CN115018215A

CN115018215A - Population residence prediction method, system and medium based on multi-modal cognitive map

Info

Publication number: CN115018215A
Application number: CN202210946542.2A
Authority: CN
Inventors: 张广志; 成立立; 于笑博; 刘畔青
Original assignee: Beiling Rongxin Datalnfo Science and Technology Ltd
Current assignee: Beiling Rongxin Datalnfo Science and Technology Ltd
Priority date: 2022-08-09
Filing date: 2022-08-09
Publication date: 2022-09-06
Anticipated expiration: 2042-08-09
Also published as: CN115018215B

Abstract

The embodiment of the application provides a population residence prediction method, a population residence prediction system and a population residence prediction medium based on a multi-mode cognitive map. The method comprises the following steps: the method comprises the steps of constructing a multi-modal cognitive map of a region intelligent body according to region data, establishing a primary region appearance cognitive system, carrying out multi-modal population data recognition and population attribute and region parking event relation extraction on the established population multi-modal data based on the multi-modal cognitive map of the region intelligent body, carrying out population attribute linkage and cognitive fusion on the extracted population data multi-modal to obtain population data multi-modal cognition, and carrying out cognitive processing according to a cognitive map and a preset logical reasoning rule to predict population of a residential region; therefore, the multi-modal cognition map of the regional intelligent agent is constructed to identify and extract the attributes and event relations of the multi-modal population data, the attribute linkage and cognition fusion are carried out on the extracted multi-modal population to obtain the multi-modal cognition of the population data, and the cognition processing is carried out according to the cognition map and the rule to predict the residential population in the region.

Description

Population residence prediction method, system and medium based on multi-modal cognitive map

Technical Field

The application relates to the field of knowledge engineering in the field of big data and artificial intelligence, in particular to a population residence prediction method, a population residence prediction system and a population residence prediction medium based on a multi-mode cognitive map.

Background

Artificial intelligence has moved from computational intelligence, perceptual intelligence, to cognitive intelligence stages. Cognition is the process of acquiring, processing and applying knowledge by an individual, which is a high-level information processing mode of human brain; cognitive intelligence enables a machine to have the capabilities of reading and understanding semantics, logical reasoning and learning judgment. Two cores of machine-aware intelligence are "understanding" and "interpretation". The realization of cognitive intelligence needs to take knowledge as a driving force, which relates to key technologies such as knowledge representation, semantic understanding, associative reasoning, intelligent question answering, emotion calculation, decision planning and the like.

With the rise of deep learning, artificial intelligence is facing a new development climax. One bottleneck in the development of artificial intelligence is how to let machines know human knowledge, and it is extremely difficult for machines to understand and know a great deal of this knowledge, which is the necessary way to develop strong artificial intelligence.

The appearance of cognitive profiles has injected "accelerators" for the development of cognitive intelligence. However, the application of the current cognitive maps is shallow, particularly the invention is invented for the social application problem related to the processing of big data, the current big data method for the investigation and statistics of regional population is lack of a dynamic, comprehensive and accurate processing means, and a method for comprehensively and accurately predicting the regional population by means of a multi-modal cognitive map is not provided.

In view of the above problems, an effective technical solution is urgently needed.

Disclosure of Invention

The embodiment of the application aims to provide a population living prediction method, a system and a medium based on a multi-modal cognitive map, which can identify and extract the attributes and event relations of population multi-modal data according to a multi-modal cognitive map of a constructed region intelligent agent, perform attribute linking and cognitive fusion on the extracted population multi-modal to obtain population data multi-modal cognition, and perform cognitive processing according to the cognitive map and rules to predict the population of the region living.

The embodiment of the application also provides a population residence prediction method based on the multi-mode cognitive map, which comprises the following steps:

establishing a multi-mode cognitive map of a region intelligent agent according to region data, and establishing a primary region appearance cognitive system;

acquiring population circulation data of the region, establishing population multi-mode data, and performing population data multi-mode recognition and population attribute and region parking event relation extraction on the population multi-mode data based on the region intelligent agent multi-mode cognitive map;

performing population attribute linking and cognition fusion on the extracted population data in a multi-mode manner to obtain population data multi-mode cognition;

and performing cognitive processing according to the cognitive map and a preset logical reasoning rule to predict regional resident population.

Optionally, in the method for predicting population occupancy based on a multi-modal cognitive map according to the embodiment of the present application, the building a multi-modal cognitive map of a geographic intelligence entity according to geographic data, and building a preliminary geographic aspect cognitive system includes:

acquiring region characteristic data of a target region, wherein the region characteristic data comprises region characteristic data, house capacity data, region function data and building characteristic data;

constructing a space coordinate system and a scale of the target region and region graphic unit data according to the region characteristic data;

establishing a region scene model according to the region graphic unit data, and carrying out digital description on the region scene model;

extracting color information of the region model scenery model and combining the digital descriptor to perform rasterization processing to construct a virtual reality scene of the target region;

and constructing a multi-modal cognitive map of the region of the virtual reality scene according to the region characteristic data, mapping the position relation of various objects in the region scene on the space and the incidence relation of various logics according to a space coordinate system, and establishing primary cognition for the region appearance.

