objects information visualization analytic system according to claim 1, it is characterized in that: described data preprocessing module comprises format analysis submodule, key frame of video extracts submodule, video image semantic processes submodule, audio frequency semantic processes submodule, text semantic process submodule, wherein: format analysis submodule: realizing reading in the parsing of dissimilar multimedia file format and data, is plain text by the text resolution such as webpage and word; Key frame of video extracts submodule: the video data obtained for multi-media objects information acquisition module, completes the extraction of key frame representative in video, for representing the main contents of this video; The video data that video image semantic processes submodule obtains for multi-media objects information acquisition module, uses OCR technology and speech recognition technology, obtains the word content in video; Audio frequency semantic processes submodule: the voice data obtained for multi-media objects information acquisition module, by speech recognition technology, transfers main contents to text document;
Text semantic process submodule: content of text data be converted to carries out the process of participle stop-word, obtains the semantic classification of these data, and sets up semantic indexing to data, deposit in local data base through semantic analysis.
3. multi-media objects information interactive analytic system according to claim 2, it is characterized in that: key frame of video extracts the video data that submodule obtains for multi-media objects information acquisition module, carry out the structuring process of video content, carry out the extraction of key frame representative in video on this basis, for representing the main contents of this video.
4. multi-media objects information interactive analytic system according to claim 3, it is characterized in that: key frame of video extracts submodule and first video is divided into camera lens, using as basic indexing units, then extract the key frame of each camera lens to represent the content of whole video; After completing shot boundary detector and key-frame extraction, the retrieval association of the huge video of data volume is changed into the retrieval of key frame images is associated.
5. multi-media objects information visualization analytic system according to claim 1, it is characterized in that: described multi-media objects information searching module comprises user and inputs submodule, knowledge acquisition submodule, character search submodule, Image Retrieval submodule and result for retrieval encapsulation submodule, wherein: user inputs submodule: the retrieval request obtaining user's input, resolve target information retrieval of content, range of search and retrieval type, retrieval type comprises exact matching and fuzzy matching, described exact matching refers to that result for retrieval comprises all input keywords, as long as described fuzzy matching refers to that comprising a certain keyword just counts result for retrieval by the document, knowledge acquisition submodule linking objective knowledge base, obtains target relevant knowledge according to user's targeting, thus realizes the expansion of query contents, as house, and expansion room, building, house, house, room and buildings, object knowledge comprises: the object knowledge picture that the synonym of target information, near synonym, conjunctive word, target information are corresponding, character search submodule: foundation according to user's targeting and retrieval request, and from the target information expansion vocabulary that knowledge acquisition submodule obtains, realizes the character search in aforementioned semantic indexing, Image Retrieval submodule is according to text event detection target information, target expansion Word message and the Target Photo of acquisition, and the similarity according to vision content carries out the retrieval of image and key frame of video, replaces video with the key frame of video retrieved,
Result for retrieval encapsulation submodule: result for retrieval is become XML data file according to the data structure organization of design; Multimedia title, data type, media time, source, content and theme is defined in XML.
6. a kind of multi-media objects information visualization analytic system according to claim 1, it is characterized in that: described visualized graph interface module comprises timeline information submodule, calendar information submodule, network information submodule and Graph Control submodule, wherein: timeline information submodule take time as coordinate, and all target information key elements are carried out tissue line with seasonal effect in time series order; Calendar information submodule take time as coordinate, and all target information key elements are carried out tissue line with the order of calendar; Network information submodule represents incidence relation between target information key element in the mode of network chart; Graph Control submodule, according to user instruction operation and input parameter, adjusts multi-media objects information graphic visual means.
7. multi-media objects information interactive analytic system according to claim 1, it is characterized in that: described target information analysis module comprises statistical study submodule, cluster analysis submodule, forecast analysis submodule and anomaly analysis submodule, wherein, statistical study submodule uses cake chart, histogram, associated statistical information in broken line graph display systems, as target type with quantity, multimedia messages relate to theme, multimedia messages relates to personage; The multimedia document of cluster analysis submodule to same subject or other attributes is sorted out; Forecast analysis submodule carries out prediction to target development activity trend and judges; The target that provides anomaly analysis submodule develops the abnormal judged result in active procedure.
