51 results about "Range searching" patented technology

The invention provides a method and a device for improving the long-range deep search reasoning capability of a large language model. According to the method for improving the long-range deep search reasoning ability of the large language model, the long-range search reasoning ability of the large language model in cross-domain complex problems is remarkably improved, and the large language model is improved through high-difficulty data automatic generation driven by the knowledge graph and a three-layer difficulty screening mechanism. And by combining a two-stage training strategy of supervised fine tuning and GRPO reinforcement learning, the multi-hop reasoning depth of the model and the collaboration of external tool calling are effectively enhanced.

The invention relates to the field of binocular stereo vision, and discloses a binocular stereo matching method and system based on multi-scale iterative optimization and related equipment. The method comprises the steps of performing semantic structure feature extraction on a corrected left view and a corrected right view through a feature extraction network; determining a cost space pyramid and a probability matrix according to the left view feature map and the right view feature map output by the feature extraction network; extracting a multi-scale context feature of the left view through a context sensing network, and initializing the multi-scale context feature to obtain a hidden state and input of a recursive network; during recursive network iteration, according to the probability matrix, a multi-range search strategy is adopted to index local cost from the cost space pyramid; and inputting the local cost into the recursive network, combining the hidden state obtained by the context features and the input, and carrying out iterative updating on the parallax value to obtain a parallax map, thereby improving the accuracy of parallax calculation.

The invention discloses a strabismus mode ground interference source positioning method and system based on a satellite platform, a storage medium and electronic equipment, and the method comprises the following steps: firstly, carrying out the range pulse compression of a received interference signal according to the related parameters estimated by the prior information of an orbit and the time-frequency relation; constructing a matched filter, realizing linear range migration correction, and eliminating a first-order component of range migration in a range frequency domain-azimuth time domain; a CTF search algorithm is used, and under different search distances, distance direction search is achieved through a cooperation mechanism of global search and step length constraint local optimization; constructing matched filters at different distances to realize quadratic term range migration correction, and eliminating a range migration quadratic term in a two-dimensional frequency domain; azimuth distance compression is carried out at different distances, so that signals are focused; and obtaining a two-dimensional positioning position opposite to the interference source through coordinate transformation. According to the invention, the problem of signal defocusing caused by the distance-orientation strong coupling effect is solved in the squint mode, and the interference source positioning precision is significantly improved.

To support efficient derivation of an optimum facility capacity.SOLUTION: The information processor 10A includes a processing unit 20. The facility-capacity optimizing unit 20A of the processing unit 20 searches for the provisional facility capacities of the supplying facilities 32 in descending order of priority from a narrower search range as the priority becomes lower, based on the priority of the supplying facilities 32 that are included in the energy system 30 and supply energy, the input search range of the facility capacities, the number of searches, and the target energy introduction ratio of the energy system 30. The simulation execution unit 20B of the processing unit 20 derives the introduction cost of the supply facility 32 having the searched temporary facility volume and the energy introduction ratio of the energy system 30 including the supply facility 32 having the temporary facility volume every time the temporary facility volume is searched. The processing unit 20 repeatedly executes the search for the provisional facility capacity and the derivation of the introduction cost and the energy introduction rate every time the target energy introduction rate is changed.SELECTED DRAWING: Figure 1

The invention discloses an implicit generation method for a thin-wall lattice sandwich structure, which belongs to the technical field of lightweight design and comprises the following steps of: dividing hexahedral grids in a to-be-filled area of a thin-wall model, constructing a shape follow-up distance field in hexahedral unit cells, and constructing a shape follow-up distance field in combination with the lattice type of the current hexahedron, design parameters and lattice information of adjacent hexahedrons. Generating a conformal dot matrix distance field; meanwhile, voxelization processing is conducted on the thin-wall grid model, boundary voxels are extracted, corresponding patch information is stored, a boundary hexahedron searches for the nearest boundary voxel patch information by expanding the search range, a model boundary distance field is generated through implicit bias, and finally Boolean fusion is achieved with a dot matrix distance field. According to the thin-wall interior filling method, the dot matrix microstructure serves as the center, high-quality parallel filling of the cross-scale microstructure of the large-size thin-wall model interlayer can be efficiently achieved, and the problems that an existing mainstream implicit filling method is insufficient in precision, and an explicit filling method is difficult to fuse and sharp in transition can be solved.

