102 results about "Space transformation" patented technology

The invention discloses a power transmission line inspection unmanned aerial vehicle obstacle intelligent detection and obstacle avoidance method. The method comprises the steps that multi-source sensing data is acquired, and alignment is completed through calibration and timestamp matching; heterogeneous data preprocessing and feature enhancement; constructing a high-precision environment fusing a geometric structure and a semantic tag, mapping a two-dimensional target detection result output by the recognition network to a three-dimensional coordinate system through spatial transformation, and fusing the two-dimensional target detection result with a point cloud structure to construct a semantic occupation grid map; performing preliminary route planning according to a preset power grid topological structure and task coverage requirements, and generating a barrier-free flight path covering the whole inspection area; reinforcing learning of a dynamic obstacle avoidance strategy; track dynamic reconstruction and energy consumption optimization scheduling are carried out; the technical problems that an existing technical system has defects in the aspects of obstacle recognition accuracy, complex environment adaptability, data fusion capacity and obstacle avoidance strategy intelligence, and the requirements for high-reliability, low-energy-consumption and high-efficiency unmanned aerial vehicle power transmission line inspection are difficult to meet are solved.

The invention provides a video coding and compression transmission method for intelligent equipment, which belongs to the technical field of video coding and compression transmission, realizes advanced channel quality estimation by establishing a channel prediction model enhanced by a spatial transformation network, configures layered coding parameters according to a prediction result and optimizes motion search in combination with inertial measurement unit data. An optimal coding mode is selected by an application rate distortion optimization algorithm, enhancement layer sending is adjusted according to real-time feedback by adopting a dynamic transmission scheduling strategy, and differential forward error correction protection and a time-space domain error concealment mechanism are matched; the technical problem that video transmission is unstable due to violent fluctuation of channel quality when intelligent equipment moves and works in complex electromagnetic environments such as a transformer substation is solved.

The invention relates to a privacy protection technology in a large language model, and provides a method for realizing privacy protection through embedded space conversion in retrieval enhancement, which comprises the following steps of: generating a target training sample by utilizing a pre-training language model based on near-end optimization training; and the potential privacy leakage query statement is mapped to the security target in the embedding space through the conversion of the embedding space to realize the protection of privacy. According to the scheme, on numerous representative large language models and data sets, it is proved that the normal function of the retrieval enhancement generation model is guaranteed, and meanwhile the defense efficiency is excellent.

This application discloses a method, apparatus, device, and storage medium for determining the pose of a photovoltaic panel, relating to the field of photovoltaic technology. The method includes: performing coordinate transformation and spatial filtering based on the original point cloud and corner pixel coordinates of the target photovoltaic panel to obtain a region of interest (ROI) point cloud of the target photovoltaic panel; performing filtering and matching processing on the ROI point cloud to obtain a photovoltaic target point cloud of the target photovoltaic panel; constructing a photovoltaic panel coordinate system of the target photovoltaic panel based on the spatial transformation relationship between the photovoltaic target point cloud and the point cloud template; and obtaining the target spatial pose of the target photovoltaic panel in the camera coordinate system based on the axial correspondence between the photovoltaic panel coordinate system and the camera coordinate system. By fusing the point cloud information and corner pixel coordinates of the photovoltaic panel, and further performing filtering and transformation processing to construct the photovoltaic panel coordinate system, the accuracy of photovoltaic panel pose recognition is improved.

This invention relates to the field of robot intelligent control and precision operation technology, and discloses a charging gun insertion positioning and error correction method and device based on a robotic arm. The method guides the charging gun into the target charging port based on constant force control. When edge contact jamming occurs, the constant force is maintained, and a six-dimensional force sensor is used to collect force and torque information in real time. By establishing three spatial coordinate systems at the end of the robotic arm, spatial transformation and decoupling analysis of force and torque are performed. Using the constraint that the z-direction of the charging gun end plane is zero, the actual contact position of the charging port relative to the charging gun end plane is calculated, obtaining the jamming correction vector. The jamming correction vector is transformed into a coordinate system based on the end of the robotic arm to obtain the target movement vector. The robotic arm is then controlled to move along this direction, achieving precise error correction and smooth insertion. This invention can accurately identify the jamming point location, achieve directional correction, and improve the charging gun insertion success rate.

