69 results about "Semantic mapping" patented technology

The invention provides a method for generating an output indicative of acoustical sound transmission in a room. By using e.g. a point cloud representation of an acoustic environment, it is possible to calculate its acoustics from the interior information obtained from the depth camera. This approach is suitable e.g. for run-time applications since it is not based on an audible excitation that can disturb running audio. Also, the point-cloud model can be updated in real time according to the scene changes detected by depth-camera. This allows efficient acoustical simulation of dynamic, interactive environments. Although only geometrical information of a room is provided, high amount of surface details leaves possibility for implementation of material recognition algorithms that involve semantic mapping. This can provide information of reflective properties of surfaces or objects at a point level. Also, a high amount of details allows a good approximation of complex geometries, e.g. porous materials, and rough surfaces, thus a more natural simulation of wave phenomena like diffraction and scattering is possible.

The invention relates to the field of cross fusion of computer vision and deep learning, in particular to a semantic mapping method based on visual SLAM and two-dimensional semantic segmentation. Themethod comprises the following steps: S1, calibrating camera parameters, and correcting camera distortion; S2, acquiring an image frame sequence; S3, preprocessing the image; S4, judging whether the current image frame is a key frame or not, if so, turning to the step S6, and if not, turning to the step S5; S5, performing dynamic fuzzy compensation; s6, carrying out semantic segmentation, extracting ORB feature points for the image frames, and carrying out semantic segmentation by using a mask region convolutional neural network algorithm model; s7, pose calculation: utilizing a sparse SLAM algorithm model to calculate the pose of the camera; s8, using the semantic information for assisting in dense semantic map construction, and achieving three-dimensional semantic map construction of theglobal point cloud map. According to the invention, the performance of the unmanned aerial vehicle semantic mapping system can be improved, and the robustness of feature point extraction and matchingfor a dynamic scene is significantly improved.

The invention discloses a novel semantic mapping system based on instant positioning and mapping and three-dimensional point cloud semantic segmentation, and belongs to the technical field of computervision and artificial intelligence. According to the method, a sparse map is established by utilizing instant positioning and mapping, key frames and camera poses are obtained, and semantic segmentation is carried out based on the key frames by utilizing point cloud semantic segmentation. A two-dimensional target detection method and point cloud splicing are utilized to obtain a truncated body suggestion, a Bayesian updating scheme is designed to integrate semantic tags of candidate truncated bodies, and points with final correction tags are inserted into an established sparse map. Experiments show that the system has high efficiency and accuracy.

The invention discloses a fusion method of building BIM model and live-action three-dimensional model. The fusion and conversion process mainly comprises model reconstruction, geometric information conversion, model three-dimensional registration and semantic mapping. The fusion method comprises the steps that firstly, a building BIM model is split, and json attribute files and ifc geometric filesof all components are generated; then, intermediate format exchange is carried out, and ifc is converted into obj; obj is converted into glTF, and three-dimensional geometric figure data transmissionis further achieved; then attributes and geometric data are connected, and glTF and json generate b3dm; and finally, a data description file tileset.json are added, and the data description file tileset.json and b3dm are converted into 3DTiles data. According to the method, the semantic attribute characteristics in the BIM original model are reserved while the geometric figures are completely expressed, and the spatial positions of the two models are unified in the same environment, so that an internal and external integrated display effect is achieved.

The invention provides a terrain semantic perception method based on vision and vibration touch fusion. The terrain semantic perception method comprises the steps: firstly, giving an implementation method of vision three-dimensional semantic mapping based on ORB_SLAM2 and semantic segmentation; secondly, in combination with a terrain semantic classification and recognition method based on CNN-LSTM, giving a realization thought and a fusion strategy of vision / touch fusion; and finally, based on the blue whale XQ unmanned vehicle platform, the Kinect V1.0 visual sensing unit and the vibration sensing unit, carrying out algorithm testing in a real object environment. Therefore, the semantic marking precision of the method obtained by comparing a test result with a real environment can meet the application requirements; and meanwhile, whether the terrain semantic cognition is good or not can be obviously compared according to the fusion result of whether the vibration touch exists or not,so that more reliable sensing capacity can be provided for the patroller through fusion of the vibration touch and the terrain semantic cognition, and the vibration touch can still provide terrain cognition precision within a limited range even under the condition of visual failure.