Methods and apparatus for sparse column-aware encodings for numeric data types, including integer data and floating-point data (float, double, etc.). The encoding schemes are tailored to take advantage of column addressable memories such as stochastic associative memories (SAM) to enable Stochastic Associative Search (SAS), which is a highly efficient and fast way of searching through a very large database of records (order of Billions) and finding similar records to a given query record (search key). Techniques are also disclosed for performing range searches for both integer and floating-point data types. The integer or float data is converted to Hexadecimal form and encoded using an m-of-n constant weight encoding. Only the columns with set bits in search keys need to be read, which significantly reduces the number of reads required for searches.

The application relates to a focusing method and device, a computer device and a storage medium. The method comprises the following steps: obtaining an estimated object distance of a target object; obtaining a corresponding focus value curve based on the estimated object distance, and determining a plurality of first focusing positions based on a peak point of the focus value curve; obtaining a corresponding farthest object distance curve and a nearest object distance curve based on the estimated object distance, and determining a focusing range based on a real-time zoom ratio, the farthest object distance curve and the nearest object distance curve; searching for a second focusing position based on the plurality of first focusing positions and the focusing range, wherein a focus value of the second focusing position for imaging the target object satisfies a preset condition. Through the application, the technical problem of low focusing efficiency in the related art is solved, the search process is shortened, the computing resource overhead is reduced, and the focusing efficiency is improved.

This invention provides a digital measurement method and system for aircraft skin depression areas based on point cloud scanning, relating to the field of parts inspection technology. The method includes: acquiring point clouds of the skin surface; locating feature regions based on local curvature; extracting and fitting feature contour lines using projection and envelope methods; selecting measurement points on the contour lines; determining the search range on the opposite side; searching and calculating the shortest geodesic length between the two contour lines; establishing a cross-section at the geodesic endpoint and projecting the local point cloud; fitting a linear segment and determining the theoretical contour points based on a deviation threshold; calculating the projection compensation distance; and obtaining the final size of the depression area by combining the geodesic length. The advantages are: acquiring point clouds through 3D scanning eliminates the need for dedicated molds, reducing costs and improving inspection accuracy and efficiency; the feature extraction method based on local curvature perception accurately extracts the required feature contour lines; and the use of the shortest geodesic and projection distance compensation effectively reduces contour recognition errors, achieving accurate measurement of the depression area.

The invention discloses an aircraft maintenance task allocation and scheduling method based on a cutting strategy and time constraint, and relates to the technical field of aviation maintenance management and operation planning optimization. The method comprises the following steps: determining a maintenance opportunity that each airplane stays on the ground and meets a maintenance resource condition according to a known flight timetable; according to the technical scheme, a mathematical heuristic solving algorithm is designed, a hybrid algorithm combining a constructive heuristic algorithm and an accurate solving technology is developed, an infeasible solution is dynamically repaired through two core processes of resource monitoring and airworthiness monitoring, and an approximate optimal solution is quickly found in a large-range search space; a maintenance task is decomposed into feasible parts through a cutting strategy to adapt to a limited time box, and the problem of infeasibility caused by resource conflicts is solved; and meanwhile, solving is performed through a mathematical heuristic algorithm, so that the total cost including direct maintenance cost and extra waste cost caused by advanced maintenance is minimized while the calculation efficiency is ensured.

A method in a data processing system and apparatus for organizing electronic data, structured or unstructured, with associated metalabels, on a recordable medium of a data processing system. The data items are structured in a heterogeneous metalabel search structure. Searching the plurality of metalabel search structures can then be done with a heterogenous-range query including at least one term and one range query. The query can also be specified in a natural language which is converted to the specific heterogenous-range query. Also provided is a method for searching using geometric shapes that are mapped into range-string queries where the ranges are d-dimensional hyper-rectangles or cubes that cover the geometric shape. A method to aid the selection of the appropriate shape is provided as well as a method to cost-efficiently select the hyper-rectangles and cubes to cover the shape.

