51 results about "Spatial localization" patented technology

This invention discloses a small target tracking method based on image enhancement and attention mechanisms, comprising: performing adaptive histogram equalization on shallow feature maps to enhance local contrast, and fusing with deep semantic features to construct a multi-scale feature pyramid; generating attention weight maps in the horizontal and vertical directions through a coordinate-decoupled attention module to highlight the salient regions of small targets; introducing a dynamic anchor box generation algorithm based on Gaussian distribution, dynamically sampling and generating anchor boxes based on a Gaussian distribution model constructed from the target position in the previous frame; and using the deep cross-correlation layer of a Siamese network to calculate the similarity response map between the template and the search region to determine the center coordinates and bounding box of the small target. This invention effectively solves the problems of insufficient feature extraction, low spatial positioning accuracy, and inflexible anchor box generation in existing methods, significantly improving the accuracy and robustness of small target tracking.

This invention provides a method and apparatus for component detection and localization based on multimodal data. In an augmented reality-based manufacturing conformity inspection system, this method uses a multimodal neural network algorithm that fuses RGB and depth images. Features are extracted from and fused from both the RGB and depth images respectively. Finally, the two-dimensional detection results and depth information are used to perform three-dimensional spatial localization of the components. This method improves the robustness of the component detection algorithm in complex environments such as insufficient lighting, and enhances the model's accuracy, recall, and mean precision. Furthermore, this method can perform three-dimensional spatial localization of components based on the center points of the 2D bounding boxes output after detection and the aligned depth image.

This invention relates to the field of power grid operation and maintenance technology, and in particular to a method and system for intelligent power grid defect identification and inspection based on unmanned aerial vehicles (UAVs). The method includes the following steps: using UAVs to complete 3D modeling and synchronous acquisition and precise registration of multi-source data, outputting the registered multi-source raw data; performing hierarchical image preprocessing and feature-level fusion to output a multi-source fused image with 3D spatial coordinates; based on the multi-source fused image, performing real-time end-side inference and identification using a preset lightweight defect identification model, outputting initial defect identification results; based on the initial defect identification results, combining spatial positioning data to complete precise 3D location and level classification of defects and early warning push, outputting leveled defect location early warning data. This invention relies on UAVs to achieve multi-source data processing, intelligent end-side defect identification and precise early warning for power grid inspection, constructing an inspection closed loop in the cloud, solving multiple technical problems, and significantly improving the level of intelligent inspection.

This invention discloses a non-cooperative signal localization method based on radiation source decoupling, belonging to the field of wireless signal processing and spatial positioning. The method includes: transforming multi-source joint localization into multiple independent single-source localization problems by constructing a source-level decoupling theoretical framework. First, the method utilizes multiple receiving nodes to acquire mixed signals; then, it uses a mask-based separation network (MS-Net) to perform blind source separation on the mixed signals, obtaining the decoupled components of each radiation source. This network does not require prior information such as pilot signals; finally, for each decoupled component, a broadband Doppler model considering time-scale effects is used for direct single-source localization. This invention significantly reduces computational complexity, improves model adaptability and localization accuracy in high-speed motion scenarios, and has good engineering application prospects.

This invention relates to the field of medical control technology, specifically to a glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs. It includes a brain imaging data acquisition module for preoperative acquisition of multimodal brain imaging data and an acquisition module for acquiring intraoperative craniotomy photographs for ECoG electrode placement. The system further aligns the multimodal brain imaging data to the same three-dimensional spatial coordinate system; then segments and corrects the multimodal brain imaging data into different regions; and establishes an individualized brain region atlas based on a healthy brain tissue texture model and segmented brain regions. Vascular structures are extracted, and a three-dimensional spatial coordinate system for vascular feature points and electrode contacts is established and transformed. The minimum Euclidean distance is calculated based on the electrode contacts, and the individualized brain region atlas is labeled. This invention addresses the problems of insufficient spatial positioning accuracy of electrode contacts and difficulty in quantifying tumor-electrode distance during surgery for glioma patients, facilitating intraoperative determination of the anatomical location of the epileptogenic focus by surgeons.