8. multi-media objects information interactive analytic system according to claim 1, is characterized in that: multi-media objects information visualization analytic system provides character search and Image Retrieval two kinds of retrieval modes.
9. multi-media objects information interactive analytic system according to claim 8, it is characterized in that: described character search mode is carried out knowledge augmented to inputted target text information and retrieved generating expanded text collection together with other words of described Word message synonym, the visual signature of Image Retrieval foundation object knowledge image carries out the retrieval of associated images and video.
10. multi-media objects information interactive analytic system according to claim 8, it is characterized in that: in text based retrieval process, using expanded text collection as a Text eigenvector, the coupling of proper vector is carried out by word index, all words in superset are delivered in searching system with "or" relation, higher grade can be given with query text collection document relatively in the text degree of correlation, according to this grade, the text of some and the multimedia messages of correspondence comprise picture, Voice & Video is out selected, form the document sets that a text is relevant.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510263196.8A CN105005578A (en) | 2015-05-21 | 2015-05-21 | Multimedia target information visual analysis system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510263196.8A CN105005578A (en) | 2015-05-21 | 2015-05-21 | Multimedia target information visual analysis system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105005578A true CN105005578A (en) | 2015-10-28 |
Family
ID=54378254
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510263196.8A Pending CN105005578A (en) | 2015-05-21 | 2015-05-21 | Multimedia target information visual analysis system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105005578A (en) |
Cited By (41)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106021485A (en) * | 2016-05-19 | 2016-10-12 | 中国传媒大学 | Multi-element attribute movie data visualization system |
CN106484804A (en) * | 2016-09-22 | 2017-03-08 | 北京金堤科技有限公司 | Key element based on electric field simulation and relation methods of exhibiting |
CN106528534A (en) * | 2016-11-09 | 2017-03-22 | 天津赛因哲信息技术有限公司 | Relevant word extraction method based on proper nouns |
CN106547891A (en) * | 2016-10-28 | 2017-03-29 | 浙江大学 | For the quick visualization method of the pictured text message of palm display device |
CN106649585A (en) * | 2016-11-18 | 2017-05-10 | 福建中金在线信息科技有限公司 | Retrieval method and device |
CN106844636A (en) * | 2017-01-21 | 2017-06-13 | 亚信蓝涛(江苏)数据科技有限公司 | A kind of unstructured data processing method based on deep learning |
CN106980689A (en) * | 2017-03-31 | 2017-07-25 | 邢加和 | A kind of method that data visualization is realized by interactive voice |
CN107305555A (en) * | 2016-04-20 | 2017-10-31 | ***通信集团河北有限公司 | Data processing method and device |
CN107343221A (en) * | 2017-08-14 | 2017-11-10 | 彭君毅 | A kind of online multimedia interaction systems and method |
CN107656760A (en) * | 2017-09-28 | 2018-02-02 | 联想(北京)有限公司 | Data processing method and device, electronic equipment |
WO2018040102A1 (en) * | 2016-09-05 | 2018-03-08 | 华为技术有限公司 | Audio processing method and device |
CN108197295A (en) * | 2018-01-22 | 2018-06-22 | 重庆邮电大学 | Application process of the attribute reduction based on more granularity attribute trees in text classification |
CN108319904A (en) * | 2018-01-24 | 2018-07-24 | 江苏金陵科技集团有限公司 | A kind of virtual interactive interface method assisting face or Car license recognition |