The invention relates to trimming search space for nearest neighbor determination in point cloud compression. A search space for performing a nearest neighbor search for encoding point cloud data may be tailored. Points to be excluded from the search space may be identified using a range of space filling curves. In addition, neighboring voxels may be searched based on the ranges of the space filling curve to identify any neighboring points that are lost during the tailored search.

Aiming at the problems of unbalanced energy and low coverage efficiency of the existing WSN, the invention provides a WSN energy efficient coverage method and device for improving a bacterial foraging algorithm (EABFO). According to the method, an adaptive variation method is adopted during position updating, before a predicted peak value is reached, the variation probability is increased along with the increase of the number of iterations, the randomness of bacteria is increased, global large-range search is realized, omission of an optimal solution is avoided, local optimum is tended, after the predicted peak value is reached, the variation probability is reduced, and local detail search is realized; time and space waste is reduced, and algorithm efficiency is improved. In addition, an elite breeding strategy is innovatively provided, the optimal fitness value is reserved, reverse growth of the fitness value is avoided, and the whole flora is evolved in a better direction. Compared with a genetic algorithm, the method has the advantages that the service life of the sensor matrix is prolonged by 200%, the method is suitable for energy-limited scenes such as large-scale static monitoring and dynamic target tracking, and the life cycle of the WSN is effectively prolonged.

The present invention relates to a method for reconstructing a quadratic delay model of a radio source for space-based VLBI, comprising: determining parameter information for space-based VLBI according to a preset observation plan, and obtaining radio source baseband signal sequences from a first station and a second station; determining a search range and a search step size within a two-dimensional time delay-delay rate space, and dividing the two-dimensional time delay-delay rate space into a grid according to the search range and the search step size; segmenting the radio source baseband signal sequences from the first station and the second station for each grid point, obtaining the maximum correlation amplitude mean corresponding to each grid point; finding the grid point with the largest maximum correlation amplitude mean; and reconstructing a quadratic delay model of the radio source for space-based VLBI based on the grid point with the largest maximum correlation amplitude mean. The present invention can obtain a high-precision radio source delay model while maintaining an order of magnitude increase in computational complexity.

This application relates to a method and apparatus for re-acquiring a lost-lock signal based on a short-time burst synchronization signal. The method includes: designing a short-time burst synchronization signal based on a short-time burst system; reading a certain length of baseband received signal from the starting position of the short-time burst synchronization signal and generating a corresponding local modulation sequence; calculating a carrier signal; performing coherent integration on the carrier signal to obtain the amplitude; determining whether the carrier Doppler frequency range search is complete; if not, recalculating the amplitude according to the Doppler search step; if complete, determining whether the phase search in the time domain is complete; if not, shifting the starting position forward by a certain distance and calculating a new amplitude; if complete, finding the maximum amplitude; and determining whether the maximum amplitude of the acquisition search is greater than a preset threshold; if greater, the signal is considered successfully acquired; otherwise, the signal is considered not acquired. This method can improve the success rate of acquisition.

The invention relates to a fault interval positioning method and system based on distribution network fault characteristics and instantaneous power characteristics, and the method comprises the steps: extracting fault signals in feedback signals in response to the obtained feedback signals of a monitoring node of a distribution network, and the number of the feedback signals is multiple; positioning the fault signal and determining a fault position range according to a positioning result; the method comprises the following steps: analyzing the direction of instantaneous power in a fault signal, positioning in a fault position range according to the direction of the instantaneous power, determining a fault position point, and normalizing the instantaneous power in the fault signal according to the value of the instantaneous power when positioning in the fault position range according to the direction of the instantaneous power. According to the fault interval positioning method and system based on the distribution network fault characteristics and the instantaneous power characteristics, range searching is carried out in a mode of signal extraction and direction determination, and the fault can be quickly positioned in the mode.