This application relates to an AR glasses collaborative system combining visual large-space positioning. The system, applied in a large-scene space, includes: AR glasses, a visual positioning server, and a coordinate synchronization server. The visual positioning server is used to obtain the second pose of the AR glasses in the map coordinate system by matching real-time images in a point cloud map and returning it; and to calculate the first relative pose of the AR glasses' local coordinate system relative to the map coordinate system based on the first and second poses. The coordinate synchronization server is used to obtain the second relative pose between different AR glasses based on the second poses of multiple AR glasses and return it to the corresponding AR glasses. The AR glasses are also used to locally display specific AR content associated with the real scene in the target scene based on the first relative pose, and to collaboratively display specific AR content in the target scene based on the second relative pose.

This invention discloses a method and system for suppressing false rejection behavior and selectively forgetting harmful semantics in large language models across vertical domains based on feature subspace decoupling, belonging to the field of fine-tuning and alignment technology for large artificial intelligence models. The method includes steps for target definition and data construction, feature subspace localization, key level optimization, orthogonal decoupling weight correction, constrained norm renormalization, selective fine-tuning, false rejection calibration, and closed-loop verification monitoring and document output. The system includes modules for concept discovery and data construction, representation subspace modeling, false rejection localization and calibration, and verifiable evaluation and continuous monitoring. This invention can significantly reduce the false rejection rate and maintain a safety baseline while preserving the model's generalizability, meeting the compliance requirements of generative artificial intelligence services.

This invention relates to the field of injection mold inspection technology, specifically a method and system for defect detection in injection molds based on visual recognition. The method includes the following steps: using an industrial camera, acquiring multi-exposure images of the injection mold cavity surface; combining the grayscale undulation trajectory of the seamless area on the parting line; identifying the set of turning and stopping distributions; and based on the grayscale direction of pixels in the core area and neighboring pixels, completing the grayscale reconstruction of the image structure and outputting the spatial localization result of the defect. This invention sequentially correlates the grayscale changes of the same pixel under different exposure states to construct a continuous change trajectory to characterize the pixel response features. It combines the turning positions and stopping sections of grayscale changes to form a discrimination criterion, introduces the consistency screening of neighboring pixels with the same exposure position change direction, completes spatial structure merging, and reorganizes the grayscale distribution based on the correspondence between grayscale trajectories inside and outside the area, so that the defect area forms a continuous structural expression, enhancing the recognition consistency of dark texture defects in complex texture environments.

According to an aspect of the present inventive concept there is provided an ingestible device for analysis of the gastrointestinal tract, the ingestible device comprising: at least one sensor configured for sensing a biochemical parameter, wherein the at least one sensor is configured to generate biochemical parameter data for identifying an anatomical localization of the ingestible device in the gastrointestinal tract, and a magnetic field sensor configured to receive an alternating magnetic field for generating spatial location data representing a spatial localization of the ingestible device, and wherein the ingestible device is configured to provide a combined localization of the ingestible device in the gastrointestinal tract based on the anatomical localization and the spatial localization.

This invention proposes a neural network model enhancement and electronic device for small target defect detection in oilfield inspection images under complex field scenarios. It addresses the problems of severe feature degradation, scarcity of real samples, poor spatial localization ability, low anchor box matching degree, and limited computing power at the edge in small target defect detection in oilfield inspection images under complex field scenarios. This invention constructs synthetic and real datasets through dual-source data processing; performs four-fold super-resolution amplification of the input data through residual dense block concatenation and sub-pixel convolution, and introduces a generative adversarial network detail enhancement branch to generate super-resolution reconstructed images; extracts multi-scale basic features through a lightweight backbone; generates coordinate attention weights through global pooling of height and width dimensions to enhance the spatial features of small targets; inputs the enhanced feature map into a defect detection layer embedded with anchor boxes specific to small targets for classification and regression; and optimizes model training using a hard sample mining loss function.