CN108416264A (en) * | 2018-01-29 | 2018-08-17 | 山东汇贸电子口岸有限公司 | A kind of searching method and search module of supporting OCR to input |
WO2018177162A1 (en) * | 2017-03-27 | 2018-10-04 | 华为技术有限公司 | Target retrieval method and device |
CN108829765A (en) * | 2018-05-29 | 2018-11-16 | 平安科技(深圳)有限公司 | A kind of information query method, device, computer equipment and storage medium |
CN109063198A (en) * | 2018-09-10 | 2018-12-21 | 浙江广播电视集团 | Melt the multidimensional visual search recommender system of media resource |
CN109101500A (en) * | 2017-06-20 | 2018-12-28 | 北京国双科技有限公司 | A kind of methods of exhibiting and device of file data |
CN109146190A (en) * | 2018-08-31 | 2019-01-04 | 饶智彬 | Intelligent sound based on interior way system and expansion visualizes real-time situation sensory perceptual system |
CN109344253A (en) * | 2018-09-18 | 2019-02-15 | 平安科技(深圳)有限公司 | Add method, apparatus, computer equipment and the storage medium of user tag |
CN109508453A (en) * | 2018-09-28 | 2019-03-22 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Across media information target component correlation analysis systems and its association analysis method |
WO2019101067A1 (en) * | 2017-11-21 | 2019-05-31 | 众安信息技术服务有限公司 | Information processing method and apparatus for data visualization |
CN109948009A (en) * | 2019-03-12 | 2019-06-28 | 福建奇点时空数字科技有限公司 | A kind of data visualization engine system |
CN109977155A (en) * | 2019-03-12 | 2019-07-05 | 福建奇点时空数字科技有限公司 | A kind of big data visualization system and its control method based on human-computer interaction |
CN110012349A (en) * | 2019-06-04 | 2019-07-12 | 成都索贝数码科技股份有限公司 | A kind of news program structural method and its structuring frame system end to end |
CN110347561A (en) * | 2019-06-11 | 2019-10-18 | 平安科技(深圳)有限公司 | Monitoring alarm method and terminal device |
CN110781237A (en) * | 2019-07-01 | 2020-02-11 | 厦门美域中央信息科技有限公司 | Information visualization analysis system based on artificial intelligence |
CN110795392A (en) * | 2019-10-29 | 2020-02-14 | 北京四方启点科技有限公司 | Packaging method of accounting electronic voucher and electronic equipment |
CN110995774A (en) * | 2019-09-30 | 2020-04-10 | 大唐可再生能源试验研究院有限公司 | Universal SCADA system for wind power plant |
CN111291896A (en) * | 2020-02-03 | 2020-06-16 | 深圳前海微众银行股份有限公司 | Interactive random forest subtree screening method, device, equipment and readable medium |
CN111382282A (en) * | 2018-12-28 | 2020-07-07 | 北京国双科技有限公司 | Method, device, storage medium and processor for processing data |
CN111402935A (en) * | 2019-01-03 | 2020-07-10 | 李庆成 | Method for playing audio and video data |
CN111611457A (en) * | 2020-05-20 | 2020-09-01 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
CN111784615A (en) * | 2016-03-25 | 2020-10-16 | 北京三星通信技术研究有限公司 | Method and device for processing multimedia information |
CN112307269A (en) * | 2020-08-24 | 2021-02-02 | 山东理工职业学院 | Intelligent analysis system and method for human-object relationship in novel |
CN112667185A (en) * | 2021-03-16 | 2021-04-16 | 北京航空航天大学 | Maintainability analysis visual display method and system |
CN112818275A (en) * | 2021-04-16 | 2021-05-18 | 泰德网聚(北京)科技股份有限公司 | Image-text resource management system based on audio and video |
CN115985509A (en) * | 2022-12-14 | 2023-04-18 | 广东省人民医院 | Medical imaging data retrieval system, method, device and storage medium |
CN116189193A (en) * | 2023-04-25 | 2023-05-30 | 杭州镭湖科技有限公司 | Data storage visualization method and device based on sample information |
CN116719785A (en) * | 2023-08-09 | 2023-09-08 | 北京量子伟业信息技术股份有限公司 | Database management system based on metadata |
CN118093923A (en) * | 2024-04-26 | 2024-05-28 | 鸣启数字科技(山东)有限公司 | Data element visual management system based on big data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101894170A (en) * | 2010-08-13 | 2010-11-24 | 武汉大学 | Semantic relationship network-based cross-mode information retrieval method |