The invention provides a meta-system design method based on a data model, relates to a meta-system design based on a data structure, and aims to simplify the processing logic of a large number of service table data in a service system, extract table information and field information by arranging the table information and the field information of a service table, and store the table information and the field information into a meta-system definition table. Meanwhile, a data element and a value domain code are extracted based on the service table and service field information, a data element catalogue and a value domain code catalogue are maintained in a meta-system, a reference relationship between the field information and the data element is established, semantic mapping is realized to ensure that each service field has a uniform format specification, and uniform accessing and extension processing are carried out on the service table data by the meta-system through these defined data.

The invention provides a trademark image consistent semantic extraction method and a trademark retrieval method. The trademark retrieval method comprises the steps of firstly performing similarity calculation on trademarks for trademark image retrieval, then adopting a similar image semantic extraction method based on shape representation and matching of the trademarks, and finally providing a trademark retrieval strategy based on calculated trademark image similarity class and semantic mapping relation. The trademark image consistent semantic extraction method and the trademark retrieval method can be applied after trademark images are segmented to obtain target objects, and accordingly accurate matching is performed on object level. In addition, by means of the trademark retrieval method, a user can further retrieve similar trademarks by using the trademark images or texts, and both the precision ratio and the recall ratio of retrieval results are improved.

The invention discloses a wheeled robot semantic mapping method and system fusing point cloud and images, solves the problem that a map constructed by laser SLAM only supports robot navigation, constructs a map embedded with semantic annotations, and belongs to the field of mobile robot SLAM. The main content of the method is that semantics are extracted from a two-dimensional point cloud, and semantics is stored, updated and optimized in a two-dimensional grid map. The method mainly comprises the steps of target detection based on deep learning on an image, point cloud segmentation, boundingbox and segmentation matching, map unit semantic updating and semantic optimization based on SLAM global optimization and clustering. The method has the advantages of being real-time, simple in device, rich in map information and the like, and aims at achieving intelligent navigation and man-machine interaction of the indoor mobile robot.

The invention discloses a semantic mapping and positioning method based on priori laser point cloud and depth map fusion. The method comprises the steps of S1, collecting priori laser point cloud data; S2, acquiring a depth image and an RGB image, generating RGB-D point cloud based on the depth image, and initializing and registering priori laser point cloud and RGB-D point cloud; S3, camera poseconstraints are provided by the registered prior laser point cloud to perform camera pose correction; S4, a three-dimensional geometric point cloud map is created by adopting a front and back window optimization method; S5, geometric increment segmentation is carried out on the three-dimensional geometric point cloud map, object recognition and semantic segmentation are carried out on the RGB image, geometric increment segmentation and semantic segmentation results are fused, and a 3D geometric segmentation map of semantic enhanced geometric segmentation is obtained; and S6, semantic association and segmentation probability allocation updating is carried out on the object to complete construction of a semantic map. Accumulated errors of large-scale indoor mapping and positioning can be effectively eliminated, and precision and real-time performance are high.

The invention discloses a semantic-based multi-modal spatio-temporal data association method, and belongs to the technical field of geographic space information systems. According to the method, the constructed ontology model is used for carrying out feature-semantic mapping and conceptualized semantic expression on the data, complex semantic information of the multi-modal spatio-temporal data iseffectively described in a unified format, and the problem that a computer is difficult to understand semantic content information of the data is solved; meanwhile, based on ontology semantic expression of the spatio-temporal data, time association, space association and content object semantic association between the spatio-temporal data are established through three factors of time, space and objects contained in data content, and association between the multi-modal spatio-temporal data is effectively carried out. The method is mainly used for cross-modal retrieval of multi-modal spatio-temporal data, an accurate and comprehensive data link association processing result is provided for a user, and then the cross-modal retrieval performance is improved.