Optionally, in the method for predicting population occupancy based on the multi-modal cognitive map according to the embodiment of the present application, the acquiring population circulation data of the region and establishing population multi-modal data, and performing population data multi-modal recognition and population attribute and region parking event relationship extraction on the population multi-modal data based on the region intelligent agent multi-modal cognitive map includes:

acquiring population circulation data of the target region, wherein the population circulation data comprises population flow data, people flow image data, people flow video monitoring data and census data;

establishing population multi-modal data according to the population circulation data;

recognizing the population multi-modal data according to the region intelligent agent multi-modal cognitive map and pre-training data, and establishing correspondence and dependency relationship of the population multi-modal data;

population image recognition, population data recognition, target population extraction, population relationship extraction, population attribute extraction and docking event extraction are performed based on the population multimodal data.

Optionally, in the method for predicting population occupancy based on a multi-modal cognitive map according to the embodiment of the present application, the performing population image recognition, population data recognition, target population extraction, population relationship extraction, population attribute extraction and parking event extraction based on the population multi-modal data includes:

the population image recognition comprises people stream image segmentation, target detection and recognition, frequency threshold comparison and appearance similarity calculation are carried out according to segmented people stream individuals and images in the multi-modal cognitive map of the regional intelligent agent, and if the similarity probability exceeds a preset threshold, the same target population individual is judged;

the population data identification comprises data word segmentation processing, keyword labeling and population individual identification;

extracting atomic information elements in the population multimodal data to perform the target population extraction based on a knowledge base and a dictionary;

the population relationship extraction and the population attribute extraction comprise population attribute relationship extraction, human-house relationship extraction, region and stream relationship extraction and building population relationship extraction based on preset rules;

the parking event extraction is to extract and structurally express parking event information between people flow and regional buildings, and comprises open domain or limited domain parking event extraction and parking reason relationship extraction.

Optionally, in the method for predicting population occupancy based on a multi-modal cognitive map according to the embodiment of the present application, performing population attribute linking and cognitive fusion on the extracted population data in a multi-modal manner to obtain population data multi-modal cognition, including:

corresponding the obtained same population individual to the same correct population individual in the cognitive library;

judging whether the same individual or related individuals exist according to the population individuals in the preset population database;

acquiring population individual objects through population attribute extraction and obtaining multi-mode population data links corresponding to correct population individuals in the cognitive library;

merging the multi-modal cognitive maps of the geographical intelligent agents into the cognitive library according to the constructed multi-modal cognitive maps to complete multi-modal cognitive combination, wherein the merging comprises merging of a data layer and a mode layer;

the data layer fusion comprises fusion of population individuals and fusion of population attributes;

the fusion of the mode layer comprises the fusion of the upper and lower bit relations of the data and the fusion of the definition of the data attribute.

Optionally, in the method for predicting population occupancy based on a multi-modal cognitive map according to the embodiment of the present application, the predicting the population of the region occupancy by performing cognitive processing according to the cognitive map and a preset logical inference rule includes:

performing cognitive processing according to the multi-modal cognitive map of the regional intelligent agent and a preset logical reasoning rule, wherein the cognitive processing comprises body construction, cognitive reasoning and result evaluation;

the ontology is constructed in a data automation driving mode, and the ontology construction process comprises population parallel relationship similarity calculation, population superior-inferior relationship extraction and ontology generation;

the cognitive inference enriches the multi-modal cognitive maps of the regional intelligent agents by acquiring new associations between population individuals and new associations between individual regions through acquiring relationships between population regions and relationships between individual region parking events according to a logical inference rule on the basis of the multi-modal cognitive maps of the regional intelligent agents;

the result evaluation includes accuracy and coverage evaluation.

Optionally, in the method for predicting population occupancy based on a multi-modal cognitive map according to the embodiment of the present application, the cognitive inference enriches the multi-modal cognitive map of the domain agent by obtaining new associations between population individuals and new associations between individual domains based on the multi-modal cognitive map of the domain agent according to a logical inference rule based on relationships between population domains and relationships between individual domain parking events, including:

the reasoning mode of the logical reasoning rule comprises deductive reasoning, inductive reasoning, analogy reasoning, cause reasoning, deterministic reasoning and uncertainty reasoning;

and carrying out logical reasoning by the numerical model method of uncertainty reasoning based on the fuzzy theory.

In a second aspect, the present application provides a population occupancy prediction system based on a multi-modal cognitive map, the system including: a memory and a processor, wherein the memory includes a multi-modal cognitive map-based population occupancy prediction method program, and the multi-modal cognitive map-based population occupancy prediction method program when executed by the processor implements the following steps:

establishing a multi-modal cognitive map of a region intelligent agent according to region data, and establishing a primary region appearance cognitive system;

and performing cognitive processing according to the cognitive map and a preset logical reasoning rule to predict the residential population of the region.