The invention discloses a positioning device for registration and fusion of a multi-modal medical image, and relates to the field of image auxiliary equipment, the positioning device comprises a transfer plate and a support, the transfer plate and the support are detachably connected, and the transfer plate is provided with a positioning assembly; according to the positioning device, by arranging the positioning assembly, the guide rail and the sliding block structure of the positioning assembly, a fixed physical track is provided for supporting tool adjustment, randomness of manual positioning is avoided, the position of the supporting tool can be rigidly fixed through a rack and an insertion block type locking component, no displacement risk exists after adjustment, an air bag on the inner side of the supporting tool forms flexible physical constraint, and body micro-movement of a patient in the examination process is limited; the whole set of design ensures that the body position of a patient has mechanical structure guarantee in the whole process from positioning, fixing to examination during each time of scanning, the body position difference of two times of scanning is extremely small, the body position repeatability is greatly improved, a software algorithm can give full play to the space transformation effect, perfect image registration is easily achieved, and a solid support is provided for the high-precision requirement of multi-modal image fusion.

This invention relates to the field of prefabricated building installation accuracy control technology, specifically disclosing an intelligent calibration system for prefabricated building installation accuracy that integrates point cloud data and BIM. The system collects on-site point clouds and acquires design geometric information; after distance attenuation compensation for the point cloud reflection intensity, it generates on-site and design projection intensity maps along the normal direction of the component installation plane, and obtains the in-plane pre-translation amount through phase correlation calculation; using this pre-translation amount as the prior center, it constructs a spatial transformation objective function containing an exponentially growing dynamic penalty coefficient, and solves the final spatial transformation parameters through weighted centroid closed-loop inversion iteration; it calculates the signed vertical distance along the normal of the aligned point cloud and removes outliers to form a deviation field, outputting translation or rotation fine-tuning amounts according to the median deviation of the joint sub-region; this invention solves the in-plane sliding problem in the registration of flat components, improving the accuracy of installation deviation calculation and calibration reliability.

The invention discloses a multi-mode fault diagnosis method and device based on automatic machine learning and a medium, and the method comprises the following steps: S1, collecting a signal sample of a rotating mechanical device, and carrying out the preprocessing; the signal sample comprises an acoustic signal and a vibration signal; s2, according to the processed signal samples, acoustic signal features and vibration signal features are extracted respectively, the features are fused and normalized, and multi-modal fusion features are obtained; s3, according to the multi-modal fusion features, feature space transformation and dimension reduction processing are carried out through an automatic machine learning process, and an optimized feature set is obtained; s4, performing model selection, hyper-parameter optimization and training according to the optimized feature set to obtain a plurality of base models; s5, based on the plurality of base models, performing model fusion by adopting a double-layer integrated learning strategy to obtain an integrated diagnosis model; and S6, inputting a to-be-diagnosed sample into the integrated diagnosis model, and outputting a fault type.

This invention discloses a multi-label classification method for medical images based on spatiotemporal integration and adaptive normalization, belonging to the field of medical image processing and computer-aided diagnosis technology. The invention proposes an improved scheme. Firstly, it constructs a dynamically balanced data input stream through frequency inverse sampling and cosine annealing strategies. Secondly, it introduces group normalization to replace batch normalization in the feature extraction network, solving the training convergence problem under limited GPU memory. During training, it employs a focal loss function to mine hard-to-classify samples. In the inference stage, the invention constructs a spatiotemporal integration mechanism based on multi-view space transformation and multi-stage model weight fusion, combined with an adaptive threshold decision algorithm based on maximizing F1 scores to output diagnostic results. This invention effectively improves the recognition accuracy of key pathological features such as cardiac hypertrophy and edema, while significantly improving the recall rate of small lesions, exhibiting high robustness and clinical auxiliary value.

