173results about How to "Increase calculation rate" patented technology

The invention provides a pollution source prediction method, a pollution source prediction device and electronic equipment, and relates to the technical field of pollution prediction, which comprisesthe following steps of: obtaining mesoscale meteorological prediction data and emission list data of a to-be-predicted destination; Wherein emissions and emission sources corresponding to the emissions are recorded in the emission list data; according to the Industry type, combining the emission sources to obtain reference emission data, inputting the mesoscale meteorological prediction data and the reference emission data into an air quality numerical model to obtain air pollution prediction data, and calculating the air pollution refined prediction contribution rate according to emission list data and the air pollution prediction data. The calculation amount of the air quality numerical model can be reduced, and the calculation speed is increased.

The invention discloses a sparse direct method-based monocular visual odometry (VO) method of a quadrotor unmanned-aerial-vehicle. The method is characterized by: carrying out depth estimation on a key frame, wherein feature points of the key frame are determined by a feature point method, an eigenmatrix between two adjacent frames is calculated, the eigenmatrix is decomposed and a rotation matrix and a translation matrix between the two adjacent frames are calculated to obtain an external parameter matrix, and then depths of the feature points are calculated according to a triangulation method; and after obtaining depth values of the feature points, obtaining the pose of the quadrotor unmanned-aerial-vehicle by solving through a sparse-matrix direct method, and carrying out motion estimation on all frames, wherein sparse feature points are extracted, the direct method is used to calculate a position of each feature point in the next frame, and grayscale information of each pixel point in pixel blocks which have a fixed size and are around the feature points is utilized to optimize grayscale differences between the two adjacent frames to obtain the motion pose of a camera. The method has the advantages that cumulative errors are avoided, higher accuracy is maintained for a long period, and the calculation amount can also be reduced.

The invention provides a method for assisting mobile unloading of users by multiple unmanned aerial vehicles, and belongs to the field of mobile cloud computing. The method comprises the steps of carrying out modeling on mobile migration of uploading and downloading requirements of multiple users from three aspects of unmanned aerial vehicle resource allocation, flight path planning and task allocation; and then converting and relaxing the model by a three-stage iterative algorithm, and performing iterative solution by using a branch-and-bound method to obtain an optical resource allocation, task allocation and flight path planning scheme. The mobile unloading method provided by the invention has the beneficial effects that the uploading demand and the downloading demand of the users are simultaneously taken into consideration, and the calculation rate of the users is greatly improved; and the mobile unloading method provided by the invention realizes the fairness of the users by maximizing the minimum calculation rate in all of the users.

The invention discloses a remote sensing image classification method of a double-branch fusion multi-scale attention neural network. The method comprises the following steps: reading a multispectral image from a data set; after an image matrix is obtained, preprocessing image data by using superpixels; performing normalization operation on the data, and taking a block from each pixel in the normalized image matrix to form an image block-based feature matrix; selecting a training set and a test set; constructing a classification model of the convolutional neural network based on two-channel sparse feature fusion; training the classification model by using the training data set; and utilizing the trained classification model to classify the test data set. According to the method, consideringfrom the characteristics of the image, the characteristics can be extracted by self-adapting to the size of the target region object in the image, and a new central pixel offset strategy is adopted for the boundary pixels, so that the classification accuracy of the boundary pixels is improved, and the operation speed of the whole training process is also improved.

A method for extracting a territorial sea baseline based on natural coastline data mainly comprises a natural coastline thinning method based on an improved Douglas-Peucker algorithm and a territorial sea baseline encryption algorithm based on earth spherop geodesic line resolving, and territorial area baseline delimitation is conducted. According to the method for extracting the territorial sea baseline based on the natural coastline data, in the thinning process, natural coastline salient points which have a larger influence on a sea area are extracted and taken as segment points, then two adjacent segment points are taken as a first point and a tail point for data thinning, an optimal distance threshold is selected through a fitting curve based on a least square method, thinning is conducted segment by segment, and the natural maritime rights and interests can be guaranteed under the circumstance that higher precision and compression ratio are guaranteed. The encryption algorithm is based on the earth spherop geodesic line resolving on the ground, a geodesic line is fitted between adjacent territorial sea base points, and therefore encryption calculation is conducted on adjacent base points by using a geodetic theme resolving principle. Improvement is conducted on distance judgment conditions and encryption point positions in the calculation process, the calculation rate is increased, and the position distribution of the territorial sea base points is more reasonable.