Optionally, in the system for predicting population residence based on a multi-modal cognitive map according to the embodiment of the present application, the building a multi-modal cognitive map of a domain intelligent agent according to domain data, and building a preliminary domain appearance cognitive system include:

In a third aspect, the present application further provides a computer-readable storage medium, where the computer-readable storage medium includes a multi-modal cognitive map-based population occupancy prediction method program, and when the multi-modal cognitive map-based population occupancy prediction method program is executed by a processor, the method implements the steps of the multi-modal cognitive map-based population occupancy prediction method described in any one of the above.

As can be seen from the above, the population residence prediction method, the system and the medium based on the multi-modal cognitive map provided in the embodiment of the present application construct the multi-modal cognitive map of the domain agent according to the domain data and establish a preliminary domain appearance cognitive system, perform multi-modal identification of population data and extraction of population attributes and domain parking event relations on the established multi-modal data of the population based on the multi-modal cognitive map of the domain agent, perform population attribute linking and cognitive fusion on the extracted population data in multiple modes to obtain multi-modal cognition of the population data, and perform cognitive processing according to the cognitive map and preset logical reasoning rules to predict the population of the domain residence; therefore, the multi-modal cognition map of the regional intelligent agent is constructed to identify and extract the attributes and event relations of the multi-modal population data, the attribute linkage and cognition fusion are carried out on the extracted multi-modal population to obtain the multi-modal cognition of the population data, and the cognition processing is carried out according to the cognition map and the rule to predict the residential population in the region.

Additional features and advantages of the present application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the present application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and that those skilled in the art can also obtain other related drawings based on the drawings without inventive efforts.

Fig. 1 is a flowchart of a population occupancy prediction method based on a multi-modal cognitive map according to an embodiment of the present application;

fig. 2 is a flowchart of a method for predicting population occupancy based on a multi-modal cognitive map for constructing a multi-modal cognitive map of a geographic intelligence entity and a preliminary geographic landscape cognitive system according to an embodiment of the present disclosure;

fig. 3 is a flowchart of multimodal identification of population data and extraction of relationship between population attributes and geographic parking events in the population residence prediction method based on the multimodal cognitive map according to the embodiment of the present application;

fig. 4 is a schematic structural diagram of a population occupancy prediction system based on a multi-modal cognitive map according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that like reference numerals and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, in the description of the present application, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.

Referring to fig. 1, fig. 1 is a flow chart of a population occupancy prediction method based on a multi-modal cognitive map in some embodiments of the present application. The population residence prediction method based on the multi-modal cognitive map is used in terminal equipment, such as computers, mobile phone terminals and the like. The population residence prediction method based on the multi-modal cognitive map comprises the following steps of:

s101, establishing a multi-modal cognitive map of a region intelligent agent according to region data, and establishing a primary region appearance cognitive system;

s102, acquiring population circulation data of the region, establishing population multi-mode data, and performing population data multi-mode recognition and population attribute and region parking event relation extraction on the population multi-mode data based on the region intelligent agent multi-mode cognitive map;

s103, performing population attribute linking and cognition fusion on the extracted population data in a multi-mode manner to obtain population data multi-mode cognition;

and S104, performing cognitive processing according to the cognitive map and a preset logical reasoning rule to predict the residential population of the region.

The technology comprises the steps of carrying out entity recognition, relationship extraction and cognitive fusion on population multi-modal data established by acquired population circulation data in a region by constructing a region intelligent body multi-modal cognitive map to obtain population data multi-modal cognition, finally carrying out cognitive processing according to a cognitive map and a logic inference rule to predict a region resident population to obtain a technology for carrying out recognition, extraction, cognition and processing on the population multi-modal data according to the region cognitive map to obtain the region resident population prediction The cognitive processing of the relational multi-modal network atlas according to the cognitive atlas and a preset logical inference rule comprises ontology construction, cognitive inference and result evaluation cognitive processing, wherein the ontology construction is a semantic basis for communication of events in the intelligent multi-modal cognitive atlas and is automatically constructed by deep learning drive, the cognitive inference is based on the existing intelligent multi-modal cognitive atlas, new associations among population individuals and new associations among individual domains are found by calculating the relationships among the population domains and the relationships among the individual domain parking events according to the preset logical inference rule, and the method is an important means for updating the intelligent multi-modal cognitive atlas, and the result evaluation is a final inspection link of the cognitive processing and ensures the reasonability of the intelligent multi-modal cognitive atlas.

Referring to fig. 2, fig. 2 is a flowchart illustrating a geographic intelligence multi-modal cognitive map and a preliminary geographic context cognitive system in a multi-modal cognitive map-based population residence prediction method according to some embodiments of the present disclosure. According to the embodiment of the invention, the method specifically comprises the following steps:

s201, obtaining region characteristic data of a target region, wherein the region characteristic data comprises region characteristic data, house capacity data, region function data and building characteristic data;

s202, constructing a space coordinate system and a scale of the target region and region graphic unit data according to the region characteristic data;

s203, establishing a region scene model according to the region graphic unit data, and carrying out digital description on the region scene model;

s204, extracting color information of the region model scene model and performing rasterization processing by combining the digital description element to construct a virtual reality scene of the target region;

s205, a multi-modal geographical cognition map of the virtual reality scene is constructed according to the geographical feature data, the spatial position relation and various logical incidence relation of various objects in the geographical scene are mapped according to a spatial coordinate system, and preliminary cognition is established on geographical appearances.