This invention belongs to the field of spatial location technology for geological defects, and provides a method and system for spatially locating water-conducting fracture zones in coal mines. The method involves establishing a three-dimensional prior geological model of the target area in the coal mine; setting at least one tracer injection point and multiple receiving points in the target area; injecting fluorescent nano-tracers; acquiring tracer concentration-time response data at each receiving point and extracting characteristic parameters; constructing a three-dimensional forward model based on the three-dimensional prior geological model and the extracted characteristic parameters, which includes a groundwater flow model and a tracer transport model; comparing the measured response with the forward model output, constructing a joint objective function, iteratively inverting the parameters to be estimated, obtaining the optimal parameter combination, and obtaining the three-dimensional spatial location, development height, extension direction, connectivity direction, influence range, and uncertainty results of the water-conducting fracture zone. This invention achieves water inrush source identification, water-conducting fracture zone location, and risk quantification assessment.

This invention discloses a method, system, and electronic device for intelligent recognition and digital modeling of substation electrical drawings, relating to the field of image processing technology. The method includes: electrical drawing input and multimodal semantic parsing; spatial localization of text and graphic elements based on small models; semantic result association based on text similarity; text-graphic element binding incorporating spatial distance and semantic constraints; inclusion relationship parsing and digital logic construction; and electrical connection recognition and digital model generation. This method achieves high-precision recognition of complex electrical drawings, effectively reducing text-graphic element mismatch problems. It can automatically identify inclusion relationships represented by solid and dashed boxes, construct a multi-level digital logic structure, and the generated digital model supports electronic drawing reconstruction, facilitating manual verification and correction, significantly improving engineering practicality. Furthermore, it is not dependent on fixed templates and is applicable to substation electrical drawings from different drawing standards and manufacturers.

This invention proposes a training-free open-vocabulary fabric defect segmentation method based on semantic-geometric fusion. The method acquires a dataset of fabric defect images and preprocesses them. A training-free open-vocabulary fabric defect segmentation network model is constructed, comprising a visual-linguistic pre-trained model as the backbone and a visual base pre-trained model for pixel-level boundary calibration. A multi-scale adaptive feature fusion module (MAFF) and an entropy adaptive sparsity module (EA-TopK) are introduced into the visual encoder of the visual-linguistic pre-trained model. A semantic-geometric uncertainty fusion module (SGUF) is introduced between the visual-linguistic pre-trained model and the visual base pre-trained model. Furthermore, a multi-view visual consistency strategy is introduced in the prediction stage of the visual-linguistic pre-trained model to weightedly fuse logits from different enhanced viewpoints to obtain a coarsely segmented image. This invention improves the spatial localization accuracy, boundary localization accuracy, and robustness of defect segmentation for small, low-contrast defects.

This invention discloses a weakly supervised video anomaly localization method based on dynamic Gaussian spraying, belonging to the field of computer vision technology. The method includes: extracting spatiotemporal features from the video and initializing a set of Gaussian distribution models to characterize potential anomaly regions; recursively updating the network to drive the parameters of the Gaussian distribution models to dynamically evolve with the video time sequence; using an anomaly confidence scoring network to calculate anomaly confidence based on the features of each frame and the Gaussian distribution parameters; filtering anomaly Gaussian distributions based on the confidence to generate spatial localization results, and calculating anomaly scores for video segments by integrating the intensity parameters of each frame; constructing a loss function that includes temporal smoothing and intensity discrimination terms, and jointly training the network using video-level labels. This invention enables the detection and spatial localization of anomalous events in videos using only video-level labels.

This application discloses an image processing method, a computer-aided diagnosis method for tumors, and a computing device, relating to large model technology and the field of image processing. The method includes: performing image segmentation on the image to be processed to obtain a mask image of the detection region; determining the spatial location information of the detection region in the image to be processed based on the mask image; performing denoising processing on the image to be processed based on the spatial location information to obtain an initial enhanced image; and performing detail restoration on the initial enhanced image based on a preset resolution to obtain a target enhanced image, wherein the display clarity of the detection region in the target enhanced image is higher than the display clarity of the detection region in the image to be processed. This application solves the technical problem of poor image processing performance in related technologies.