The invention discloses a method for constructing a spatial semantic database based on BIM and GIS data integration. The method comprises the following steps: analyzing spatial data of BIM and GIS models; establishing mapping rules of object model related attribute description fields of IOT equipment and BIM and GIS model object attribute description fields, and constructing a relational database;extracting spatial semantic information, establishing a spatial semantic mapping relation table, and constructing a graph database; constructing a time sequence database according to the time sequence data transmitted by the IOT equipment; storing spatial semantic information in a graph database, storing relational data and static business data in a relational database, storing Internet-of-thingsdata in a time sequence database, and reading data interfaces of all the databases through a visualization engine. According to the method, a data storage standard of a BIM + GIS model is established, and fusion application is realized, so that a digital twin system application mode and a technical framework integrated with key technologies such as ontology, semantic mining, Internet-of-things, big data analysis and the like are constructed.

A system for associating a string of natural language with items in a relational database includes a first subsystem having a pre-trained first artificial neural network configured to apply a semantic tag selected from a predefined set of semantic labels to a segment of a plurality of tokens representing the string of natural language. A second subsystem includes a second artificial neural network configured to convert the plurality of labeled tokens into a first multi-dimensional vector representing the string of natural language. A third subsystem is configured to rank the first multi-dimensional vector against a second multi-dimensional vector representing a plurality of items in the relational database.

The invention discloses a coupled indoor three-dimensional semantic mapping and modeling method and a medium. The method comprises the steps of obtaining initial point cloud data, extracting feature points, estimating poses corresponding to the initial data frames and generating a local point cloud map according to the poses, generating a training data set, performing semantic annotation on the local point cloud map based on the deep neural network and the training data set, and feeding back a semantic annotation result of the local point cloud map to the initial data frame, optimizing the pose to obtain a first pose optimization result, extracting a semantic structure surface, and associating the semantic structure surface with the global plane, carrying out nonlinear optimization to obtain a second pose optimization result, and generating a final point cloud map, a semantic point cloud map and a building internal semantic line structure model. A semantic point cloud map and an internal wireframe structure model of an indoor environment can be accurately reconstructed; and meanwhile, the situation that pose estimation is lost due to violent movement of the acquisition platform during laser-based mobile scanning is prevented.

The invention discloses a vision-based automatic signboard semantic mapping positioning method and system. The method comprises the steps: acquiring continuous images and corresponding positioning information in the vehicle driving process; identifying signboard information in the continuous images; determining the positioning information of the signboard based on the signboard information in thecontinuous images and the positioning information corresponding to the continuous images; establishing a semantic map based on the signboards based on the signboard information in the continuous images and the positioning information of the signboards corresponding to the signboard information; and acquiring signboard information identified in the current image of the vehicle, and retrieving positioning information of the signboard in the semantic map based on the signboard to determine the current positioning information of the vehicle. According to the method and the system, the semantic mapis established according to the signboard based on a visual means, real-time positioning is provided for the automatic driving vehicle, and the method and the system are suitable for various indoor and outdoor scenes.

The invention belongs to the field of emotion recognition in artificial intelligence, and particularly relates to a music healing system based on brain wave emotion recognition and a processing method thereof. The system comprises an emotion recognition model based on brain waves, a music emotion classification model based on audio and an emotion semantic mapping relation based on music healing, and specifically comprises: a data acquisition and database management module which is used for acquiring electroencephalogram data and performing classification management to construct a database body; an emotion recognition module which is used for performing emotion recognition on the electroencephalogram signal data; the music classification module is used for carrying out music classification on the audio signal data; and a music healing module which is used for mining the association relationship between the emotion change and the music type of the user, constructing a complete emotion mapping relationship set and performing music healing on the user according to the complete emotion mapping relationship set. The electroencephalogram emotion recognition variety is rich, the music emotion category is wide, the emotion and the music type are combined, the music emotion label is given, the emotion recognition accuracy is high, and the music healing reliability is high.