It should be noted that, in order to establish initial cognition on the geographical features, a geographical multi-modal cognitive map generating a virtual reality scene is constructed through the acquired characteristic data of the target geographical region, and a spatial coordinate system is combined to map the spatial position relationship and logical association relationship of each characteristic object in the geographical scene, including regional building buildings and the like, wherein the spatial coordinate system includes a model coordinate system, a world coordinate system and an observation coordinate system; the model in the model coordinate system is a three-dimensional object, each object has its own model coordinate system, the model coordinate system is an imaginary coordinate system, the relative position of the coordinate system and the object is invariable all the time, the world coordinate system is a real 3D scene of our life, the model coordinate in the model coordinate system is transformed into world coordinates after being multiplied by the model matrix, the observation coordinate system is a Camera coordinate system, the Camera coordinate system can also be called uvn coordinate system, and corresponds to the three XYZ axes of the world coordinate system; a multi-mode cognitive map for building a region scene model is characterized in that the cognitive map is based on a space coordinate system and time sequence mapping position relations and logic incidence relations of various objects in a region scene, entities in the cognitive map are divided into logic entities and three-dimensional graphic entities, the logic entities refer to entities on character concepts, the three-dimensional graphic entities refer to visual three-dimensional graphics, the logic entities are further divided into logic entities and event entities, the logic entities can correspond to the three-dimensional graphic entities one by one, the event entities are combinations of a series of dynamic change processes of related objects, the entities can have various attributes such as flow, identity, gender, age, flow direction, time density and the like of population flows, the attributes can be in a character concept form such as census data, and can also be in a graphic or video form such as capturing video pictures or screen capture images, the entities may have various relationships, such as spatial, temporal, or logical relationships, such as individual relationships between population groups, group classification, group attribute relationships, and the like; establishing a region scene model including objects of various real regions such as region landform, building layout, region arrangement, region space and the like, wherein the basic steps of establishing main graphic operation of the scene are firstly establishing a scene model according to a basic graphic unit and mathematically describing the established model, then placing the scene model at a proper position in a three-dimensional space and setting a view point to observe a target scene and then calculating the colors of all objects in the model, wherein the colors are determined according to application requirements, simultaneously determining an illumination condition and a texture pasting mode, and finally converting the mathematical description of the scene model and the color information thereof to a computer screen for rasterization.

Referring to fig. 3, fig. 3 is a flow chart of multimodal identification of population data and population attribute and geographic parking event relationship extraction in a population residence prediction method based on multimodal cognitive mapping according to some embodiments of the present application. According to the embodiment of the invention, acquiring the population circulation data of the region and establishing population multi-modal data, performing population data multi-modal identification and population attribute and region parking event relation extraction on the population multi-modal data based on the region intelligent agent multi-modal cognitive map specifically comprises the following steps:

s301, acquiring population circulation data of the target region, wherein the population circulation data comprises population flow data, people flow image data, people flow video monitoring data and census data;

s302, establishing population multimodal data according to the population circulation data;

s303, identifying the population multi-modal data according to the region intelligent agent multi-modal cognitive map and pre-training data, and establishing correspondence and dependency relationship of the population multi-modal data;

and S304, performing population image identification, population data identification, target population extraction, population relation extraction, population attribute extraction and parking event extraction based on the population multimodal data.

It should be noted that the multi-modal cognitive map of the domain intelligent agent has a multi-modal data recognition capability, the recognition capability of the multi-modal cognitive map of the domain intelligent agent to the multi-modal population data can be realized by training the cognitive map by using the known corresponding relationship and the classification attribute multi-modal data as pre-training data, the multi-modal population data is established by data acquired by the population circulation condition of a target domain, population flow data such as images, audios and videos, people flow image data, people flow video monitoring data and population census data, the corresponding and dependency relationship of the multi-modal population data is identified and established in the multi-modal cognitive map of the domain intelligent agent according to the multi-modal population data, then classification entity extraction and event extraction are carried out, the entity extraction means that a specific element label is identified in a multi-modal data source and is linked with a label in an entity library, the target population extraction is to identify attribute tags conforming to the target population according to population attributes and link the attribute tags with tags in a population attribute library, the entity relationship extraction is to find the relationship among entities in the multi-modal data source, the population relation extraction can be divided into global extraction and local extraction, wherein the population relation extraction is to find the relations between population individuals and between individuals and populations in a population multi-mode data source, the entity attribute extraction is the relation between entities and attributes thereof, namely the correlation between population individuals and population attributes, and the event extraction is to extract and structurally express the event information in the multi-modal data source and comprises the steps of event extraction, event relation extraction, the system is characterized in that population individuals and groups in the population multimodal data source are extracted and structurally represented in regional parking time, location, origin and parking pass, front and back parking processes and the relationship between parking and regions.

