24 results about "Point graph" patented technology

The invention discloses a Gaussian representation SLAM method based on dense matching priori and factor graph constraint, which comprises the steps of inputting a current image and a key frame image, outputting a point graph corresponding to the image through a pre-trained model, returning a matching condition of two frame image points and respective point cloud information, and obtaining a point-level matching result based on a point-level matching result. The method comprises the following steps: constructing a joint optimization problem of a current frame and a key frame by taking a luminosity consistency error and a geometric projection error as targets, performing joint estimation on a camera pose and a point cloud of the current frame, realizing high-precision pose solution, generating point diagram data after Gaussian scene representation and rasterized rendering processing, and transmitting the point diagram data to a rear end for global optimization. And the rear end receives the pose and point cloud data, executes loopback detection to identify repeated key frames, and performs Gaussian rendering through an optimized key frame image to complete global dense three-dimensional reconstruction. The method effectively solves the problem of track drift and scene inconsistency caused by lack of pose priori and global geometric constraints in an existing system.

The invention discloses a homework semantic evaluation method based on a multi-modal large model, and relates to the technical field of education intelligent evaluation, and the method comprises the steps: 1, normalizing multi-modal answers into a step sequence, a scoring key point map, an acceptable answer set and an initial scoring anchor point, and building a step and key point bidirectional index; 2, performing consultation according to teacher roles, generating candidate scores and key credits, and calculating a key point coverage degree, a consistency degree and a contradictory proportion; and step 3, reading the conditional reliability ledger, generating a comprehensive weight, aggregating the comprehensive weight to obtain a final score, and outputting an uncertainty index at the same time. 4, giving out hierarchical personalized explanation, determining a minimum correctable step, executing step-level echoing, and generating a training list; 5, carrying out anchor question online calibration and fusion parameter updating, and setting a trigger proportion upper limit and an early stop threshold according to a budget perception route; the link realizes unified measuring scale, steady decision, traceable learning and controllable cost.

The invention discloses a test paper generation method and system, and belongs to the technical field of intelligent education, and the method comprises the steps: obtaining student answer records, question-knowledge point mapping, a test question bank and knowledge text information, and constructing a knowledge point graph; constructing a student-question-knowledge three-layer heterogeneous graph, and predicting the answering performance of the students on the test paper; taking the answer expression as environment feedback input of reinforcement learning, taking a question set of the current test paper as a state, replacing questions in the test paper or keeping the questions as an action, determining a comprehensive reward value of the test paper in each dimension by constructing a multi-target reward function, and generating an optimal test paper through iterative optimization by taking maximization of the comprehensive reward value as a target; on the basis of the adjacency relation in the knowledge point graph, candidate test questions are selected from a test question bank to form a candidate sub-question set, and multiple sets of parallel test papers equivalent to the optimal test paper are generated through an equivalent question replacement strategy. Through the method, more efficient, reasonable and fair parallel test paper generation can be realized.

The invention provides an attack detection and source tracing method and system for a multi-mode AI system, and relates to the technical field of artificial intelligence. The method comprises the steps that input data, model internal states, system logs and behavior data are collected, and time mark high-frequency records are unified to form a multi-modal data set; encoding each time window graph snapshot of the dynamic graph by applying a neural network, and generating a graph overall embedded vector and a node embedded vector to capture spatial structure and time sequence dependence; acquiring a graph embedding representation of a current time point according to the graph overall embedding vector and the node embedding vector, executing attack detection based on the graph embedding representation, and triggering an alarm when an anomaly is detected; attack traceability is carried out in an abnormal time point diagram snapshot, a node-level abnormal score is calculated through node embedding, and key nodes, edges and sub-graphs are identified by using attention weight of a neural network and a graph interpreter so as to locate attack entry points, propagation paths and influence ranges.