CN102024056A (en) * | 2010-12-15 | 2011-04-20 | 中国科学院自动化研究所 | Computer aided newsmaker retrieval method based on multimedia analysis |
CN102110140A (en) * | 2011-01-26 | 2011-06-29 | 桂林电子科技大学 | Network-based method for analyzing opinion information in discrete text |
US20120137367A1 (en) * | 2009-11-06 | 2012-05-31 | Cataphora, Inc. | Continuous anomaly detection based on behavior modeling and heterogeneous information analysis |
CN102609512A (en) * | 2012-02-07 | 2012-07-25 | 北京中机科海科技发展有限公司 | System and method for heterogeneous information mining and visual analysis |
CN103838826A (en) * | 2014-01-23 | 2014-06-04 | 北京东方泰坦科技股份有限公司 | Integration method of dynamic heterogeneous space information plotting data |
-
2015
- 2015-05-21 CN CN201510263196.8A patent/CN105005578A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120137367A1 (en) * | 2009-11-06 | 2012-05-31 | Cataphora, Inc. | Continuous anomaly detection based on behavior modeling and heterogeneous information analysis |
CN101894170A (en) * | 2010-08-13 | 2010-11-24 | 武汉大学 | Semantic relationship network-based cross-mode information retrieval method |
CN102024056A (en) * | 2010-12-15 | 2011-04-20 | 中国科学院自动化研究所 | Computer aided newsmaker retrieval method based on multimedia analysis |
CN102110140A (en) * | 2011-01-26 | 2011-06-29 | 桂林电子科技大学 | Network-based method for analyzing opinion information in discrete text |
CN102609512A (en) * | 2012-02-07 | 2012-07-25 | 北京中机科海科技发展有限公司 | System and method for heterogeneous information mining and visual analysis |
CN103838826A (en) * | 2014-01-23 | 2014-06-04 | 北京东方泰坦科技股份有限公司 | Integration method of dynamic heterogeneous space information plotting data |
Cited By (54)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111784615A (en) * | 2016-03-25 | 2020-10-16 | 北京三星通信技术研究有限公司 | Method and device for processing multimedia information |
CN107305555A (en) * | 2016-04-20 | 2017-10-31 | ***通信集团河北有限公司 | Data processing method and device |
CN106021485B (en) * | 2016-05-19 | 2019-05-14 | 中国传媒大学 | A kind of polynary attribute cinematic data visualization system |
CN106021485A (en) * | 2016-05-19 | 2016-10-12 | 中国传媒大学 | Multi-element attribute movie data visualization system |
WO2018040102A1 (en) * | 2016-09-05 | 2018-03-08 | 华为技术有限公司 | Audio processing method and device |
US11042587B2 (en) | 2016-09-05 | 2021-06-22 | Honor Device Co., Ltd. | Performing behavior analysis on audio track data to obtain a name of an application |
CN106484804A (en) * | 2016-09-22 | 2017-03-08 | 北京金堤科技有限公司 | Key element based on electric field simulation and relation methods of exhibiting |
CN106547891A (en) * | 2016-10-28 | 2017-03-29 | 浙江大学 | For the quick visualization method of the pictured text message of palm display device |
CN106528534A (en) * | 2016-11-09 | 2017-03-22 | 天津赛因哲信息技术有限公司 | Relevant word extraction method based on proper nouns |
CN106649585A (en) * | 2016-11-18 | 2017-05-10 | 福建中金在线信息科技有限公司 | Retrieval method and device |
CN106844636A (en) * | 2017-01-21 | 2017-06-13 | 亚信蓝涛(江苏)数据科技有限公司 | A kind of unstructured data processing method based on deep learning |
WO2018177162A1 (en) * | 2017-03-27 | 2018-10-04 | 华为技术有限公司 | Target retrieval method and device |
CN106980689B (en) * | 2017-03-31 | 2020-07-14 | 江苏赛睿信息科技股份有限公司 | Method for realizing data visualization through voice interaction |
CN106980689A (en) * | 2017-03-31 | 2017-07-25 | 邢加和 | A kind of method that data visualization is realized by interactive voice |
CN109101500A (en) * | 2017-06-20 | 2018-12-28 | 北京国双科技有限公司 | A kind of methods of exhibiting and device of file data |
CN107343221B (en) * | 2017-08-14 | 2020-01-03 | 彭君毅 | Online multimedia interaction system and method |