According to the embodiment of the invention, the population image recognition, population data recognition, target population extraction, population relationship extraction, population attribute extraction and parking event extraction based on the population multimodal data specifically comprise:

The image segmentation comprises the steps of inputting an image into a network to obtain a corresponding feature map, using an RPN structure to generate a candidate frame, projecting the candidate frame onto the feature map to obtain a corresponding feature matrix, carrying out scaling on each feature matrix to obtain the feature map, flattening, carrying out scaling on each feature matrix through a series of full-connection layers, carrying out convolution to carry out deeper feature extraction, and finally attaching the feature matrix to a corresponding position in an original image to obtain a result map of example segmentation, wherein target detection and identification are carried out by comparing a segmented population individual with an existing individual image in a cognitive map, calculating a similarity probability through a similarity degree calculation method, judging the individual to be the same individual if the similarity probability exceeds a preset threshold value, and carrying out image comparison through searching if the similarity probability does not reach the result compared with the existing image in the cognitive map, and carrying out data word segmentation processing by a dictionary-based method (forward maximum matching algorithm, forward maximum matching algorithm, and forward maximum matching algorithm, Reverse maximum matching algorithm and two-way maximum matching method) and a method based on statistics, wherein the keyword labeling adopts a hidden Markov model, a perceptron and a conditional random field method, the population individual identification is to firstly combine the population attribute library of the existing cognitive map to assign weights to each rule, then judge the type according to the conformity degree of the individual and the rule, then use the hidden Markov model, the maximum entropy model and the conditional random field to label the locked individual recognition task as a sequence based on the pre-labeled sentences, and the target population mainly extracts the atomic information elements in the multi-modal population data, wherein the method based on the knowledge base and the dictionary is to use the knowledge base and the dictionary established by the existing cognitive map to match with the mode and the character string as the main means, and the method based on the statistics is the hidden Markov model based on the learning method of the counting machine, the method of the counting machine and the method of the multi-mode, Maximum entropy, support vector machines, conditional random fields, event extraction including open or defined domain resident event extraction and resident reason extraction by relationship, divided into meta-event extraction and topic event extraction, wherein meta-event represents occurrence of resident action of population or change of resident state, driven by verb and can also be triggered by noun capable of representing action, including individual resident place, time and associated individual or group participating in the resident action behavior, meta-event extraction method includes meta-event extraction based on pattern matching, meta-event extraction based on machine learning, extraction method based on neural network, topic event includes core event or activity and all directly related events and activities, can be composed of multiple meta-event segments, topic event extraction method includes topic event extraction based on event framework, topic event extraction based on ontology, the residence time extraction comprises residence event extraction of a residence event frame and residence relationship event extraction based on residence individuals, the population relationship extraction and the population attribute extraction comprise population attribute relationship extraction, people-room relationship extraction, region people flow relationship extraction and building population relationship extraction based on preset rules, extraction reflecting population individual and population attribute relationship, extraction of individual and population and region building relationship, extraction of the building population flow relationship of each region of the region and extraction of building and population individual and population relationship.

According to the embodiment of the invention, the population attribute linking and cognition fusion are performed on the extracted population data in a multi-mode manner to obtain the population data multi-mode cognition, and the method specifically comprises the following steps:

It should be noted that from the two aspects of population attribute layer and population individual layer, the population attributes, population individuals and resident events in a plurality of cognitive maps or information sources are linked through the modes of alignment, association, combination and the like of cognitive libraries to form a more uniform and dense intelligent multi-modal cognitive map, which is an important method for realizing cognitive sharing and reasoning, the cognitive fusion of the population attribute layer is mainly expressed as population attribute alignment and resident event alignment, which means the process of determining the mapping relationship between population attributes, concepts and relationships and the like, and determining resident events, resident event relationships and resident event attributes, the deep learning algorithm based on the intelligent multi-modal cognitive map is generally realized by calculating the similarity between population individuals and the similarity between resident events, and according to the natural language types, the single language alignment and the cross-language alignment are realized, the cognitive fusion of the population individual layer is mainly expressed as coreference resolution, population individual alignment and resident event alignment, wherein the coreference resolution is used for uniformly resolving different labels of the same individual and the same resident event in the same population individual information source, and the population individual alignment and the resident event alignment are used for uniformly resolving the same individual and the same resident event in different information sources so as to generate connection between the information sources.

According to the embodiment of the invention, the cognitive processing is performed according to the cognitive map and the preset logical reasoning rule to predict the residential population of the region, and the method specifically comprises the following steps:

the result evaluation includes accuracy and coverage evaluation.