The invention discloses a semantic mapping-based object classification method and system, and the method comprises the steps: obtaining a first image according to a first collection instruction, converting the format of the first image according to a first document conversion instruction, and obtaining a second file; performing document disassembly evaluation on the second file according to the first document disassembly evaluation instruction to obtain a first document disassembly scheme; obtaining a third file; performing keyword extraction on the third file according to the first keyword extraction instruction to obtain a first keyword extraction result; constructing a first mapping retrieval classification relationship between the keyword and the third file; obtaining a first semantic aggregation result; constructing a second mapping retrieval classification relationship between semantics and a third file according to the first semantic aggregation result to obtain a third mapping retrieval classification relationship; and according to the third mapping retrieval classification relationship, carrying out mapping relationship-based file classification on the first file. The technical problem that in the prior art, file classification and arrangement are not scientific and accurate enough during file classification and arrangement is solved.

The invention belongs to the technical field of image processing and computer vision, and discloses an RGBD multi-camera calibration method and system based on multi-dimensional semantic mapping and application, and the method comprises the steps: employing a hard synchronization triggering mechanism to synchronously obtain the color image data and depth image data of a plurality of RGBD cameras under the same clock; based on RGBD multi-camera initial calibration, shooting a two-dimensional code calibration plate image in a common-view area of all RGBD cameras, and obtaining initial poses among the RGBD cameras; then, associating the semantic information of a two-dimensional color image with three-dimensional depth space information to obtain a space point overlapping region between the adjacent cameras; and finally, obtaining accurate poses among the RGBD multiple cameras by adopting an ICP algorithm. The RGBD multi-camera calibration method has relatively high calibration precision,and can realize the accurate calibration of RGBD multiple cameras in a complex scene.

The invention discloses a zero-sample deep-sea biological picture classification method based on multiple classifiers. The method comprises the following steps that (1) image feature data importing, wherein a data set is composed of visual features, semantics and label information of visible categories and invisible categories, common marine organism categories with labels serve as the visible categories, and deep-sea organism categories lack of data serve as the invisible categories; (2) class specific classifier training: utilizing visible class data training to establish different visual semantic mappings for different classes on the basis of reserving a manifold structure so as to improve the classification accuracy of the visible classes; and (3) inferring the invisible category, migrating the weight between the visible category and the invisible category semantics, synthesizing a mapping matrix of the invisible category by using the mapping matrix of the visible category, and obtaining the label of the invisible category according to the distance. According to the method, the influence of mapping domain drift is reduced to a certain extent, and the method is accurate and feasible.

The invention discloses a semantic mapping method based on track alignment. The method comprises the following steps: S1, synchronously collecting RGB data, laser data, IMU data and odometer data along the same moving track; s2, through semantic ORB-SLAM, while ORB tracking is carried out by using RGB data, carrying out target detection on each frame of key frame image, constructing a triangulation relation, and extracting a 3D semantic landmark; S3, filtering the laser data by using downsampling operation, and performing calculation processing on IMU data and odometer data; s4, processing the laser data, the IMU data and the odometer data, and constructing a 2D grid map; and S5, solving a similarity transformation relationship between the ORB-SLAM and the track sequence by using the ORB-SLAM and the track sequence, and fusing the 3D semantic landmark and the 2D raster map to generate a semantic map. According to the scheme, the accurate grid map with the semantic landmark can be constructed.

