151results about How to "Reduce double counting" patented technology

The invention relates to the technical field of computers, in particular to an information pushing method and device based on a graph network. The method comprises the steps of determining a sub-graphstructure taking the to-be-processed node as a central node in a graph network, the sub-graph structure comprising nodes connected with the to-be-processed node and a connection relationship; inputting the subgraph structure into a node feature extraction model to obtain an aggregation feature vector corresponding to a to-be-processed node output by the node feature extraction model; screening nodes matched with the to-be-processed nodes as information pushing nodes based on the aggregation feature vectors of the nodes in the graph network; and pushing the service information corresponding tothe information pushing node, so that the graph structure feature vector is added on the basis of the aggregation feature vector during training, namely, the graph structure information is added forconstraint, the convergence speed and accuracy of training are improved, and the sub-graph structure can be input during application. And the node representation accuracy and the information pushing accuracy in the graph network are improved.

The invention discloses a high-performance electrical network real-time topology analysis method. The objective of the invention is to solve technical problems of high complex degree, poor scalability and low real-time performance of an electrical network topology analysis method in the prior art. According to the method of the invention, electrical network topology analysis is established based on specific switch actions; the switch actions of an electrical system are monitored; and calculation results of an adjacency matrix and a full-connectivity matrix before the action of the switch are fully utilized, so that a new adjacency matrix and a new full-connectivity matrix can be obtained; and calculated data can be utilized efficiently, and a large quantity of repeat calculation processes can be decreased. The high-performance electrical network real-time topology analysis method of the invention can be widely applied to topology analysis and topology coloring of various kinds of automatic monitoring fields involving complex circuits.

The invention discloses a spectrum resource allocation method for a wireless return link heterogeneous internet of things. The method comprises the following steps: 1) dividing user terminals into two groups, wherein one group is directly connected to a macro cell base station and referred to as macro cell users, the other group is connected to a micro cell access point and referred to as micro cell users, and the micro cell access point is connected with the macro cell base station via a wireless return link; 2) respectively providing, by the macro cell base station and the micro cell access point, a communication service for the macro cell users and the micro cell users; and 3) calculating, by the macro cell base station, an optimal allocation coefficient eta* using statistical channel state information, and allocating a spectrum resource according to the eta* value, wherein the wireless return link occupies eta*, and the remaining spectrum resource is used for a wireless access network. Under the condition that the spectrum resource is limited, the principle of maximum total rate of a system is considered, and the spectrum resource is allocated to the wireless access network and the wireless return link, so that the capacity of the system is improved and then the spectral efficiency is improved.

The invention provides a method and a system for image noise reduction. According to the method, a boundary threshold of a high frequency area and a low frequency area of an image is determined, whether pixel points of the image belong to the high frequency area or the low frequency area is determined respectively, the pixel points in the high frequency area or the low frequency area of the image can be independently processed according to demands, problems of serious loss of detail portions of the image and the fuzzy effect of the image after noise reduction existing in an image noise reduction method in the prior art can be solved, relatively strong adaptability is realized, and quality of the image after noise reduction is improved, moreover, a shift summation method taking columns as units is employed to calculate sum of gray values of all pixel points in each window, calculation repetition is reduced, the calculation process is simple, and the calculation speed is accelerated.

The application aims at providing a method and device for distributed diagram calculation. Compared with the prior art, the method provided by the invention comprises the following steps: firstly acquiring original diagram data, and processing the original diagram data according to a diagram algorithm to acquire structured diagram data corresponding to the diagram algorithm, thereby adapting to the diagram algorithms in different types; and then distributing the calculation tasks corresponding to the diagram algorithms to multiple calculation nodes to be executed, wherein a persistence condition should be satisfied in the execution process, the persistence operation is performed, the data dependence is cutoff, the repeated calculation amount is reduced, and the processing efficiency is improved. Furthermore, the application comprises firstly performing a combination operation on the diagram data before an aggregation operation and a connection operation, thereby improving the operation efficiency and relieving the network transmission pressure. Furthermore, the application adopts a data serialization and deserialization method to facilitate the transmission of intermediate data produced in the calculation process between the calculation nodes.