The present application relates to the technical field of sharing management. The present application relates to a sharing resource management terminal service platform based on the Internet of Things. It comprises a user connection unit, a range searching unit, a vehicle connection unit, a vehicle management unit, a management background unit and a sharing resource management terminal management unit; the user connection unit opens the operation authority of the sharing resource management terminal to the user with accurate information; the range searching unit detects the shaking of the vehicle, and when the shaking is detected, the shaking vehicle and the nearby vehicle are synchronously started to search the equipment, and the sharing resource management terminal to be matched within the effective range of the shaking vehicle is screened; through the collaborative design of the range searching unit and the vehicle connection unit, the present application optimizes the equipment matching logic; when the shaking of the vehicle is detected, the distance threshold screening and the real-time position tracking dynamically update the list of the equipment to be matched, so as to ensure the effectiveness of the candidate equipment; this dynamic screening combined with the priority matching mechanism reduces the probability of invalid matching.

The present invention discloses a multi-user encrypted data cross-domain search method based on constant-level query tokens, which relates to the field of encrypted data range search. The system model consists of a user group, data owners in different search domains, a blockchain, and a cloud server. The method includes: an initialization phase, a key distribution phase, an index construction phase, a token generation phase, and an authorization query phase. The data owner constructs a sequential weighted inverted index to implement a randomized constant-level query token, uses a binary tree sorting tree to store different values to determine the search range, reduces the query overhead and improves the security of the query process; introduces a request chain and an authorization chain to save the key information related to the user and the data owner and to construct and search the index, thereby achieving a secure search. The present invention reduces the overhead of multiple users in the range search process, improves the security of the query process, and ensures the security and integrity of the query information during multi-data domain range search.

The application discloses a DEM sunshine duration method and system based on GPU and terrain entropy optimization, and belongs to the field of geographic information science and spatial computing technology. The method firstly calculates terrain information entropy by using digital elevation model data to quantize the terrain complexity of different regions; and then dynamically adjusts the solar visibility range search radius and search step length by taking the terrain information entropy as an adaptive driving factor, so that the terrain shielding relationship is finely simulated in the mountainous or complex terrain area while the unobstruction is quickly determined in the flat terrain area. In addition, the application is realized by using a CUDA parallel architecture, and the GPU parallel strategy of block partitioning is combined, so that the calculation load of high-resolution data is effectively reduced, and the overall operation efficiency and memory utilization are improved.

The invention belongs to the technical field of scene matching, and particularly relates to a rapid scene matching method for a large-range search area, which greatly improves the scene matching speed during large-range search and solves the problem of high time consumption in initial positioning of current scene matching. According to the method, a four-dimensional descriptor is firstly designed, so that the robustness of a back-end matching strategy can be greatly improved; furthermore, on the basis of the multi-dimensional reference descriptor and the multi-dimensional real-time descriptor, a convolution coarse matching and fine matching strategy is designed, firstly, the multi-dimensional reference descriptor is segmented to perform software acceleration, coarse matching is performed on the obtained nine regions by using convolution, and fine matching is performed on the obtained nine regions; cutting an image with the same size as the real-time image on the original reference image according to the matching position for accurate matching, and finally designing a matching judgment rule to judge a matching result. According to the method, a mode of combining segmentation rough matching and fine matching is adopted for a large-range search region, so that the matching speed of the large-range region is greatly improved.

The invention belongs to the technical field of ocean information analysis, and particularly relates to a sound velocity estimation method for a high-resolution three-dimensional sound field in an ocean search and rescue area and application. Gridding an established search and rescue area, and filling corresponding grids with surface layer temperature and salinity data; according to profile values of temperature, salinity and depth of a search and rescue area, calculating surface sound velocity and re-sampling the surface sound velocity into the constructed grid; similarly, re-sampling the existing sound velocity data on each depth layer in the vertical direction to the nearest grid point; constructing a vertical-space joint feature data set; designing and training a neural network model, and performing lower-layer sound velocity data missing value reconstruction by using complete vertical structure information from a surface layer to an upper layer of a target layer; and when the next layer of data is reconstructed, the newly constructed layer is used as new upper layer data. According to the invention, high-resolution three-dimensional numerical value stratified rapid estimation is carried out on the sound field of the established ocean search and rescue area, high-precision data support is provided for search and rescue decision, and the method is suitable for large-range search and rescue sea areas.