CN107343221A (en) * | 2017-08-14 | 2017-11-10 | 彭君毅 | A kind of online multimedia interaction systems and method |
CN107656760A (en) * | 2017-09-28 | 2018-02-02 | 联想(北京)有限公司 | Data processing method and device, electronic equipment |
WO2019101067A1 (en) * | 2017-11-21 | 2019-05-31 | 众安信息技术服务有限公司 | Information processing method and apparatus for data visualization |
CN108197295A (en) * | 2018-01-22 | 2018-06-22 | 重庆邮电大学 | Application process of the attribute reduction based on more granularity attribute trees in text classification |
CN108197295B (en) * | 2018-01-22 | 2022-03-11 | 重庆邮电大学 | Application method of attribute reduction in text classification based on multi-granularity attribute tree |
CN108319904A (en) * | 2018-01-24 | 2018-07-24 | 江苏金陵科技集团有限公司 | A kind of virtual interactive interface method assisting face or Car license recognition |
CN108416264A (en) * | 2018-01-29 | 2018-08-17 | 山东汇贸电子口岸有限公司 | A kind of searching method and search module of supporting OCR to input |
CN108829765A (en) * | 2018-05-29 | 2018-11-16 | 平安科技(深圳)有限公司 | A kind of information query method, device, computer equipment and storage medium |
WO2019227582A1 (en) * | 2018-05-29 | 2019-12-05 | 平安科技(深圳)有限公司 | Information query method and apparatus, computer device, and storage medium |
CN109146190A (en) * | 2018-08-31 | 2019-01-04 | 饶智彬 | Intelligent sound based on interior way system and expansion visualizes real-time situation sensory perceptual system |
CN109063198A (en) * | 2018-09-10 | 2018-12-21 | 浙江广播电视集团 | Melt the multidimensional visual search recommender system of media resource |
CN109063198B (en) * | 2018-09-10 | 2022-02-11 | 浙江广播电视集团 | Multi-dimensional visual search recommendation system for fusing media resources |
CN109344253A (en) * | 2018-09-18 | 2019-02-15 | 平安科技(深圳)有限公司 | Add method, apparatus, computer equipment and the storage medium of user tag |
CN109508453A (en) * | 2018-09-28 | 2019-03-22 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Across media information target component correlation analysis systems and its association analysis method |
CN111382282A (en) * | 2018-12-28 | 2020-07-07 | 北京国双科技有限公司 | Method, device, storage medium and processor for processing data |
CN111402935A (en) * | 2019-01-03 | 2020-07-10 | 李庆成 | Method for playing audio and video data |
CN109977155A (en) * | 2019-03-12 | 2019-07-05 | 福建奇点时空数字科技有限公司 | A kind of big data visualization system and its control method based on human-computer interaction |
CN109948009A (en) * | 2019-03-12 | 2019-06-28 | 福建奇点时空数字科技有限公司 | A kind of data visualization engine system |
CN110012349A (en) * | 2019-06-04 | 2019-07-12 | 成都索贝数码科技股份有限公司 | A kind of news program structural method and its structuring frame system end to end |
CN110347561A (en) * | 2019-06-11 | 2019-10-18 | 平安科技(深圳)有限公司 | Monitoring alarm method and terminal device |
CN110781237A (en) * | 2019-07-01 | 2020-02-11 | 厦门美域中央信息科技有限公司 | Information visualization analysis system based on artificial intelligence |
CN110995774A (en) * | 2019-09-30 | 2020-04-10 | 大唐可再生能源试验研究院有限公司 | Universal SCADA system for wind power plant |
CN110995774B (en) * | 2019-09-30 | 2024-01-05 | 大唐可再生能源试验研究院有限公司 | Universal SCADA system for wind farm |
CN110795392A (en) * | 2019-10-29 | 2020-02-14 | 北京四方启点科技有限公司 | Packaging method of accounting electronic voucher and electronic equipment |
CN110795392B (en) * | 2019-10-29 | 2022-04-19 | 北京四方启点科技有限公司 | Packaging method of accounting electronic voucher and electronic equipment |
CN111291896A (en) * | 2020-02-03 | 2020-06-16 | 深圳前海微众银行股份有限公司 | Interactive random forest subtree screening method, device, equipment and readable medium |
CN111611457A (en) * | 2020-05-20 | 2020-09-01 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
CN111611457B (en) * | 2020-05-20 | 2024-01-02 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