The invention discloses a cross-modal data retrieval system based on text semantic mapping and a retrieval method thereof. The retrieval system comprises a data domain, an offline domain, a semantic domain and an online domain, wherein the data domain comprises a text data set and a non-text data set; the offline domain comprises a label extraction module, a text semantic extraction module, a datalabeling module and a non-text semantic model training module; the semantic domain comprises text semantic space; and the online domain comprises a text semantic extraction module, a non-text semantic extraction module, a query expression module, a relevancy calculation module and a relevancy sorting module. According to the method, the non-text data semantics are mapped to the text semantic space, so that the efficiency and operability of system data labeling, model training and retrieval accuracy evaluation are improved, the non-text data semantic mapping frequency is reduced, the originalsemantics of the data are greatly reserved, and the data retrieval accuracy can be effectively improved.

The invention discloses a method for constructing a building information model, a terminal and a storage medium, and the method comprises the steps: obtaining indoor three-dimensional point cloud data, and obtaining geographic space data according to the three-dimensional point cloud data; according to a semantic mapping method, obtaining a target modeling component corresponding to an indoor component in the geographic space data; and obtaining a building information model according to the parameters of the indoor components in the geographic space data. Therefore, automatic reconstruction ofthe building information model is realized, and the labor cost is saved.

The invention relates to a semantic mapping-based combination simulation method and device, equipment and a medium, and the method comprises the steps: obtaining a combination simulation task, wherein the combination simulation task comprises a plurality of simulation events, and each simulation event is formed by connecting a plurality of simulation parts; converting the simulation parts into atomic models by defining internal execution logic and external interface logic of the simulation parts, and converting the combination simulation task into a semantic layer combination simulation model; according to the combination simulation task, obtaining a pre-stored grammar layer combination simulation model; and according to the semantic layer combination simulation model and the grammar layer combination simulation model, obtaining a mapping relation of conversion from the semantic layer to the grammar layer, establishing an executable layer combination simulation model corresponding to each simulation event, and inputting the executable layer combination simulation models into a simulation system for combination simulation. By adopting the method, the relation between simulation system construction and simulation context constraint can be established, the modeling process is simple and rapid, and modeling results are convenient to combine.

The invention provides an artificial intelligence auxiliary innovation method and device, and a computer readable storage medium. The method comprises the steps: obtaining a semantic ontology, reconstructing an invention information base, cooperatively labeling a reference text, and realizing semantic mapping of invention information and the reference text; and performing comprehensive retrieval sorting by utilizing a semantic ontology concept and an instance extension retrieval formula and combining a scene and a concept allocation weight, and screening a text to generate an auxiliary innovation scheme. According to the method, a semantic bridge between the known invention information base and the reference text is effectively established, and a feasible scheme is provided for precision invention innovation, technical transformation and auxiliary industry upgrading.

The invention provides a knowledge graph semantic mapping construction method and system, and the method comprises a training set management step: managing a text training set and a test set, and iterating a Boolean model; a text analysis step: performing word segmentation processing on the input text, and converting the text into a combination with a syntactic structure; and a Boolean logic expression step: calculating the mapping between the text and the knowledge graph entity relationship according to the obtained converted combination with the syntactic structure. More syntactic structuresand matching relations related to text semantics are obtained through Boolean logic based on syntactic structure analysis, and the Boolean logic constructed through the relations can describe matching and modifying structures between words more accurately; therefore, the matching relationship between the words of the syntactic structure can be intervened transparently and manually.

A method for performing tasks on items located in a space using a robot, the items being located proximate fiducial markers, each fiducial marker having a fiducial identification. The method includesreceiving an order to perform a task on at least one item and determining the fiducial identification associated with the at least one item. The method also includes obtaining, using the fiducial identification of the at least one item, a set of coordinates representing a position of the fiducial marker with the determined fiducial identification, in a coordinate system defined by the space. The method further includes navigating the robot to the coordinates of the fiducial marker associated with said determined fiducial identification.