The embodiment of the invention discloses a three-dimensional model cutting method and device, and the method comprises the steps: receiving at least one bottom surface vertex coordinate set and at least one box height value of at least one bounding box inputted by a user, and at least one spatial transformation matrix corresponding to the at least one bounding box; determining at least one initial bitmap based on the at least one vertex coordinate set and the at least one spatial transformation matrix; performing coding processing on the at least one initial bitmap to obtain a target bitmap; based on the target bitmap and the at least one box height value, determining whether at least one pixel point in a to-be-cut model is located in the at least one bounding box, and obtaining at least one judgment result; cutting the to-be-cut model according to the at least one judgment result; according to the embodiment of the invention, rapid cutting judgment is realized, and the model cutting efficiency is improved.

The invention relates to a semantic abstract generation and security verification method and system for multi-modal data, and belongs to the technical field of data security and information process.The method comprises the steps that feature extraction is carried out, specifically, a client processes a multi-modal plaintext data stream, and key features of the multi-modal plaintext data stream are extracted; vector mapping: the client performs vector space transformation on the extracted features to generate a unique corresponding data abstract; local persistence: the system stores the generated data abstract back to a user local database for subsequent use; abstract collection, wherein the detector obtains generated data abstract information from the data stream; and integrity verification: carrying out consistency comparison on the data stream with the data abstract by the detection and verification device, and judging whether the data stream is complete and reliable or not. The data abstract generation system comprises a client module, a server module and a middleware module. According to the method, efficient and reliable data stream abstract generation and credible management are realized on the premise of fully guaranteeing user data privacy.

An electronic device performs a method of decoding video data, including: receiving video data corresponding to a coding unit from a bitstream, where the coding unit is encoded in an inter prediction mode or an intra block copy mode; receiving a first syntax element from the video data, where the first syntax element indicates whether the coding unit has any non-zero residual; in accordance with a determination that the first syntax element has a non-zero value: receiving a second syntax element from the video data, where the second syntax element indicates whether the coding unit has been encoded using an adaptive color space transform (ACT); in accordance with a determination that the first syntax element has a zero value: assigning the zero value to the second syntax element; and determining whether to perform inverse ACT on the video data of the coding unit according to the value of the second syntax element.

The application relates to the technical field of precision microelectronic and semiconductor devices, and discloses a micro TEC cold and hot ceramic sheet stacking and mounting production line control method and system.The method comprises the following steps: collecting a surface image of a ceramic sheet to be mounted and extracting key points to generate a real-time deformation distribution benchmark; registering the benchmark with a previous layer benchmark and performing differential operation to construct a deformation difference field; solving a three-dimensional deviation when a displacement amplitude exceeds a limit, fusing historical interlayer cumulative deviation to determine a total cumulative offset vector and a total cumulative rotation angle; decomposing the total deviation into a rigid translation and a distortion deformation component, generating a space transformation compensation instruction based on a warping mode match, projecting and transforming a theoretical coordinate with the instruction and synthesizing a vector to generate a final adjustment mounting coordinate to drive a mechanical arm to move, and finally updating compensation parameters in a closed loop iteration according to actual interlayer deviation feedback. The method can solve the problem of low non-linear thermal deformation compensation precision in the multi-layer ceramic mounting process of the prior art.

The application discloses an aluminum veneer engraving path automatic correction method and system based on positioning image feedback, and particularly relates to the technical field of digital program control path correction, and is used for solving the problem that the path correction method includes the protective film pseudo edge into the space transformation calculation, resulting in the deviation of the corrected path from the real contour of the aluminum veneer; the positioning image containing the protective film state of the aluminum veneer surface is acquired, and the boundary features are extracted; the plate material contour in the original tool path data is called to match each boundary feature with the predetermined cutting contour to determine the candidate pseudo features; the brightness gradient asymmetry is calculated along the normal direction of the boundary feature to identify the protective film tearing pseudo edge; the process feature point coordinates are called to review the protective film tearing pseudo edge; after the protective film tearing pseudo edge determined in the pseudo edge judgment and the candidate pseudo features are removed, the remaining boundary features form a trusted boundary feature set; and only the trusted boundary feature set is used as a constraint to calculate the space transformation relationship and to perform overall transformation on the original tool path data.