This invention relates to the fields of computer vision and measurement technology, and proposes a method for precise localization of images without location information based on a real-world 3D model. The method includes four steps: Step 1: virtual projection image generation based on the real-world 3D model; Step 2: feature extractor construction and training; Step 3: feature library construction and feature matching; and Step 4: image pose calculation after retrieval. The virtual projection image generation method uses a real-world 3D model as a foundation, sets up a virtual camera according to preset observation distance, viewing angle range, and attitude parameters, performs multi-view virtual projection, and generates model image data associated with spatial location and height information. This invention requires no known shooting pose or spatial prior conditions, relying solely on the real-world 3D model as a reference data source to achieve high-precision spatial localization of images, filling the technical gap of being unable to locate images without location information, and greatly expanding the range of sources of locatable images.

This application discloses an asset identification method and terminal based on video spatial gridding. It uses a video spatial gridding mechanism to discretize the registered geographic video stream into grid units with geocoded mappings, and relies on multimodal feature extraction to interpret and separate asset categories and precise contours. This approach effectively solves the problems of spatial positioning errors and unstable recognition of dynamic objects across edges in traditional image analysis. The coupling between multimodal features enhances the anti-interference capability of the identification, and the coded integration realizes a structured association between asset attributes and spatial location. Ultimately, it not only significantly improves the completeness and consistency of asset extraction, eliminating boundary truncation and contour drift, but also provides digital records with semantic attributes and high-precision spatial coordinates for large-scale asset inventory and intelligent operation and maintenance monitoring.

This invention relates to the field of medical ultrasound therapy technology, and more particularly to a method and system for controlling ultrasound emission parameters in transcranial acoustic field adaptive correction for the acoustic dynamics treatment of gliomas. It includes: acquiring spatial localization information of the glioma target area and individualized skull characteristics of the patient; pre-compensating the emission parameters of each element of a multi-element phased-array ultrasound transducer based on this characteristic information to generate corrected emission parameters; driving the transducer to emit an ultrasound beam, causing the beam to pass through the skull and form a transcranial focused acoustic field in the target area based on the spatial localization information; applying ultrasound to the target area using this acoustic field in the presence of a sonosensitizer to activate the sonosensitizer and generate an acoustic dynamic effect; monitoring changes in the acoustic field state and / or tissue acoustic properties during treatment, and updating the emission parameters in real time based on the monitoring results to maintain stable focusing. This invention can compensate for acoustic beam distortion caused by the skull, improve the accuracy and stability of transcranial focusing, and achieve adaptive correction during treatment.

This invention relates to a hyperspectral camouflage target detection method based on spatial-spectral feature fusion. It combines deep learning and spatial-spectral feature extraction to establish a 3D network model suitable for spectral feature learning and target spatial localization. Anomaly detection and spatial-spectral constraints are incorporated into the model to improve detection accuracy. Simultaneously, a false positive / false negative removal training method based on spatial receptive domain is proposed to further adjust the network and improve its training effect. The RX-3DSSRF algorithm effectively compensates for the problems of missed detections and false alarms in the traditional RXD algorithm, achieving good segmentation results on low-altitude hyperspectral camouflage target detection data. By removing false positive features of the target and false negative features of the background, the final optimized target detection result is obtained.

The present application relates to the field of automation control technology, in particular to a remote-controlled port container portal crane, the crane comprising a spatial positioning mapping module, a channel priority scheduling module, an action gap compensation module, an execution frozen control module and a channel isolation module. In the present application, three-dimensional labels are generated through job coordinates and compared with logical rules to realize precise binding of control instructions and spatial tasks, enhance the mapping clarity between remote control and the job area, extract control channels and task identification to establish execution sequence and partition relationship, realize multi-instruction priority sequencing and conflict avoidance, identify potential gap intervals of the motion chain to reconstruct action continuity, dynamically generate operation lock values combined with task cycles to ensure stable operation of key sections and consistent instruction timing, extract job area proximity information to optimize channel configuration, adjust frequency band activation logic to reduce communication interference and improve control collaboration efficiency and job safety.