The invention belongs to the technical field of power grid dispatching operation and relay protection, and particularly relates to a power distribution network line protection constant value checkingalgorithm. The power distribution network line protection constant value checking algorithm includes: setting a node of allowing new energy to access a power distribution network and a corresponding parameter; determining the number of protections which are affected by new energy access and need to perform constant value checking according to the features of a looped network structure and open loop operation of the power distribution network; manually inputting or online-acquiring the output power of a new energy power supply in an EMS system, and setting the output power as a short-circuit calculation initial value; encoding a fault calculation result according to the number of the protections to be checked and branches where the protections are located; performing constant value checkingon a phase current protection, a zero sequence current protection, a phase distance protection, and a grounding distance protection, and determining the selectivity and the sensitivity of the constant values; and finally outputting the checking result. The objective of the invention is to achieve protection constant value adaptation checking according to the features of the power distribution network after mass new energy is accessed currently.

The invention discloses a H.264 standard-based deblocking filter boundary strength computing device and method, which are used for computing boundary strength of boundary points during H.264 deblocking filtering. According to the invention, the boundary point with same modes are pre-judged according to a coding mode of current macroblocks and a coding mode of adjacent macroblocks and according tothe condition that boundaries of the macroblocks exist, and can be directly assigned. For the boundaries which can not be directly assigned, logic operation and arithmetic operation are adopted for replacing skip judgment when the boundary strength is computed. The invention has the advantages of being capable of remarkably reducing a large quantity of repetitive computation and complex conditionjudgment of a deblocking filtering link of the H.264 standard and increasing program execution efficiency, and is especially suitable for embedded processors such as a DSP (Digital Signal Processor).

The invention provides a solid rocket engine performance progressive matching design method. The solid rocket engine performance progressive matching design method comprises the steps of acquiring task objectives of total indexes of a solid rocket engine; determining a time interval division number according to the working time of a preset thrust demand in the task target; obtaining an initial training sample set of the preset thrust demand matching degree based on an initial balance pressure method; according to the initial training sample set, operating the combustion surface and internal ballistic trajectory calculation model in the initial time interval to obtain N thrust curves, and constructing a thrust approximation model according to the N thrust curves; carrying out inaccurate sampling to obtain alternative solutions; substituting the alternative solution into a combustion surface calculation and internal trajectory calculation model, and calculating the performance of the internal trajectory within simulation time to obtain an alternative thrust curve; when the simulation time is not less than the working time, adding the alternative solution as a new sample point into asample set to obtain an updated sample set; and when a preset convergence condition is reached, outputting the alternative thrust curve as an optimal solution.

The invention discloses a Hadoop platform time series data incremental computation method and system. The method includes the steps that when a time series data incremental computation task is started, the historical computational state of time series data is obtained from a cache server; incremental computation is carried out by means of a segmented time series data incremental computation method containing a SubCp sub-operation and a ReduceCp sub-operation according to the historical computational state, wherein the SubCp sub-operation is used for self-definition of segmented time series data and storage of intermediate results, the ReduceCp sub-operation is carried out in an operation merging stage and used for merging operation on computed results of the segmented time series data according to the self-defined operation, and the computational state of the SubCp sub-operation and the computational state of the ReduceCp sub-operation are maintained through the cache server. By the adoption of the method and system, plenty of unnecessary repetitive computation can be saved through incremental computation, and therefore data processing efficiency is improved.

This invention discloses a method for acquiring paths with protection entities in the intelligent optical network including the following steps: setting up different TE links for data links of different protection kinds, setting up a same TE link for the data links having the same protection attribute and belonging to the same protect entity, carrying out flooding of the protection entity information to diffuse the protection entity information in the entire route zone and acquiring paths with protection entities in the network according to the information.

The invention discloses a sample ingredient content measuring method based on an online sequential limit learning machine. The method comprises the following steps that: collecting the spectroscopic data sample of a sample, and utilizing an online sequential limit learning machine algorithm for modeling; and utilizing an established model to measure the ingredients of the sample. The online sequential limit learning machine algorithm is used for modeling, data which is previously used can not be kept, and only knowledge which is learnt previously is kept for standby; and when new spectroscopicdata arrives each time, the implicit strata output of new data needs to be calculated, and then, the knowledge which is learnt previously is used for dynamically updating output weight between a middle implicit strata and the output layer to carry out quick modeling. Compared with a traditional modeling method, the method disclosed by the invention improves modeling speed, reduces unnecessary repeated calculation amounts and consumption for a data storage space and improves the accuracy and the generalization performance of the model, in addition, data which arrives one by one can be processed, and data which can arrives by one block can be processed.