The invention discloses a multi-process thermal mechanical deformation monitoring and compensating system for an engine support, and relates to the technical field of image monitoring. The system firstly collects visible light and infrared data to generate a fusion image, and recognizes the chip shielding state of the surface of a workpiece; if yes, driving a physical cleaning device to perform directional cleaning until a judgment logic is met, and generating a hot historical log containing cleaning process data; calculating a current three-dimensional full-field deformation vector of the workpiece in combination with the recovered pure surface temperature field and a thermal history log; and finally extracting the displacement and rotation amount of the key feature points, and generating a machine tool coordinate system space transformation instruction for compensation. According to the method, the problems of infrared temperature measurement distortion caused by inner cavity chip accumulation under the minimal quantity lubrication working condition and thermal lag prediction deviation generated by internal heat accumulation after chip removal are solved.

A method for planning an autonomous landing safe trajectory suitable for UAV inspection is proposed. After the UAV takes off, it acquires the current pose of the UAV, the point cloud map information of the surrounding environment, and the spatial orientation of the target point. By establishing the spatial transformation relationship between the camera coordinate system and the body coordinate system, the orientation vector relative to the ground target is calculated, and the target point is generated within the radar-sensible area. Based on the generated target point, an optimization function that minimizes the fourth derivative is constructed based on multiple high-order polynomials to generate a smooth trajectory with continuous position, velocity, and acceleration. The initial trajectory is transformed into a B-spline curve, and local trajectory segments with collision risk are optimized. A segmented penalty mechanism is designed based on the boundary threshold to guide the trajectory away from the boundary area. The time span is adjusted by calculating the trajectory derivative over-limit ratio, and a trajectory similarity cost function is introduced to ensure that the final optimized trajectory maintains the original geometric structure, thereby achieving trajectory control optimization and safe landing of the UAV.

The application discloses a medical image and physical space registration method and system based on automatic recognition of reference points and a system, and relates to the technical field of medical navigation. The method comprises the following steps: setting reference points which can be sensed in medical image space and physical space simultaneously; automatically recognizing and positioning the image space coordinates of the reference points in the medical image; obtaining the physical space coordinates of the reference points in the physical space; and automatically calculating the space transformation relationship between the image space and the physical space based on multiple sets of corresponding point coordinates. In the method, the positioning of the reference points in the image space adopts a multi-stage method from coarse to fine, candidate regions are extracted through coarse positioning, and a deep learning model is combined to obtain sub-pixel level three-dimensional coordinates. The application realizes full-automatic recognition and high-precision positioning of the reference points, does not require manual point selection, the registration accuracy can reach the sub-millimeter level, the calculation speed is fast, the application is suitable for intraoperative real-time navigation, and the accuracy and efficiency of minimally invasive interventional surgery can be significantly improved.

The application provides a power prediction method based on probability space transformation and neural network fusion, comprising the following steps: S1, constructing a one-dimensional probability analyzer; S2, probability space mapping and transformation; S3, feature fusion mechanism; S4, inputting the fused feature representation into multiple series probability transformation modules to form a multi-layer stacking structure, and performing layer-by-layer feature extraction combined with a residual connection mechanism; S5, outputting the final power prediction result through a fully connected regression network; S6, model training and evaluation: performing end-to-end training on the overall model, adopting the model with the minimum error in the verification set as the optimal training model, and testing the accuracy of the model prediction in the test set. The method can effectively cope with the non-Gaussian and multi-peak distribution problems of meteorological variables such as wind speed and wind direction, has strong adaptability to complex atmospheric environment, significantly improves the accuracy and stability of wind power prediction, and has good engineering application prospect.

The invention belongs to the technical field of unit vibration non-contact measurement, and discloses an interference fringe image contrast improvement method based on a polynomial rooting principle, aiming at solving the problem that the detection precision of a coherent dispersion spectral imaging device is interfered due to too low interference fringe image contrast caused by a large optical path difference. The method comprises the following steps: converting an image into a brightness space with uniform perception, providing a basic brightness channel, converting local brightness distribution of the image into expression of a polynomial coefficient, and converting static polynomial coefficient expression into a root state space; and mapping the brightness distribution of the low-contrast interference fringe image to the target range of the high-contrast interference fringe image, re-decoding a result after root space transformation into a perception brightness graph, and migrating color information from a low-contrast interference fringe image domain to a high-contrast interference fringe image domain. The method has the advantages of excellent local adaptability, accurate control, high flexibility and nonlinear dynamic range mapping.