It should be noted that the population parallel relationship similarity calculation is suitable for examining the index measure of how much any given two population individuals belong to the same attribute classification, and the higher the similarity is, the more likely the two population individuals belong to the same classification, so that the parallel relationship is relative to the longitudinal concept membership, and there are two methods for calculating the population parallel relationship similarity: the method comprises a mode matching method and a distribution similarity, wherein the mode matching method adopts a method of predefining individual pair modes, the frequency of common occurrence of given keyword combinations in the same semantic unit is obtained through mode matching, the similarity between individuals is calculated according to the frequency, the distribution similarity method is based on the premise that semantic similarity exists between the individuals frequently appearing in similar context pipe diameters, population upper and lower relation extraction is used for determining the membership relation between concepts, the main method is to extract individual pairs based on grammar modes or judge individual relations and distinguish upper and lower terms by using a probability model, and help to train a model by means of concept classification knowledge to improve algorithm precision, the main task of generating the body is to cluster the concepts obtained by each level of the population, carry out semantic class calibration on the concepts, and assign one or more common upper terms to the individuals in the class, the result evaluation is the final inspection link of cognitive processing, and the reasonability of the intelligent multi-modal cognitive map is ensured, wherein the accuracy rate refers to the degree that an individual and a relationship correctly represent a phenomenon in real life, and the accuracy rate can be further subdivided into three dimensions: syntactic accuracy, semantic accuracy and timeliness, and coverage refers to avoiding missing elements related to a domain or possibly generating incomplete query results or derived results, biased models.

According to the embodiment of the invention, the cognitive inference enriches the multi-modal cognitive maps of the regional intelligent agents by acquiring new associations between population individuals and new associations between individual regions through acquiring relationships between population regions and relationships between individual region parking events based on the multi-modal cognitive maps of the regional intelligent agents according to a logical inference rule, and specifically comprises the following steps:

the reasoning mode of the logic reasoning rule comprises deductive reasoning, inductive reasoning, analogy reasoning, cause reasoning, deterministic reasoning and uncertainty reasoning;

It should be noted that, according to the logic inference rule, the multi-mode cognition map of the regional intelligent agent is enriched by acquiring the new association between population and individuals and the new association between individual regions according to the relationship between population and regions and the relationship between individual region parking events, the cognition map is enriched by acquiring the association between population and individuals and the new association between individual regions according to the logic inference rule, the deductive inference is also called logic inference, from general to special, the inductive inference is from special to general, the analog inference is from special to special, because the inference is also called inverse deductive inference, from special to explanation, the deterministic inference means that the knowledge and evidence used in the inference are determined, the deduced conclusion is also determined, the true value is either true or false, the knowledge and evidence used in the inference of the deterministic inference are not both determined, and the conclusion is also uncertain, the uncertainty reasoning method adopts a numerical model method, and the numerical model method adopts a credibility method based on a fuzzy theory reasoning method, an evidence theory and a Bayesian reasoning method based on probability.

As shown in fig. 4, the present invention also discloses a population residence prediction system based on the multi-modal cognitive map, which includes a memory 41 and a processor 42, wherein the memory includes a population residence prediction method program based on the multi-modal cognitive map, and when executed by the processor, the population residence prediction method program based on the multi-modal cognitive map implements the following steps:

According to the embodiment of the invention, the multi-modal cognitive map of the region intelligent agent is constructed according to the region data, and a preliminary region appearance cognitive system is established, and the method specifically comprises the following steps:

It should be noted that, in order to establish initial cognition on the geographical features, a geographical multi-modal cognitive map generating a virtual reality scene is constructed through the acquired characteristic data of the target geographical region, and a spatial coordinate system is combined to map the spatial position relationship and logical association relationship of each characteristic object in the geographical scene, including regional building buildings and the like, wherein the spatial coordinate system includes a model coordinate system, a world coordinate system and an observation coordinate system; the model in the model coordinate system is a three-dimensional object, each object has its own model coordinate system, the model coordinate system is an imaginary coordinate system, the relative position of the coordinate system and the object is invariable all the time, the world coordinate system is a real 3D scene of our life, the model coordinate in the model coordinate system is transformed into world coordinates after being multiplied by the model matrix, the observation coordinate system is a Camera coordinate system, the Camera coordinate system can also be called uvn coordinate system, and corresponds to the three XYZ axes of the world coordinate system; a multi-mode cognitive map for building a region scene model is characterized in that the cognitive map is based on a space coordinate system and time sequence mapping position relations and logic incidence relations of various objects in a region scene, entities in the cognitive map are divided into logic entities and three-dimensional graphic entities, the logic entities refer to entities on character concepts, the three-dimensional graphic entities refer to visual three-dimensional graphics, the logic entities are further divided into logic entities and event entities, the logic entities can correspond to the three-dimensional graphic entities one by one, the event entities are combinations of a series of dynamic change processes of related objects, the entities can have various attributes such as flow, identity, gender, age, flow direction, time density and the like of population flows, the attributes can be in a character concept form such as census data, and can also be in a graphic or video form such as capturing video pictures or screen capture images, the entities can have various relationships, such as spatial, temporal, or logical relationships, such as individual relationships, group classifications, group attribute relationships, and the like between population groups; establishing a region scene model including objects of various real regions such as region landform, building layout, region arrangement, region space and the like, wherein the basic steps of establishing main graphic operation of the scene are firstly establishing a scene model according to a basic graphic unit and mathematically describing the established model, then placing the scene model at a proper position in a three-dimensional space and setting a view point to observe a target scene and then calculating the colors of all objects in the model, wherein the colors are determined according to application requirements, simultaneously determining an illumination condition and a texture pasting mode, and finally converting the mathematical description of the scene model and the color information thereof to a computer screen for rasterization.