This invention relates to a method for detecting foreign objects (FOOs) on airport runways based on biomimetic binocular vision. The method includes: simultaneously receiving images at least two scales from cameras with different focal lengths; performing cooperative target detection based on spatial correspondence; mapping and fusing the detection results to output supervised detection results; employing an unsupervised learning algorithm for anomaly detection and outputting unsupervised detection results; cross-validating and fusing the supervised and unsupervised detection results to generate a comprehensive FEO detection result; receiving left and right view images from a biomimetic binocular camera; performing three-dimensional spatial localization of FEOs in the comprehensive FEO detection result based on a binocular stereo vision algorithm; and fusing the comprehensive FEO detection result with the FEO location information to generate a final detection report. This method integrates multiple technological advantages, forming a complete, high-precision, and robust all-weather FEO detection solution for airport runways.

The application provides a three-dimensional nuclear magnetic resonance footprint area spatial positioning and morphological analysis method and system, comprising the following steps: S1, obtaining a weighted 3D MRI magnetic resonance image, and performing footprint area layer reconstruction; S2, obtaining an image parallel to the footprint area plane, and determining a spatial positioning area; S3, performing footprint area outlining, and segmenting the footprint area; and S4, performing spatial positioning analysis and morphological analysis on the footprint area. By adopting the 3D MRI reconstruction technology, the application can accurately evaluate the ACL femoral footprint area in the oblique sagittal plane which is tangent to the attachment point of the ACL and the femur, solves the inaccurate problem of the ACL femoral footprint area morphological evaluation in the traditional 2D MRI evaluation method, and thus provides more accurate ACL femoral footprint area morphological information for doctors, and further optimizes the effect of the ACL reconstruction surgery.

This invention provides a non-equilibrium grouting method based on the multi-dimensional characteristics of tunnel rock mass. By constructing a technical system that integrates multi-dimensional geological feature identification and model-based decision-making, it achieves precise and intelligent grouting control for complex fractured rock masses in tunnels. First, fractures are comprehensively classified based on their width, connectivity, and filling state, forming spatial distribution data. This transforms geological conditions into structured, spatially locatable digital information, providing precise target objects for grouting construction. Then, through a grouting parameter decision model, the current fracture type is automatically mapped to a strictly matching combination of grout mix ratio, grouting pressure, and grouting rate, achieving intelligent and standardized decision-making from geological features to construction parameters. This process solidifies mature experience and theoretical understanding into repeatable rules, ensuring a high degree of compatibility between the initial construction plan and the inherent hydraulic characteristics of the fractures, thereby significantly improving the targeting and controllability of grouting construction.

A high-efficiency UHPC (Ultra-High-Performance Computation) fence recognition and detection method, belonging to the fields of computer vision and object detection, includes the following steps: Step 1: Construct a fence center hole dataset by acquiring fence images under different lighting conditions, angles, and occlusion levels using a binocular depth camera; Step 2: Improve the original YOLOv11n-seg network to construct a YOLOv11n-DSCA-seg model; Step 3: Train the improved YOLOv11n-DSCA-seg model using the constructed dataset; Step 4: Input the fence image to be detected into the trained YOLOv11n-DSCA-seg model; Step 5: Calculate the 3D spatial coordinates of the center hole by combining the 3D spatial localization with the depth information acquired by the binocular depth camera. This invention provides reliable technical support for the automated handling of building components and has significant practical value for promoting the development of the intelligent construction field.

The present invention relates to an additive manufacturing machine (100) for large parts (8) with good manufacturing accuracy, said additive manufacturing machine (100) comprising a nozzle (6) for depositing beads (16), a moving platform (4), a pre-positioning system (18), and a final positioning system (5) cooperating with a spatial localization system (13) to position a reference point (Pr) of said moving platform (4) along a predefined target trajectory (Ti) with a final positioning accuracy (P3), said reference point (Pr) corresponding to a reference position of the nozzle (6) outlet orifice (20). The invention also relates to a method for implementing the additive manufacturing machine (100). Figure 1

This invention provides an AR remote assistance method and system for the installation and debugging of communication equipment. By constructing a three-in-one collaborative architecture of AR smart terminal, expert collaboration platform, and cloud service brain, it achieves remote immersive assistance throughout the entire process of communication equipment installation and debugging. The method includes: the AR smart terminal collecting first-person perspective data and environmental information from the site, and uploading it to the cloud service brain via ultra-low latency transmission technology; the cloud service brain achieving spatial positioning and virtual-real overlay based on multi-source fusion SLAM technology, and completing equipment identification, instrument reading extraction, and fault diagnosis through AI models; the expert collaboration platform supporting multi-user collaborative annotation and real-time interaction, combined with guidance pushed from a communication industry-specific knowledge base; and achieving three-dimensional spatial consistency of multi-terminal virtual annotations through spatial anchor point sharing technology. This invention solves the problems of poor immersion, inaccurate information transmission, and low utilization of expert resources in traditional remote assistance, improving the efficiency and accuracy of communication equipment installation and debugging.