Embodiments of the invention provide a map data processing method and apparatus, an electronic device and a storage medium. The processing method comprises the steps of obtaining a first pattern spot and first attribute data corresponding to the first pattern spot; according to the first attribute data, combining second pattern spots with adjacent positions and the same land utilization type in the first pattern spots, and combining second attribute data corresponding to the second pattern spots in the first attribute data to obtain third pattern spots and target attribute data; in the third pattern spots, eliminating non-sporadic ground feature pattern spots smaller than a preset area, and extracting long and narrow pattern spots and sporadic ground feature pattern spots to obtain target pattern spots; the long and narrow pattern spots are converted into target linear patterns, and the sporadic ground feature pattern spots are converted into target point-shaped patterns; and generating data of a target map according to the target pattern spot, the target linear graph, the target point graph and the target attribute data. According to the embodiment of the invention, the display effect after the map is shrunk can be effectively ensured.

The application discloses a kind of Gauss representation SLAM methods based on dense matching prior and factor graph constraint, including input current image and key frame image, pre-trained model, output the point graph corresponding to image pair, and return the matching situation of two frame image points and respective point cloud information, based on point level matching result, with photometric consistency error and geometric projection error as target, construct the joint optimization problem of current frame and key frame, jointly estimate the camera pose and point cloud of current frame, realize high-precision pose solution, after Gauss scene representation and rasterization rendering processing, generate point graph data, and pass to back end and carry out global optimization, back end receives pose and point cloud data, executes loop detection to identify repeated key frame, Gauss rendering is carried out through optimized key frame graph, complete global dense three-dimensional reconstruction, the present application effectively solves the trajectory drift and scene inconsistency problem caused by lack of pose prior and global geometric constraint in existing system.

The invention discloses a streaming three-dimensional scene reconstruction method based on explicit spatial point memory, and aims to solve the problems of memory redundancy and information loss existing in an implicit memory method in existing streaming three-dimensional reconstruction. According to the method, firstly, explicit spatial point memory is initialized through feature coding of a first frame of image, and scene features and explicit spatial positions are directly associated through the memory; each time a frame of real-time RGB image is received, image features are firstly extracted and efficiently interacted with stored spatial point memory features, and three-dimensional prior is provided by combining three-dimensional hierarchical position embedding to obtain feature representation under a global coordinate system; decoding point diagrams and camera poses in the global coordinate system, and encoding to generate a new spatial point memory unit; and finally, updating the global explicit spatial point memory through a memory fusion module. According to the method, the non-registration pose streaming visual input is unified to the global coordinate system online, the pixel-by-pixel alignment point diagram is output, and the reconstruction efficiency, uniformity and flexibility are improved.

The application provides a kind of city multi-source spatial data fusion method and system based on graph comparison, it is related to data fusion technical field, wherein the method comprises the following steps: S1: obtaining the city multi-source spatial data of the region to be processed, and the city multi-source spatial data is preprocessed;S2: the city multi-source spatial data after preprocessing is extracted, and the corresponding point graph data set, surface graph data set and line graph data set are obtained;S3: point graph data set, surface graph data set, line graph data set are fused and conflict is judged, and the corresponding fusion result is generated;The application realizes the high-quality fusion of multi-source data, provides reliable data support for city digital construction and intelligent management, and has significant technical advantages and practical application value.

The invention discloses a large language model adaptive content generation method based on a multilayer knowledge graph and preference perception learning. The method comprises the steps of constructing a hierarchical data set according to questions q and answers in an encyclopedia book, constructing a training data set according to a preset selection instruction p and the hierarchical data set, constructing an encyclopedia knowledge natural language model, training the encyclopedia knowledge natural language model according to the training data set to obtain a trained encyclopedia knowledge natural language model, and obtaining a knowledge point graph G according to the question and answer data set E, and obtaining a reasoning answer according to the knowledge point graph G, the cognitive level of the questioner, the trained encyclopedic knowledge natural language model and the question q'of the questioner. According to the method, the hierarchical data set and the multi-layer knowledge graph with the aligned cognitive levels are constructed, so that the requirements of learners with different cognitive levels are effectively met, and the problem of cognitive dislocation of a current large language model in education application is solved.