According to the embodiment of the invention, the acquiring of the population circulation data of the region and the establishing of the population multi-modal data, and the performing of the population data multi-modal identification and the population attribute and region parking event relationship extraction on the population multi-modal data based on the region intelligent agent multi-modal cognitive map specifically comprise:

recognizing the population multimodal data according to the region intelligent agent multimodal cognitive map and pre-training data, and establishing correspondence and dependency relationship of the population multimodal data;

the population relationship extraction and the population attribute extraction comprise population attribute relationship extraction, people-room relationship extraction, region people flow relationship extraction and building population relationship extraction based on preset rules;

The image segmentation comprises the steps of inputting an image into a network to obtain a corresponding feature map, using an RPN structure to generate a candidate frame, projecting the candidate frame onto the feature map to obtain a corresponding feature matrix, carrying out scaling on each feature matrix to obtain the feature map, flattening, carrying out scaling on each feature matrix through a series of full-connection layers, carrying out convolution to carry out deeper feature extraction, and finally attaching the feature matrix to a corresponding position in an original image to obtain a result map of example segmentation, wherein target detection and identification are carried out by comparing a segmented population individual with an existing individual image in a cognitive map, calculating a similarity probability through a similarity degree calculation method, judging the individual to be the same individual if the similarity probability exceeds a preset threshold value, and carrying out image comparison through searching if the similarity probability does not reach the result compared with the existing image in the cognitive map, and carrying out data word segmentation processing by a dictionary-based method (forward maximum matching algorithm, forward maximum matching algorithm, and forward maximum matching algorithm, Reverse maximum matching algorithm and two-way maximum matching method) and statistics-based method, the keyword labeling adopts hidden Markov model, perceptron, conditional random field method, the population individual identification is to combine the population attribute library of the existing cognition map to assign weight to each rule, then to judge the type according to the conformity degree of the individual and the rule, then to use hidden Markov model, maximum entropy model, conditional random field to label the locked individual identification task as the sequence based on the pre-labeled sentence, the target population mainly extracts the atom information elements in the multi-modal population data, wherein the method based on knowledge base and dictionary is based on the knowledge base and dictionary established by the existing cognition map and matching the pattern and character string as the main means, the method based on statistics is based on hidden Markov model of the machine learning method, the method based on statistical method, Maximum entropy, support vector machines, conditional random fields, event extraction including open or defined domain resident event extraction and resident reason extraction by relationship, divided into meta-event extraction and topic event extraction, wherein meta-event represents occurrence of resident action of population or change of resident state, driven by verb and can also be triggered by noun capable of representing action, including individual resident place, time and associated individual or group participating in the resident action behavior, meta-event extraction method includes meta-event extraction based on pattern matching, meta-event extraction based on machine learning, extraction method based on neural network, topic event includes core event or activity and all directly related events and activities, can be composed of multiple meta-event segments, topic event extraction method includes topic event extraction based on event framework, topic event extraction based on ontology, the residence time extraction comprises residence event extraction of a residence event frame and residence relationship event extraction based on residence individuals, the population relationship extraction and the population attribute extraction comprise population attribute relationship extraction, people-room relationship extraction, region people flow relationship extraction and building population relationship extraction based on preset rules, extraction reflecting population individual and population attribute relationship, extraction of individual and population and region building relationship, extraction of the building population flow relationship of each region of the region and extraction of building and population individual and population relationship.

the cognitive inference enriches the multi-modal cognitive atlas of the regional intelligent agent by acquiring new associations between population individuals and new associations between individual regions through the relationship between population regions and the relationship between individual regional parking events according to a logical inference rule based on the multi-modal cognitive atlas of the regional intelligent agent;

the result evaluation includes accuracy and coverage evaluation.

According to the embodiment of the invention, the cognitive inference is based on the region intelligent multi-mode cognitive map, and the region intelligent multi-mode cognitive map is enriched by acquiring new relations among population individuals and new relations among individual regions through relations among population regions and relations among individual region parking events according to a logical inference rule, and specifically comprises the following steps:

and carrying out logical reasoning by the numerical model method of uncertainty reasoning based on a fuzzy theory.