The invention discloses an image copy and paste tampering detection method based on two-dimensional sliding window DFT fast calculation. The features of a tampered image are extracted by employing a multipoint sliding window DFT (Discrete Fourier Transform) fast algorithm, tampered pixel positions are detected, and a feature extraction speed and the detection quality are analyzed. By adoption of the method, a detection method combining a multipoint sliding window DFT algorithm with a single-point sliding window DFT algorithm is provided to achieve better balance between the detection speed and the detection quality. Under normal circumstances, in order to ensure the integrity of the image, tampering mainly focuses on local variation, and otherwise tampering traces can be found easily via intuitive vision. When the proportion of the tampered part of the image in the entire image is not large, the calculated amount of the improved detection method provided by the invention has relatively high computational efficiency.

The present invention discloses a vehicle body color recognition region positioning method and device. The method includes the following steps that: the edge texture image of the vehicle head region of a vehicle body is acquired, and the integral image of the edge texture image is calculated; the integral image is divided into a plurality of local regions of first preset area, wherein the local regions do not intersect with each other; the unit area edge densities of the local regions are calculated, a first preset number of local regions are selected according to the unit area edge densities so as to be adopted as first regions to be recognized; local sub-regions of second preset area are selected from each first region to be recognized so as to be adopted as second regions to be recognized, wherein the local sub-regions have the smallest unit area edge density; and a second preset number of regions are selected from the second regions to be recognized according to the unit area edge densities of the local regions and the local sub-regions so as to be adopted as vehicle body color recognition regions. Since the regions with smooth textures are selected according to the unit area edge densities, and then the vehicle body color recognition regions are selected, and therefore, external interference resistance is excellent.

The invention relates to an animation file processing method and apparatus, a computer readable storage medium and computer equipment. The method comprises steps of obtaining a bitmap image sequence corresponding to an original animation file; dividing each bitmap image in the bitmap image sequence into a color channel bitmap and a transparency channel bitmap; synthesizing the color channel bitmapand the transparency channel bitmap to obtain a synthesized bitmap; encoding the synthesized bitmap according to a video encoding mode to obtain an encoded picture corresponding to the bitmap image;and generating an animation export file corresponding to the original animation file according to the coded picture corresponding to each bitmap image in the bitmap image sequence. According to the method, decoding efficiency of the obtained animation export file is high.

The invention relates to a river regulation water demand determination method based on ecological water demand calculation. When the water demand is determined according to the determination method, the ecological target of a regulation project is clearly put forward by combining a regulation project target according to natural and hydrological conditions of a regulated river section, and the river water demand is divided into flow type water demand and consumption type water demand according to different properties of the river water demand. According to the method, the target demand of the river regulation project is taken into consideration in calculation, and the repeated calculation process is reduced, thereby being capable of providing an accurate water volume basis for planning an implementation scheme and a later-period water supplementing scheme for the project, and being capable performing simple calculation and determination on the river water demand for ensuring the river regulation effect in allusion to different hydrological conditions and different regulation targets.

The invention discloses a multi-starting-point route navigation method. The method comprises the following steps of (a) transmitting a route navigation request to a server through a mobile terminal, and calculating a first navigation route by adopting the current position of a vehicle as a starting point through the server after receiving the route navigation request; (b) if fork roads exist in the first navigation route, calculating time for the vehicle to arrive at the next fork road through the server; (c) calculating multiple second navigation routes by adopting all sideways of a fork road as starting points if the estimated time for the vehicle to arrive the next fork road is smaller than a preset threshold value; and (d) transmitting multiple navigation routes of different starting points to the mobile terminal. According to the multi-starting-point route navigation method, whether a sideway exists in front of the navigation route is preliminarily judged and the time for the vehicle to arrive at the sideway is estimated, and multiple navigation routes with different starting points and identical destination are transmitted to the mobile terminal to be selected, so that the probability for repeatedly calculating the navigation route can be greatly reduced, and the navigation efficiency is improved.

The invention discloses an inverse discrete cosine transform (IDCT) lookup fast algorithm with variable complexity, which mainly aims at reducing the redundant operation in the original acceleration IDCT algorithm and accelerating the computation speed of an IDCT module. The algorithm provided by the invention comprises the steps of decomposing an inverse quantized DCT (discrete cosine transform) coefficient matrix F into polynomial matrixes to add; performing inverse discrete cosine transform for each decomposed matrix; and finally, adding the inverse discrete cosine transformed matrixes to obtain an original matrix f. The algorithm disclosed by the invention reduces the redundant operation in the original algorithm and effectively accelerates the computation speed of the inverse discrete cosine transform.