This invention discloses a multimodal balanced learning method based on label reshaping. Starting from label-side design, this invention proposes for the first time to balance multimodal learning by reshaping the cross-modal label space. A reshaping matrix is ​​generated using the prediction output of complementary modalities to inject cross-modal inter-class information. Simultaneously, the reshaping intensity is adaptively calculated based on the confidence differences between modalities to control the degree of label space transformation, thereby balancing the mapping difficulty from the feature space to the label space for different modalities. Finally, a target parameter update strategy is employed, applying different optimization objectives to different parts of the model to ensure training stability. This method dynamically reshapes the label space corresponding to each sample and each modality, constructing a balanced soft supervision signal rich in inter-class relationships. This balances the learning difficulty of different modalities at the source and injects cross-modal inter-class information to enhance the model's discriminative ability.

The invention is suitable for the technical field of water meter shell cutting, and particularly relates to a water meter shell water inlet and outlet cutting and positioning method and device.The method comprises the steps that part of contour point cloud of a machined reference cylindrical surface on a water meter shell body is obtained, and an actual central axis is obtained through fitting; determining a theoretical central axis and a theoretical central axis according to the water meter shell; based on the theoretical central axis and the theoretical central axis, determining the relative position relation of the two in the water meter shell; obtaining required spatial transformation parameters based on the relative position relationship; obtaining a virtual central axis from the theoretical central axis based on the spatial transformation parameters; constructing a virtual reference cylindrical surface based on the virtual central axis and the nominal outer diameter parameter; obtaining a theoretical intersecting line according to the virtual reference cylindrical surface and the theoretical three-dimensional model, and obtaining actual cutting position data based on the theoretical intersecting line; according to the water meter shell water inlet and outlet cutting positioning method, the water inlet and the water outlet are formed in the weld joint, and the structural strength can be improved similar to formation of weld joint patches.

An electronic device performs a method of decoding video data, including: receiving video data corresponding to a coding unit from a bitstream, where the coding unit is encoded in an inter prediction mode or an intra block copy mode; receiving a first syntax element from the video data, where the first syntax element indicates whether the coding unit has any non-zero residual; in accordance with a determination that the first syntax element has a non-zero value: receiving a second syntax element from the video data, where the second syntax element indicates whether the coding unit has been encoded using an adaptive color space transform (ACT); in accordance with a determination that the first syntax element has a zero value: assigning the zero value to the second syntax element; and determining whether to perform inverse ACT on the video data of the coding unit according to the value of the second syntax element.

The application discloses a kind of real-time non-rigid registration method and system of surgical navigation image based on deep learning, comprising: calibration the space transformation matrix between RGB-D camera and medical imaging equipment;RGB-D camera real-time capture patient whole body optical image and depth information;Medical imaging is carried out to patient, while reconstructing the initial model of patient based on SMPL based on human posture estimation neural network of deep learning;Medical image is point cloud, and the point-to-plane spatial position matching relationship between medical image and the initial model of patient is established;Neural network reconstructs the dynamic change of patient model in operation in real time, and the non-rigid deformation field of medical image is solved by matching relationship;Registration medical image, update display.The application can realize the real-time non-rigid registration of surgical navigation medical image to different body posture patient, solve the movement, deformation of organ, soft tissue caused by patient posture change in surgical navigation, thereby cause the problem of medical image failure, help to improve the accuracy of surgical navigation.

The application provides a slender structure space transformation device for a wind tunnel laboratory, and belongs to the technical field of wind tunnel experiments.The device comprises a multi-axis frame, a suspension assembly, three-axis servo motors, a plurality of encoders and a central control system.The multi-axis frame has rotational freedom degrees around the X-axis direction, the Y-axis direction and the Z-axis direction of a three-dimensional coordinate system.The suspension assembly is connected to the inner side wall of the multi-axis frame and is used for suspending a test model to be tested.The three-axis servo motors are used for driving the rotation of the multi-axis frame.The plurality of encoders are connected to the three-axis servo motors and receive rotation angle signals.The central control system is electrically connected to the three-axis servo motors and the plurality of encoders, and is suitable for controlling the operation of the three-axis servo motors to accurately adjust the space posture of the test model.The slender structure space transformation device for the wind tunnel laboratory has the technical effects of high posture adjustment precision, simple adjustment and accurate control of the posture of the test model for the slender structure in the wind tunnel experiment.