It should be noted that, according to the logic inference rule, the multi-mode cognition map of the regional intelligent agent is enriched by acquiring the new association between population and individuals and the new association between individual regions according to the relationship between population and regions and the relationship between individual region parking events, the cognition map is enriched by acquiring the association between population and individuals and the new association between individual regions according to the logic inference rule, the deductive inference is also called logic inference, from general to special, the inductive inference is from special to general, the analog inference is from special to special, because the inference is also called inverse deductive inference, from special to explanation, the deterministic inference means that the knowledge and evidence used in the inference are determined, the deduced conclusion is also determined, the true value is either true or false, the knowledge and evidence used in the inference of the deterministic inference are not both determined, and the conclusion is also uncertain, the uncertainty inference method adopts a numerical model method, and the numerical model method adopts a credibility method based on an inference method of a fuzzy theory, an evidence theory and a Bayesian inference method based on probability.

A third aspect of the present invention provides a computer-readable storage medium, where the computer-readable storage medium includes a multi-modal cognitive map-based population occupancy prediction method program, and when the multi-modal cognitive map-based population occupancy prediction method program is executed by a processor, the method implements the steps of the multi-modal cognitive map-based population occupancy prediction method described in any one of the above.

The invention discloses a population residence prediction method, a system and a medium based on a multi-modal cognitive map, wherein the multi-modal cognitive map of a region agent is constructed according to region data, a preliminary region appearance cognitive system is established, the multi-modal cognitive map of the region agent is used for carrying out multi-modal recognition on the population multi-modal data established and population attributes and region parking event relation extraction, population attribute linkage and cognitive fusion are carried out on the extracted population data multi-modal to obtain population data multi-modal cognition, and cognitive processing is carried out according to the cognitive map and a preset logic reasoning rule to predict the population of the region residence; therefore, the multi-modal cognition map of the regional intelligent agent is constructed to identify and extract the attributes and event relations of the multi-modal population data, the attribute linkage and cognition fusion are carried out on the extracted multi-modal population to obtain the multi-modal cognition of the population data, and the cognition processing is carried out according to the cognition map and the rule to predict the residential population in the region.

In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of the unit is only a logical functional division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the coupling, direct coupling or communication connection between the components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between the devices or units may be electrical, mechanical or other forms.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units; can be located in one place or distributed on a plurality of network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated into one unit; the integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.

Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium includes: a removable storage device, a read-only memory, a random access memory, a magnetic or optical disk, or other various media that can store program code.

Alternatively, the integrated unit of the present invention may be stored in a readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. Based on such understanding, the technical solutions of the embodiments of the present invention may be essentially implemented or a part contributing to the prior art may be embodied in the form of a software product stored in a storage medium, and including several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, a ROM, a RAM, a magnetic or optical disk, or various other media that can store program code.

Claims

1. The population residence prediction method based on the multi-modal cognitive map is characterized by comprising the following steps of:

2. The population residence prediction method based on the multi-modal cognition graph according to claim 1, wherein the building of the multi-modal cognition graph of the geographical intelligent object according to the geographical data and the building of the preliminary geographical appearance cognition system comprises:

3. The population residence prediction method based on the multi-modal cognitive map as claimed in claim 2, wherein the step of obtaining population circulation data of the region and establishing population multi-modal data, and the step of performing population data multi-modal recognition and population attribute and region parking event relationship extraction on the population multi-modal data based on the region agent multi-modal cognitive map comprises the steps of:

4. The population occupancy prediction method based on the multi-modal cognitive atlas of claim 3, wherein the population multi-modal data based population image recognition, population data recognition, target population extraction, population relationship extraction, population attribute extraction and parking event extraction comprises:

5. The population occupancy prediction method based on the multi-modal cognition graph according to claim 4, wherein the performing population attribute linking and cognition fusion on the extracted population data in a multi-modal mode to obtain the population data in a multi-modal cognition, specifically comprises:

6. The population occupancy prediction method based on the multi-modal cognitive map as claimed in claim 5, wherein the cognitive processing performed according to the cognitive map and the preset logical inference rule to predict the population of the regional occupancy comprises:

the result evaluation includes accuracy and coverage evaluation.

7. The population occupancy prediction method based on the multi-modal cognitive map as claimed in claim 6, wherein the cognitive inference enriches the multi-modal cognitive map of the domain agent by obtaining new associations between population individuals and new associations between individual domains based on the multi-modal cognitive map of the domain agent according to logical inference rules, and by obtaining new associations between population domains and new associations between individual domains based on the inter-domain relationships and the inter-individual parking events of the population, the method comprises:

8. A system for predicting population occupancy based on a multi-modal cognitive profile, the system comprising: a memory and a processor, wherein the memory includes a multi-modal cognitive map-based population occupancy prediction method program, and the multi-modal cognitive map-based population occupancy prediction method program when executed by the processor implements the following steps:

9. The system for predicting population occupancy according to claim 8, wherein the system for building a multi-modal cognitive map of a geographic intelligence entity according to geographic data and building a preliminary geographic facies awareness system comprises:

10. Computer-readable storage medium, characterized in that the computer-readable storage medium comprises a multi-modal cognitive map-based population occupancy prediction method program, which when executed by a processor, implements the steps of the multi-modal cognitive map-based population occupancy prediction method according to any one of claims 1 to 7.