The invention provides a rapid high-precision gravity field forward-modeling method in the spherical coordinate system. An underground field source is divided into ruled unit bodies at equal interval,whether one unit body is dissected into a proper size is determined by determining relation between the distance from an observation point to the central point of the unit body and lengths of the unit body in the radial, broadwise and warpwise directions, gravity response of each unit body in the observation point P beyond the field source body is calculated, the gravity potential V, the gravityacceleration galpha and the gravity gradient tensor galphabeta are represented in a matrix multiplying form, a calculation process can be simplified by utilizing the equivalence of a kernel matrix, kernel matrix elements can be obtained rapidly from one kernel matrix element, the gravity response of the multiple unit bodies is accumulated according to the superposition performance of the gravity field, and a gravity field generated in the point P by all substances underground is obtained. Thus, less fewer is occupied by introducing a kernel matrix equivalence strategy and an adaptive fine-dissection strategy, and the calculation efficiency and forward-modeling precision are improved greatly.

The invention provides a processor applied to a sparse neural network. The processor includes a storage unit used for storing data and instructions, a control unit used for getting the instructions stored in the storage unit and sending out control signals, and a calculation unit used for getting the node values in a layer of a neural network and corresponding weight value data from the storage unit in order to get the node values in a next layer. When any of elements to be calculated is equal to zero, the calculation unit does not implement the multiplication of the calculation element, wherein the elements to be calculated include node values and weight values. By using the processor of the invention, the computing speed of the neural network can be improved, and power can be saved.

A melt and concrete interaction analysis method based on a particle method mainly comprises the following steps that 1, conducting particle modeling, and setting initial arrangement and parameters ofparticles; 2, establishing and dividing a background grid according to nodes, and retrieving neighbor particles; 3, calculating the enthalpy value, temperature and phase state of the particles; 4, calculating eutectic and chemical reactions, updating particle matter components and physical property parameters, meanwhile, calculating chemical heat and updating the particle enthalpy value, temperature and phase state; 6, calculating gas motion and updating the particle type, speed and position; 7, displaying and calculating a viscosity item, a surface tension item and a gravity item in the momentum equation, and estimating the particle speed and position; 8, displaying and calculating a pressure item in the momentum equation by using the estimated particle position, and correcting the particle speed and position; 9, outputting a calculation result. All phenomena of interaction between the melt and the concrete are considered. Based on a particle method, the free liquid level can be accurately captured and modeling and accurate phase change processing are facilitated and large-scale calculation is facilitated by adopting the display pressure model.

The invention discloses a line obstacle monitoring and alarming system and method based on three-dimensional imaging, and belongs to the technical field of railway safety. The line obstacle monitoringand alarming method comprises a step of judging the type of an obstacle: extracting an obstacle contour information point cloud cluster in a three-dimensional point cloud image, and judging whether the obstacles are the same obstacle or not; if the obstacle is the same obstacle, calculating the volume, angle and position of the obstacle according to the characteristics of the obstacle contour information point cloud cluster, judging whether the obstacle is in a motion state or not according to the position information of the obstacle, and calculating the moving speed of the obstacle if the obstacle is in the motion state; and performing feature matching with the point cloud sample library according to the volume, shape and moving track data of the obstacle contour information point cloudcluster to realize obstacle classification. According to the line obstacle monitoring and alarming method, the property and the residence time of the obstacle can be accurately judged, and graded alarming is carried out according to the type and the moving speed of the obstacle, and the railway safety is ensured.

The invention belongs to the technical field of computers. The invention provides an FPGA-based heterogeneous neural network calculation accelerator design method, which is suitable for large-scale deep neural network algorithm acceleration. The method comprises the following steps that: a CPU reads related parameters of a neural network, and dynamically configures an external memory and a convolution calculation unit according to obtained information; an external memory stores parameters and input data which need to be loaded into corresponding positions of an input cache through a bus; the parameters are alternately loaded into the two convolution calculation units respectively, the other convolution calculation unit performs calculation while the parameters are loaded into one convolution calculation unit, and circularly alternating is conducted until all operations of the whole convolution neural network are completed; and final output results are stored in an output cache to waitfor an external memory to access. According to the method, the FPGA is used for combining the convolutional neural network calculation units, so a computing rate of a platform can be increased while resources are saved.