The application relates to an unsupervised point cloud segmentation method and system, and relates to the technical field of target segmentation, and comprises the following steps: S1, acquiring point cloud data, estimating a neighborhood radius, and making the ratio of a first target eigenvalue and a second target eigenvalue not less than a preset value when a directionless normal of each point is extracted based on PCA; S2, acquiring a local representative subset through range search based on the estimated neighborhood radius; estimating a local tangent plane of each point in the local representative subset based on PCA, and extracting a directionless normal of each point based on the local tangent plane of each point; and S3, performing plane segmentation based on the directionless normal of each point. The neighborhood radius is estimated by calculating eigenvalues based on PCA analysis, so that the plane information in original data is fully utilized when the normal of each point is estimated. The unsupervised system does not need training data or manual intervention, and can obtain a point cloud segmentation result in a short time and at a low calculation cost.

The application provides a boundary layer height inversion method based on a dynamic height threshold, and relates to the technical field of boundary layer height inversion. The method comprises the following steps: acquiring a laser radar signal and preprocessing; identifying a cloud layer height according to a zero-crossing point of the signal based on a slope method; determining a height range searched when there is a cloud, adjusting an upper limit height according to the size of the signal in the range, and generating a limited range searched when there is a cloud; under the condition of a cloud, judging whether the cloud layer is in the boundary layer according to the slope size of the signal in the corresponding limited range, and in the two cases that the cloud layer is in the boundary layer and the cloud layer is out of the boundary layer, then recording the first target height or the second target height in the corresponding limited range as the boundary layer height according to the slope size. Through the method of dynamically adjusting the height threshold, the upper limit height threshold is adjusted in real time according to different situations, and the influence of complex stratification such as the cloud layer, the suspended aerosol layer and the residual layer is effectively eliminated.

The present application provides a content addressable memory (CAM) device, a content addressable memory cell and a data search comparison method thereof. The content addressable memory device includes a plurality of content addressable memory strings and an electrical characteristic detection circuit coupled to the content addressable memory strings. When performing a data search, a search data is compared to a stored data stored in the content addressable memory strings. The content addressable memory strings generate a plurality of memory string currents. The electrical characteristic detection circuit detects the memory string currents or detects a plurality of match line voltages of a plurality of match lines coupled to the content addressable memory strings to generate a plurality of search results. The stored data and the search data are a range stored data and a single bit search data, or the stored data and the search data are a single bit stored data and a range search data.

A method in a data processing system and apparatus for organizing electronic data, structured or unstructured, with associated metalabels, on a recordable medium of a data processing system. The data items are structured in a heterogeneous metalabel search structure. Searching the plurality of metalabel search structures can then be done with a heterogenous-range query including at least one term and one range query. The query can also be specified in a natural language which is converted to the specific heterogenous-range query. Also provided is a method for searching using geometric shapes that are mapped into range-string queries where the ranges are d-dimensional hyper-rectangles or cubes that cover the geometric shape. A method to aid the selection of the appropriate shape is provided as well as a method to cost-efficiently select the hyper-rectangles and cubes to cover the shape.

The invention relates to the technical field of computer security, and discloses a Boolean range search method supporting access mode and search mode privacy protection, comprising the following steps: a cloud server receives a search request from a client; the cloud server receives an encryption index constructed by a data owner; the plurality of cloud servers search on the stored encrypted index according to the received search request to respectively obtain part of search results; and sending the obtained partial search results to the client for summation to obtain a final search result. On the basis of effectively protecting the privacy of the access mode and the search mode, efficient Boolean range search can be realized, and meanwhile, in the process of dynamically updating the data, forward security is realized, and the security of previous search is not influenced by data updating.