The application discloses a tunnel point cloud multi-target segmentation integrated method and system based on causal reasoning, and the method comprises the following steps: collecting point cloud data of a tunnel and converting the point cloud data into a two-dimensional point graph based on a circle projection algorithm; extracting global features of the point cloud data based on the two-dimensional point graph and a two-dimensional Unet model, and extracting local features of the point cloud data based on the two-dimensional point graph and edge convolution; performing data enhancement on the point cloud data of the tunnel based on a causal reasoning model; respectively classifying the point cloud data of the tunnel after data enhancement according to the global features and the local features; and performing feature fusion on the classification results based on an improved D-S evidence-based feature fusion method. In view of the effective and rapid solution to the three-dimensional point cloud semantic segmentation problem, the application provides an integrated method based on point projection and a dynamic graph convolutional neural network, processes input point cloud, and performs good semantic reasoning, so that seepage categories can be identified and determined at high precision and high speed.

This invention discloses an interest point recommendation method based on user multi-behavior enhancement and efficient rich-information negative sampling, comprising: acquiring and preprocessing user multi-behavior data; constructing three types of context graphs using the user multi-behavior data; learning multi-view representations of users and interest points using a Skip-Gram model; constructing a pairwise dataset, calculating the probability score of each user-interest point check-in pair using a pairwise ranking model, and calculating the BPR loss function; replacing the negative sampling part in the above two models with a negative sampling method based on a baseline point graph; designing a joint learning framework that combines the loss of the Skip-Gram model with the loss function of BPR; calculating the probability score of a user checking in at any interest point in the future using a pairwise ranking model, and recommending interest points to the user. This invention's method can recommend interest points to users more accurately, and its graph construction method can be flexibly applied to different recommendation scenarios.

Embodiments of the present application disclose a multi-source vector map fusion method and device. The method of the embodiments of the present application is to obtain a global road skeleton, a historical vehicle trajectory set and original vector maps from multiple data sources, construct a point graph and a line graph of each original vector map respectively, extract vector element features, element spatial features and local topological features from the point graph and the line graph, fuse the vector element features, the element spatial features and the local topological features to obtain a vector feature vector, extract global topological features from the global road skeleton, extract trajectory time sequence features from the historical vehicle trajectory set, match the vector feature vector with the global topological features and the trajectory time sequence features based on a preset feature matching algorithm, and fuse multiple original vector maps with a matching rate higher than a preset matching rate to obtain a target vector map. The method additionally analyzes the global road skeleton and the historical vehicle trajectory set to participate in the fusion of the original vector maps, thereby ensuring topological accuracy and consistency.

The application provides a retrieval enhancement generation method based on a document knowledge base and a knowledge graph, and can be applied to the technical field of artificial intelligence. The method comprises the following steps: in response to receiving target question text, determining enhancement data for the target question text by using a prior global knowledge graph and a prior document knowledge graph; inputting the target question text and the enhancement data into a pre-trained language model to generate target reply text corresponding to the target question text, wherein the prior global knowledge graph and the prior document knowledge graph are obtained by the following operation: aligning M single-document knowledge graphs based on M documents according to a prior knowledge point graph to obtain the prior global knowledge graph; constructing a document knowledge graph representing the association relationship among the M documents based on the prior global knowledge graph; and aligning the document knowledge graph according to the prior knowledge point graph to obtain the prior document knowledge graph.

A visual language navigation method and system based on three-dimensional geometric perception and shape level representation, the method comprising: constructing a Floor-to-Floor multi-layer navigation data set containing viewpoint level three-dimensional observation data and shape perception instructions; the viewpoint level three-dimensional observation data comprises multi-view RGB images, point graphs aligned with the multi-view RGB images pixel by pixel, point graph confidence and relative orientation encoding; through geometric block extraction and mask optimal transmission, the RGB images and the point graphs are aligned in cross-modal geometric perception features to generate geometric enhanced visual features; through a geometric-language alignment module based on a cross-attention mechanism, the geometric enhanced visual features and the shape perception instructions are fused to generate multi-modal features; an incremental topological graph is dynamically constructed and maintained during navigation, and global action decision is made based on the multi-modal features and the incremental topological graph; the navigation strategy is jointly optimized and trained through behavior cloning and data aggregation strategy.