The invention discloses a heart CTA image group registration template construction method based on deep learning, and aims to solve the problems of fuzzy template, low construction efficiency and unstable alignment caused by significant deformation and intensity difference of the heart under different cardiac phases and scanning conditions. The method comprises the following steps: acquiring and preprocessing a heart CTA image; grouping is carried out based on feature similarity and topology persistent coherence; a multi-resolution three-dimensional registration network is adopted in each level to carry out coarse-to-fine registration estimation; differential deformation is realized through a space converter, and end-to-end and symmetric training is carried out by adopting combined loss of multi-scale reconstruction similarity and displacement field smooth regularization; clear sub-templates are obtained through iteration of a registration-averaging-updating strategy in the sub-groups, and finally, the sub-templates are converged step by step to obtain a global heart template, so that efficient and stable heart template construction is realized.

The invention provides a double-flow layered brain image registration method and system based on a state space model, and relates to the technical field of medical image processing, and the method comprises the steps: obtaining a moving image and a fixed image, carrying out the global feature modeling, and obtaining the global features of the moving image and the global features of the fixed image; based on the global features of the two images, explicit matching is carried out, and multi-direction matching features and initial deformation prompts are generated; performing learnable voting fusion on the multi-direction matching features to obtain globally consistent matching features; further decoding to generate a dense nonlinear deformation field; and based on the dense nonlinear deformation field, performing spatial transformation and resampling on the moving image to obtain a registration result aligned with the fixed image. Through hierarchical recursion and multi-direction modeling strategies, state space modeling and explicit feature matching are used for obtaining and sharing information most relevant to brain anatomy from multi-source direction prior, and therefore the accuracy and stability of three-dimensional brain image registration are improved.

The application relates to the technical field of building information modeling, and particularly discloses a BIM graph element reference association processing method and system based on composite space transformation, which can indicate and store state information of a reference instance by constructing a lightweight storage data model of structure data, reduces the storage cost of a processing process, stably converts all operations into modification of storage data, eliminates the positive correlation between operation times and storage cost, can greatly improve the efficiency in a synchronous scene, significantly improves the performance of BIM software, and has high application value and popularization.

The application discloses an orthodontic arch wire evaluation method based on coplanar isogonal vectors, and relates to the field of orthodontic arch wire bending evaluation. The application is based on a set of bending point information of a theoretical orthodontic arch wire curve after spatial transformation, a set of bending point information of an actual orthodontic arch wire curve after spatial transformation, judges the isogonal point distance of the orthodontic arch wire, then combines the judgment of the isogonal vector angle and the maximum included angle between the coplanar isogonal vector group of the orthodontic arch wire and a base vector, and establishes an orthodontic arch wire evaluation method based on coplanar isogonal vectors. The application effectively improves the efficiency of the parameterized evaluation of the orthodontic arch wire, realizes the quantitative evaluation of the orthodontic arch wire by calculating the isogonal point distance of the actual orthodontic arch wire curve, the isogonal vector angle and the maximum included angle between the coplanar isogonal vector group of the orthodontic arch wire and the base vector.

This invention belongs to the field of data processing and discloses a deep learning-based method for dynamic monitoring of blue-green space in urban renewal. The method includes: acquiring high-resolution remote sensing images and preprocessing them to obtain preprocessed image data; determining a UNetMamba model and improving it using a ResTv2 backbone network and ECA attention mechanism to obtain an improved UNetMamba model; determining a preset dataset and training and optimizing the improved UNetMamba model based on the preset dataset to obtain a blue-green space classification model; inputting the preprocessed image data into the blue-green space classification model and outputting the blue-green space classification results; and performing post-processing and spatial transfer econometric analysis on the blue-green space classification results to obtain a blue-green space distribution map and a blue-green space transformation Sankey diagram. This invention can effectively improve the identification accuracy of blue-green space, thereby effectively revealing the evolutionary patterns of ecological space.