The present invention relates to the technical field of model simulation, and relates to a simulation method, system, device, and medium for accurately hitting and superimposing damage effects of a gamma-ray weapon. The present invention uses a hierarchical bounding box algorithm to perform hit detection in stages, quickly screen out hit models, and accurately locate hit positions on the models, thereby improving detection efficiency. After determining the hit position, a radius range search is performed on discrete points in three-dimensional space corresponding to the model, and a temperature map and a time map are updated based on a mapping relationship between the discrete points and a UV map of the model, thereby achieving superposition of cooling and heating effects of the gamma-ray hit point and the model temperature within its influence range. During the entire simulation process, if an attacker emits a gamma-ray and the gamma-ray hits the position of an attacked party calculated by the attacker based on parameters, the hit behavior is confirmed and broadcasted in a simulation network, ultimately realizing a gamma-ray weapon simulation solution and providing ideas for judging simulation effects in a simulation network environment.

The invention relates to the technical field of driving cycles, in particular to a data processing method and system for typical driving cycle construction. The method comprises the following steps: determining a sampling moment state based on actual driving cycle data to obtain a Markov state transition matrix; combining a Monte Carlo algorithm to obtain a suboptimal driving cycle set; determining a reachable domain, and obtaining a candidate driving cycle set in the reachable domain through a Monte Carlo algorithm; carrying out resampling to obtain candidate sub-segments of the candidate driving cycle set; constructing a better driving cycle set, and selecting an iteration sub-segment of the better driving cycle set; in the reachable domain of the iteration sub-segment, carrying out random sampling through a Monte Carlo algorithm to obtain a candidate bridging path set; and replacing the iteration sub-segment with the candidate bridging path set, and carrying out repeated iteration to obtain an optimal driving cycle set. According to the method, large-range search, rapid generation and excellent feature maintenance can be synchronously realized.

The invention discloses a comprehensive management and control system and method for temporary power utilization, and the method comprises the steps: enabling all nodes to autonomously complete the position verification and the searching and negotiation of temporary power access points through the deployment of an edge calculation module, a Bluetooth module and a sound wave module at a power supply end node; direct communication among the nodes is realized through Bluetooth broadcasting, and millisecond-level quick response is obtained; a geo-fence is constructed through Bluetooth signals attached to each node, the constraint condition of a physical space is truly perceived, and wiring is convenient and fast; calculation tasks are dispersed to all nodes, a central control system does not need to process massive and high-concurrency real-time temporary power request calculation, and the nodes can be flexibly expanded according to needs; the natural ripple diffusion effect is simulated, small-range search is performed firstly, then the search range is expanded step by step, the nearest available node can be quickly positioned, and the line can be optimally utilized; and after the nodes are determined, the nodes nearby the nodes can automatically record related power supply information, so that the completeness and tampering resistance of the power supply records are ensured.

The application provides a multi-baseline phase unwrapping method based on accurate intercept correction, and relates to the technical field of radar signal processing, and comprises the following steps: acquiring parameter information corresponding to main and auxiliary images of an interferometric synthetic aperture radar in an observation scene and original interferometric phases, and constructing a reference intercept vector based on the parameter information and an external digital elevation model database; traversing actual intercepts corresponding to all pixels in the observation scene through a sliding window, obtaining a core intercept based on the frequency of the actual intercepts, correcting the core intercept by using the reference intercept vector, and obtaining a corrected intercept set; calculating a fuzzy number by using a limited range search method or a Chinese remainder theorem method based on the corrected intercept set and the number of baselines to obtain a fuzzy value; and calculating an unwrapped phase corresponding to each baseline by using a phase unwrapping equation based on the fuzzy value and the original interferometric phases. In this way, the practicability and reliability of the multi-baseline phase unwrapping method are significantly improved.

Methods and apparatus for sparse column-aware encodings for numeric data types, including integer data and floating-point data (float, double, etc.). The encoding schemes are tailored to take advantage of column addressable memories such as stochastic associative memories (SAM) to enable Stochastic Associative Search (SAS), which is a highly efficient and fast way of searching through a very large database of records (order of Billions) and finding similar records to a given query record (search key). Techniques are also disclosed for performing range searches for both integer and floating-point data types. The integer or float data is converted to Hexadecimal form and encoded using an m-of-n constant weight encoding. Only the columns with set bits in search keys need to be read, which significantly reduces the number of reads required for searches.