The application relates to the technical field of online learning, and particularly relates to a learning path generation method, device and equipment for senior members, a medium and a program product. The method comprises the following steps: firstly, in response to a question input by a senior member, extracting a question feature according to a corresponding subject type of the question; secondly, inputting the feature into a knowledge point matching model, obtaining a core knowledge point by combining a preset knowledge point description-feature keyword mapping library; thirdly, obtaining a mastery degree rating by a multi-dimensional weighting algorithm according to historical learning data of the senior member on the core knowledge point; fourthly, screening similar knowledge points from a preset knowledge point graph based on the core knowledge point, and obtaining an initial screening question list by combining a knowledge point ID-question ID list mapping relationship; and finally, secondarily screening the list and preferentially returning a question corresponding to a weak knowledge point, so that a precise and personalized learning path can be generated, and the learning efficiency of the senior member is improved.

The application discloses a test paper generation method and system, belongs to the intelligent education technical field, acquires student answering record, question-knowledge point mapping, test question bank and knowledge text information, and constructs a knowledge point graph; a student-question-knowledge three-layer heterogeneous graph is constructed, and student answering performance on a test paper is predicted; the answering performance is taken as environmental feedback input of reinforcement learning, a question set of the current test paper is taken as a state, questions in the test paper are replaced or reserved as actions, a multi-target reward function is constructed to determine a comprehensive reward value of the test paper in each dimension, and the optimal test paper is generated by iteration optimization with the maximum comprehensive reward value as a target; based on the adjacent relationship in the knowledge point graph, candidate subtest sets are formed by selecting candidate test questions from the test question bank, and multiple sets of parallel test papers equivalent to the optimal test paper are generated by using an equivalent question replacement strategy. Higher efficient, reasonable and fair parallel test paper generation can be realized by the method.

The embodiment of the invention provides a line topology identification method, electronic equipment, a storage medium and a program product. The method comprises the following steps: intercepting a to-be-processed image of a tower graph model based on a two-dimensional circuit diagram of a single circuit in a power grid; based on the symmetric state and the side view of the to-be-processed image, determining a starting and ending point graph model and a target line endpoint which are erected on the same pole in the pole tower graph model; determining a target endpoint name of the target line endpoint based on the target line endpoint of the starting and ending point graph model and the line name of the target line endpoint; based on target line endpoints and target endpoint names of the starting and ending point graph models in the plurality of single lines, constructing an association relationship between lines erected on the same pole; and in response to a topology search request, carrying out search based on the two-dimensional line diagrams of the plurality of single lines and the association relationship to obtain a line topology comprising the same-pole erected lines. The method is used for achieving the effects of obtaining the line topology containing the same-pole erected line through two-dimensional line diagram recognition and improving the recognition efficiency.

A model-free object grasping pose generation method for non-structured operation scenarios, after the collected point cloud data is generated into an original scene graph based on a neighborhood node algorithm, the original scene graph is input into a graph feature extraction network to obtain a graph embedding feature fused with multi-scale information, the graph embedding feature is sequentially input into an object surface point screening network and a high-value point screening network and a high-value point graph is obtained, the high-value point graph is input into a pose generation network to obtain a spatial six-degree-of-freedom grasping candidate pose, and then the best grasping is selected for implementation. The present application solves the problems of strong randomness of point selection, limited generalization performance, insufficient spatial feature extraction capability of the network model, and low grasping quality of the generated grasping in the prior art.

The application discloses a method for planning a flight path of a UAV and related devices, comprising: determining grid nodes of a monitoring area and edges connecting adjacent nodes; extracting a monitoring point graph from the grid structure according to monitoring points and their adjacency relations; judging whether the monitoring point graph belongs to a first preset graph set, which corresponds to monitoring point graphs satisfying a right-angle Steiner minimum tree construction condition; if yes, obtaining a target connected monitoring point graph by using a first connected graph construction algorithm based on a right-angle Steiner minimum tree problem, otherwise, using a second connected graph construction algorithm based on a minimum spanning tree; then judging whether the target connected monitoring point graph belongs to a second preset graph set, which corresponds to connected graphs capable of determining a Hamilton path within a polynomial time; if yes, planning a flight path by using a first path planning algorithm based on a Hamilton path, otherwise, using a second path planning algorithm based on tree search. Thus, the timeliness of information collection can be ensured while the path planning efficiency is improved.