This invention discloses a VR virtual-real fusion interaction system and method based on large-space positioning and dynamic rendering. The system includes: a physical space modeling module to generate a spatial risk field function; a hybrid positioning engine that uses tightly coupled extended Kalman filtering to fuse SLAM visual features and UWB / Bluetooth AoA data to calculate pose, and integrates a long short-term memory network for predictive compensation in case of visual failure; a spatial synchronization and dynamic scheduling module to allocate rendering and bandwidth resources; a virtual-real fusion rendering unit that dynamically generates a fused video stream containing virtual scenes and boundary warnings; a low-latency transmission network that distributes the video stream based on WiFi 7 with a latency of no more than 15ms; and a terminal presentation and closed-loop correction module that displays and transmits IMU data to correct pose. This invention achieves highly robust positioning, dynamic boundary warning, predictive resource scheduling, and multimodal interaction closed loop, significantly improving the immersion and safety of large-space multi-person VR experiences.

This invention relates to the field of intelligent lighting control technology, and discloses a multi-lamp collaborative control method and system based on ultra-wideband spatial positioning. The method includes acquiring ultra-wideband ranging data between lamps in a target lighting system; constructing a distance matrix based on the ultra-wideband ranging data, and reconstructing the spatial coordinates of each lamp using a spatial positioning method; acquiring luminous efficacy template parameters, and calculating the target luminous efficacy state function of each lamp based on its spatial coordinates and luminous efficacy template parameters; exchanging state information with neighboring lamps using a distributed consensus collaborative method and iteratively updating the state information until convergence conditions are met; and controlling the luminous efficacy output of the corresponding lamp based on the converged state information of each lamp. This invention enables high-precision multi-lamp linkage control without relying on real-time scheduling by a central controller, significantly improving the intelligence level, deployment efficiency, and user experience of the lighting system.

This invention provides a behavior recognition method and system based on multi-channel video linkage using a 3D map, relating to the field of mine security technology. The method involves acquiring key feature data of a target area and real-time multi-channel monitoring videos; generating a lightweight 3D map model of the target area based on the key feature data using a sparse point cloud model; performing multi-feature analysis on the real-time multi-channel monitoring videos to identify candidate targets exhibiting suspected abnormal behavior in the personnel activity area and their spatial locations within the lightweight 3D map model; determining at least one monitoring video corresponding to the spatial location from the real-time multi-channel monitoring videos based on the candidate targets and the locations of each camera installation point; and performing multi-feature fusion analysis on the at least one monitoring video to determine whether the candidate target's behavior is abnormal, thereby achieving early screening and spatial localization of abnormal behavior and improving the accuracy and reliability of abnormal behavior recognition.

The invention relates to a multi-modal large language model-based three-dimensional affordability positioning method and system for a body robot, and belongs to the technical field of robot perception and body intelligence. According to the scheme, a multi-modal large model availability positioning framework is provided, the framework effectively combines semantic understanding, multi-modal fusion and 3D space positioning through an end-to-end learning mode, and the features of 3D scene data and natural language instructions are deeply mined and fully utilized. A novel scene-level affordability data representation is constructed, so that robot interaction tasks, such as door opening, drawer pulling and object grabbing, in various complex unstructured environments become more accurate and efficient; according to the method, adaptive sampling and a coarse-to-fine reasoning thought are fused, the calculation efficiency and robustness of understanding of a robot on a large-scale three-dimensional scene are improved by optimizing a data retention strategy and enhancing space consistency, and the limitation of a traditional method under the complex background and the ambiguous instruction in the real world is effectively solved.