The invention provides a rapid simulation method of a synthetic aperture sonar signal based on a time delay table, comprising the following steps: (1) setting each simulation parameter of a synthetic aperture sonar system; (2) constructing a height model of a target scene needed to be simulated; (3) projecting the height information of each target point in an oblique distance direction; (4) counting the number of target points in all directions and distances; (5) calculating the delayed time of each ping of the target point corresponding to one synthetic aperture length according to the distance and constructing the time delay table; (6) calculating the basic matrix positions of different pings according to the given speed, performing time delay superposition to a sending signal accordingto the value of the time delay table, and finally obtaining an original echoed signal of the synthetic aperture sonar. The method can improve the computational efficiency of synthetic aperture sonar simulation signals and effectively perform echoed signal simulation of synthetic aperture sonar under the condition with more target points, in particular under the condition with a large-scale landform scene.

The invention discloses a massive target performance data fast reading and precise interpolation method. The method includes that 1 target double-station radar cross sections are calculated, and first-order and second-order derivative information is required to be stored in the radar cross sections at the frequency end point positions simultaneously; 2 a useful double-station radar cross section range is determined and read according to the target state; 3 interpolation of an incidence pitch angle, an incidence azimuth angle, a scattering pitch angle and a scattering azimuth angle is conducted respectively; 4 data format organizing is conducted on the double-station radar cross sections; 5 double-station data after being organized is sent to a graphics processing unit, and frequency one-dimensional interpolation is conducted; 6 whether the interpolation of the incidence pitch angle, the incidence azimuth angle, the scattering pitch angle and the scattering azimuth angle is conducted is judged; 7 interpolation data generated by the graphics processing unit is sent back to a central processor. By means of a lagrange interpolation method, fast reading of massive target performance data is achieved, large-scale reading of the data is avoided, interpolation precision is improved, and repeated calculation is saved.

The embodiment of the invention provides a human body contour point extraction method and device, a terminal device and a computer readable storage medium, and relates to the field of data processing.The method comprises the following steps of extracting the human skeleton key points contained in an image; converting every two adjacent human skeleton key points in the human skeleton key points into a corresponding skeleton matrix; and inputting the image and each bone matrix into a contour point extraction model, and extracting the human body contour points to obtain the human body contour points contained in the image. The input human skeleton matrix reduces the extraction range of human body contour points, the contour point extraction model structure can be simplified, the calculationamount is reduced, and the speed is increased. A human skeleton matrix is used as partial input, and a general region where human contour points appear is limited, so that the accuracy is improved.

The invention provides a method for identifying user name abbreviations. The method includes the following steps: (1) filtering characters in two or more user names with English letters and numbers kept; (2) segmenting the filtered user names into a plurality of continuous fragments, and selecting the initial character of each fragment to form a new character string; (3) acquiring the length of the longest abbreviation according to the new character string, and determining that user name abbreviations exist among user names if the length value is larger than or equal to a give threshold deltaL; converting the English letters to lower-case or upper-case forms in a unified manner; the fragments are words or single characters; the fragments are obtained by segmentation according to a specified dictionary; the length of the longest abbreviation is obtained according to the new character string via a dynamic programming algorithm.

The invention discloses an interference analysis method among DAG real-time tasks based on division scheduling on a multi-core processor, and relates to the technical field of embedded real-time systems. The invention aims to improve the precision of WCRT analysis so as to improve the accuracy of predicting the schedulable rate of a task set by a real-time system. According to the method, the influence of the internal structure of the high-priority DAG task on the interference of the low-priority task is analyzed, and a more accurate interference upper bound is obtained. For the topological structure of the DAG task and the condition that the internal sub-tasks are allocated to the processor, a parallel structure is provided firstly, and it is proved that the parallel structure can cause repeated calculation of inter-task interference generated by a traditional analysis method. According to the parallel structure, a WCRT analysis strategy for reducing repeated calculation is provided,and the precision of WCRT analysis is effectively improved. The WCRT analysis strategy can reduce the WCRT numerical value of the real-time system prediction task set, so that the WCRT numerical valueis closer to the real value of the task set executed in the system, and the accuracy of the real-time system prediction task set schedulable rate is improved. The method is applied to the isomorphicmulti-core processor.

The invention discloses a corpus annotation method and device. The method comprises the following steps: detecting a text selection operation on a sentence to be labeled; After the text selection operation is finished to obtain a selection text, determining a label corresponding to the selection text; And displaying a label corresponding to the selection text at a position outside the node of theto-be-labeled sentence, and storing the position information of the to-be-labeled sentence labeled with the label corresponding to the selection text and the position information of the selection textin the to-be-labeled sentence in a preset database. The technical problem that in the prior art, when corpus annotation is carried out, annotation results are inserted into sentences or displayed atthe tails of the sentences, and the sentence positions are repeatedly calculated during multiple annotations is solved.