CN112307269B (en) * | 2020-08-24 | 2022-09-30 | 山东理工职业学院 | Intelligent analysis system and method for human-object relationship in novel |
CN112307269A (en) * | 2020-08-24 | 2021-02-02 | 山东理工职业学院 | Intelligent analysis system and method for human-object relationship in novel |
CN112667185A (en) * | 2021-03-16 | 2021-04-16 | 北京航空航天大学 | Maintainability analysis visual display method and system |
CN112667185B (en) * | 2021-03-16 | 2021-07-06 | 北京航空航天大学 | Maintainability analysis visual display method and system |
CN112818275A (en) * | 2021-04-16 | 2021-05-18 | 泰德网聚(北京)科技股份有限公司 | Image-text resource management system based on audio and video |
CN115985509A (en) * | 2022-12-14 | 2023-04-18 | 广东省人民医院 | Medical imaging data retrieval system, method, device and storage medium |
CN116189193A (en) * | 2023-04-25 | 2023-05-30 | 杭州镭湖科技有限公司 | Data storage visualization method and device based on sample information |
CN116189193B (en) * | 2023-04-25 | 2023-11-10 | 杭州镭湖科技有限公司 | Data storage visualization method and device based on sample information |
CN116719785A (en) * | 2023-08-09 | 2023-09-08 | 北京量子伟业信息技术股份有限公司 | Database management system based on metadata |
CN118093923A (en) * | 2024-04-26 | 2024-05-28 | 鸣启数字科技(山东)有限公司 | Data element visual management system based on big data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105005578A (en) | Multimedia target information visual analysis system | |
CN104239501B (en) | Mass video semantic annotation method based on Spark | |
CN106796578B (en) | Autoknowledge system and method and memory | |
Hsu | Content-based text mining technique for retrieval of CAD documents | |
US20150120777A1 (en) | System and Method for Mining Data Using Haptic Feedback | |
US20140229476A1 (en) | System for Information Discovery & Organization | |
US11989662B2 (en) | Methods and systems for base map and inference mapping | |
EP3940555A2 (en) | Method and apparatus of processing information, method and apparatus of recommending information, electronic device, and storage medium | |
CN102663010A (en) | Personalized image browsing and recommending method based on labelling semantics and system thereof | |
Lu et al. | Scalable news recommendation using multi-dimensional similarity and Jaccard–Kmeans clustering | |
Alsakran et al. | Real-time visualization of streaming text with a force-based dynamic system | |
Bernard et al. | Content-based layouts for exploratory metadata search in scientific research data | |
Sun et al. | PGCNet: patch graph convolutional network for point cloud segmentation of indoor scenes | |
KR20210129465A (en) | Apparatus for managing laboratory note and method for searching laboratory note using thereof | |
Varfolomeyev et al. | Smart personal assistant for historical tourism | |
US9875443B2 (en) | Unified attractiveness prediction framework based on content impact factor | |
Seifert et al. | Visual analysis and knowledge discovery for text | |
CN117556067A (en) | Data retrieval method, device, computer equipment and storage medium | |
CN116975403A (en) | Content retrieval model, content retrieval processing method and device and computer equipment | |
Zhuge | Dimensionality on summarization | |
JP2003316819A (en) | Object classification researching device and program for executing it | |
CN115130453A (en) | Interactive information generation method and device | |
Salmam et al. | Prediction in OLAP data cubes | |
Meinecke et al. | A visual analytics framework for composing a hierarchical classification for medieval illuminations | |
Hai et al. | Improving The Efficiency of Semantic Image Retrieval using A Combined Graph and SOM Model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151028 |
|
WD01 | Invention patent application deemed withdrawn after publication |