The invention discloses a circuit maintenance learning and drawing fusion interaction method, electronic equipment and a storage medium. The circuit maintenance learning and drawing fusion interaction method comprises the following steps: acquiring and analyzing an electronic design automation source file; extracting a netlist from the analyzed electronic design automation source file, and establishing a flattened data structure by taking the netlist as a hub; extracting schematic diagram information from the analyzed electronic design automation source file, and converting the schematic diagram information into a group of lightweight vector graphic data; and carrying out deep binding on the maintenance and learning scene data with the flattened data structure and the lightweight vector graphic data through a network list to form a maintenance learning knowledge graph. Through data reconstruction of an electronic design automation source file, a point location diagram and a schematic diagram of an electronic circuit are subjected to native fusion from a bottom layer, maintenance and learning data are deeply bound with a drawing, and maintenance and learning personnel provide integrated immersive drawing analysis experience.

The application discloses a tool wear state monitoring method and system based on visual features, and relates to the technical field of tool wear monitoring.The method comprises the following steps: acquiring a one-dimensional time domain vibration signal of a tool to be measured in a machining process; converting the one-dimensional time domain vibration signal into a two-dimensional vibration signal-symmetrical point graph based on a symmetrical point graph technology; adaptively extracting a deep visual feature map of the two-dimensional vibration signal-symmetrical point graph by using a wavelet scattering convolution network; and classifying by using a deep convolution neural network model based on the deep visual feature map, and outputting a wear state of the tool to be measured.The application realizes accurate monitoring of the tool wear state, avoids deviation caused by manual screening of features, has high adaptability to multiple working conditions, and has a wide application range.

The application discloses a face changing video tampering detection method and device based on face key point graph modeling, wherein the method comprises the following steps: acquiring face changing video data, acquiring a training set and a test set according to the face changing video data; constructing a tampering detection model based on face key point graph modeling; training the tampering detection model by using the training set; testing the trained tampering detection model by using the test set; wherein the tampering detection model comprises a front-end face key point extraction module, a key point topological graph feature generation module, a parallel graph attention neural network module, a rear-end recurrent neural network module and a decision output module. The application performs topological graph modeling on face key points, uses a graph attention neural network to extract high-dimensional geometric features of key point topological graph nodes, further enriches the feature representation of a single frame face, and improves the robustness and detection accuracy of the model. The application can be widely applied to the technical field of digital video tampering detection.

The application discloses a sparse pose-free feedforward novel view synthesis system and method based on geometric conduction. The system comprises a geometric perception feature extraction module, a geometric-to-Gaussian conduction module, a geometric consistent camera adaptation module and a trajectory condition confidence rendering module. The geometric perception feature extraction module extracts a dense feature map, a normalized three-dimensional point graph and a cross-view pixel trajectory from a sparse input image by using a pre-trained visual geometry foundation model; the geometric-to-Gaussian conduction module converts the geometric perception feature into a three-dimensional Gaussian primitive set by adopting a slot attention mechanism; the geometric consistent camera adaptation module aligns the camera parameters with the Gaussian rendering space by global similarity alignment and differentiable geometric proxy rendering; and the trajectory condition confidence rendering module constructs a prior graph by using the cross-view trajectory and predicts a confidence score of each Gaussian and each view to modulate the Gaussian splatting weight. The application realizes fast and generalizable novel view synthesis under sparse and camera pose-free input, and improves the rendering quality and geometric consistency.

This invention relates to computer vision and point cloud processing technology, specifically a scalable 3D panoramic segmentation method and system based on reinforcement learning-driven clustering. The method includes the following steps: dividing the input point cloud data into semantic voxels to construct an initial super-point graph; constructing a high-level meta-controller and a low-level controller; modeling panoramic segmentation as a sequential graph clustering problem and iteratively processing the initial super-point graph; evaluating the current super-point subgraph using the high-level meta-controller, extracting global feature embeddings, and outputting the probability of whether to terminate segmentation or continue splitting the current super-point subgraph; if the decision is to terminate segmentation, using the current super-point subgraph as the final instance; if the decision is to continue splitting, traversing the edges in the super-point subgraph to be split using the low-level controller and updating the subgraph structure; and employing a near-end policy optimization algorithm with a hybrid reward function for optimization. This invention reconstructs graph clustering into a multi-step sequential decision-making process, effectively solving the problems of under-segmentation and boundary ambiguity in large-scale point clouds.