The invention discloses an industrial operating system data lake construction method based on a data warehouse, which belongs to the technical field of industrial enterprise management, and comprises the following steps: respectively constructing a data storage layer, a data management layer and a data analysis layer in a data lake architecture, the data analysis layer is divided into a source pasting layer, a uniform data warehouse layer and an application data layer, and the data analysis layer is constructed by adopting a real-time analysis component, an offline analysis component and an inner layer analysis component. According to the industrial operation system data lake construction method based on the data warehouse, equipment state data, operation data and equipment asset data are acquired through an equipment access tool, production data, processing data and supply and demand data of a production and manufacturing system are acquired by adopting a system intervention tool, and files and model data are acquired by applying an intelligent filling tool. Unified storage, analysis and management are carried out on the data, and a data basis is provided for rapid intervention of industrial resources and agile development of applications.

The invention discloses a spectrum matching method and system based on full-wave band feature enhancement. The spectrum matching method comprises the following steps: acquiring a reflection spectrum of a sample by using a high-spectrum technology; calculating a reflectivity variance in a segment of spectrum range; defining the reflectivity variance as fluctuation intensity; building a fluctuation spectrum; calculating a similar distance of the fluctuation spectrum to realize spectrum matching. By adopting the method, spectrum fluctuation is analyzed on a wavelength dimension, transformation from the reflection spectrum to the fluctuation spectrum is built, a spectrum diagnosis feature is extracted and amplified, spectrum redundancy information is restrained, the spectrum separability is further enhanced, and the performance of a spectrum matching algorithm is improved effectively. By adopting a high-spectrum feature enhancement technology based on spectrum fluctuation, a high-spectrum imaging technology is further improved on recognition of ground objects, particularly minerals with absorbing-reflecting features.

The present invention discloses a network security detection method and detection system. The method includes the following steps that: 1) detection strategies used for detecting attack confidence-theft and leakage behaviors are received; 2) the detection strategies are executed according to the priority sequence of the detection strategies to detect messages to be matched, so that the attack confidence-theft and leakage behaviors can be obtained; and 3) alarm information is generated according to the obtained attack confidence-theft and leakage behaviors, and is reported to a management center. With the network security detection method and detection system of the invention adopted, the detection effect of the attack confidence-theft and leakage behaviors can be improved, and new types of attack confidence-theft and leakage behaviors can be detected out.

The invention discloses a video image super-resolution reconstruction method based on time domain correlation. The method comprises the steps: determining whether a reconstruction result of a preorderframe is used or not through obtaining the space-time correlation feature information of a video frame, so as to reduce the unnecessary repeated calculation; meanwhile, the reconstruction process ofthe to-be-reconstructed frame is guided and the quality is enhanced in combination with the inter-frame difference and the reconstruction condition of the preorder frame. And performing super-resolution reconstruction on the low-resolution video sequence by using a deep learning technology to obtain a high-resolution video sequence, multiplexing a reconstruction result of the preorder frame sequence, and optimizing calculation resource configuration while ensuring the reconstruction quality.

The invention discloses a multi-view stereo vision method integrated with spatial propagation and pixel-level optimization, which belongs to the field of computer vision. The multi-view stereo visionmethod comprises the following steps: camera parameters are obtained according to a plurality of pictures of a scene, a depth image of each picture is extracted according to the camera parameters, andspatial propagation is utilized to update the depth image of each picture; pixel-level optimization and filtration are iteratively carried out on the updated depth images, so that processed picturesare obtained; and depth image fusion is concurrently carried out on each pixel in the processed pictures, so that a three-dimensional point cloud of the scene is obtained. In initial depth image extraction, the multi-view stereo vision method utilizes spatial propagation to reduce redundant computation between images, thus increasing efficiency; in addition, by iterative depth image optimization and filtration, while noise points are reduced, the integrity of the depth images is increased, and ultimately, while the efficiency is increased by highly efficient depth image fusion, the density ofthe point cloud is ensured.

The invention discloses a big data based efficient search method. The method comprises the steps of determining whether a case for the querying or partial querying is in the historical record according to the new querying requirement from a user; if so, sharing the result set of the historical querying. The method can solve the problem of small search efficiency of the big data, reduce a large number of repeated calculation for the historical querying, save the search time and increase the search efficiency.