The invention relates to a face image cartoonalization processing method and device, computer equipment and a storage medium. The method comprises the following steps: extracting face feature points and image features from a real face image; constructing a three-dimensional simulation face model corresponding to the real face image based on the face feature points; obtaining difference characteristics between the simulation face model and a three-dimensional simulation face template; based on a semantic mapping relationship between the three-dimensional simulation face template and a three-dimensional cartoon face template, migrating the difference features to the three-dimensional cartoon face template to obtain a three-dimensional cartoon face model; and performing image rendering on the cartoon face model according to the image features to generate a three-dimensional cartoon image. By adopting the method, the three-dimensional cartoon image which is relatively high in accuracy and has real face features and image features can be effectively constructed.

The invention discloses a heterogeneous data pattern matching method based on an attention mechanism, and the method comprises the following steps: inputting source data and to-be-matched data, wherein the to-be-matched data is heterogeneous data; classifying the source data and the to-be-matched data, wherein the source data and the to-be-matched data comprise numeric data, character data and rare type data; performing feature extraction on the classified data according to a newly established data index system; inputting features extracted from source data into a pattern matching model based on an attention mechanism for training, inputting features of to-be-matched data into the trained model for similarity calculation of attributes among heterogeneous data, and establishing a semantic mapping relationship between elements in a source pattern and a to-be-matched pattern according to a bidirectional filtering method, namely pattern matching. According to the method, the complexity of the neural network is reduced based on the attention mechanism algorithm, interference data is further filtered, and the efficiency and accuracy of heterogeneous data pattern matching are improved.

The invention discloses a heterogeneous multi-source data association analysis system and method. The system comprises a heterogeneous multi-source data acquisition module, a heterogeneous multi-source data hybrid association module, a heterogeneous multi-source data semantic mapping module, a heterogeneous multi-source data fusion module and a heterogeneous multi-source data cleaning module. The heterogeneous multi-source data acquisition module is used for acquiring data information of different sources and transmitting the acquired data; the heterogeneous multi-source data hybrid association module provides a way for knowledge of the infrastructure and describes a space-time evolution process in a phenomenon occurrence process of the infrastructure; the heterogeneous multi-source data semantic mapping module establishes a relationship between low-level data and a corresponding label, so that conversion between bottom-level features and high-level semantics is realized; the heterogeneous multi-source data fusion module is used for judging a semantic distance between the data so as to perform fusion; and the heterogeneous multi-source data cleaning module is used for cleaning the data to ensure the validity and authenticity of the data.

The invention provides a semantic integration method for a multi-source heterogeneous database. The semantic integration method comprises the following steps: (1) entity extraction: extracting domain-related entities from an unstructured text based on an entity extraction model and identifying corresponding categories; (2) concept matching: performing matching with ontology concepts in the knowledge graph according to the corresponding categories to obtain candidate entity sets of the same category; (3) neighborhood matching: obtaining an aligned entity graph representation according to the context information of the related entities, and obtaining a candidate entity graph representation according to the domain relationship of the candidate entity set in the knowledge graph; and (4) comparison decision making: performing comparison decision making on the aligned entity graph representation and the candidate entity graph representation to obtain the most matched candidate entity arrangement as a matching result. According to the method, the deep reinforcement learning technology and multi-source heterogeneous database semantic integration are combined, the semantic mapping relationbetween knowledge in different forms is established, and semantic integration-based semantic retrieval, intelligent question and answer and other related applications can be better supported.

The invention discloses a visual question answering method and device based on question semantic mapping. The method comprises the steps of extracting visual features of images in a training set and question features in questions; carrying out feature fusion on the extracted visual features and problem features; classifying questions in the training set, and counting answer ranges of different question types in the training set; extracting answer features of answers in the answer range; establishing a mapping relation between the question features and the answer features, and obtaining answer range probability distribution; carrying out reasoning according to fusion features after feature fusion and answer range probability distribution, and obtaining a final answer. The method comprises the following steps: extracting question features through a gating loop unit, extracting consistent semantics of a question and an answer by utilizing embedded learning, mapping the semantics to the answer, and extracting high-level semantic information of an image through a question-oriented regional attention mechanism and a relationship; and fusing the image high-level semantic information with the answer range mapping result to generate a final answer.