The invention discloses a multi-DC-feed AC grid fault calculating method taking DC control characteristics into consideration. The multi-DC-feed AC grid fault calculating method comprises the following steps of inputting system parameters, DC control strategies and fault parameters of an MODC (multi-outfeed direct current) system; forming an AC system sequence network chart through the system parameters and the fault parameters of an AC grid, establishing an AC constraint equation, and establishing a DC constraint equation between the DC voltage and the DC current of a DC side through the DC control characteristics; setting initial values for all variables; establishing converter switch function models SI and SU of every DC subsystem; according to the AC constraint equation, the DC constraint equation and voltage and current switch functions, solving the transient state parameters of the AC side and the DC side of the MODC system; performing iterative computation, and when variable errors meet computational requirements, outputting the computation results of the transient state parameters of the AC side and the DC side of the MODC system. The multi-DC-feed AC grid fault calculating method taking the DC control characteristics into consideration is high in computation speed, accuracy and adaptability and has high engineering practical value in the aspect of setting and cooperation of fault computation and relay protection of multi-DC-outfeed AC grids.

The invention discloses a bearing capacity design method for a prestressed concrete structure. The existing design method for a statically indeterminate prestressed concrete structure has the problems of the influence on the transfer of the prestressed concrete structure and on design calculation results caused by lateral confinement structures, such as a post, a wall and a cylinder. The bearing capacity design method for the prestressed concrete structure comprises the following steps of distinguishing the concrete structure of an unconfined lateral structure from a concrete structure of a confined structure, and introducing a lateral confinement influence coefficient; calculating the cross-sectional area of a longitudinal prestressed tendon in a tensioned area; determining the sum of a bending moment value caused by tensioning the prestressed tendon and the designed bending moment value of a control cross section external load; establishing effective prestress tensioned to the prestressed tendon; passively providing resisting force by the part, higher than the effective prestress, of the designed tensile strength value of the prestressed tendon; calculating normal section bearing capacity of a prestressed concrete flexural component with a rectangular section of the unconfined lateral structure; introducing the lateral confinement influence coefficient for modifying to obtain a calculation formula for the normal section bearing capacity of the prestressed concrete flexural component with the rectangular section of the confined lateral structure. The bearing capacity design method is applied to the bearing capacity design of the prestressed concrete structure.

A multimode medical image registration method based on the SURF algorithm comprises the following steps: S1, employing the SURF algorithm to extract the feature points in different mode medical images; S2, determining the main directions of the feature points in each mode medical image, and constructing corresponding SURF feature descriptors; S3, performing matrix operation of the SURF feature descriptors of the feature points, and obtaining a transition matrix; and S4, employing the transition matrix to complete the registration of different mode medical images. The SURF algorithm compares to the SIFI algorithm, the calculation rate is greatly improved. The parallelization technology such as GPU and the like is employed to reach real-time linear registration of a big-resolution images.

The invention discloses a parallelization optimization method for an SM3 cryptographic hash algorithm. The SM3 cryptographic hash algorithm comprises an SM3 message extension function and an SM3 message compression function; a multi-channel parallel algorithm is adopted, cyclic operation in the SM3 cryptographic hash algorithm is unfolded, and multiple rounds of iteration are combined, so that theoperation required by the SM3 cryptographic hash algorithm is simplified; a potential parallel operation vector in the extended SM3 cryptographic hash algorithm is realized so as to realize internalparallelization of the SM3 cryptographic hash function; the parallelization optimization method can run on a computing platform supporting simultaneous execution of a vector instruction and a non-vector instruction. According to the method in the invention, the internal parallelization of the SM3 cryptographic hash function is realized, and the operation speed is further increased; and when the method is applied to the computing platform capable of executing the vector instruction and the non-vector instruction at the same time, the computing speed is higher.

The invention discloses a method for ranking turbine rotor blades based on Monte Carlo tree search method for ranking turbine rotor blades based on Monte Carlo tree search, belonging to the mechanicalassembly field. Firstly, the leaf parameters are extracted. Then, the root node of Monte Carlo tree is initialized, and different leaf sequences are generated by selecting nodes, expanding nodes andsimulating nodes for many times, and the unbalance quantity is calculated, and the weight value in Monte Carlo tree is updated. Secondly, the node with the largest weight and the corresponding optimalsequence are taken as output, the root node of the Monte Carlo tree is updated again, a new Monte Carlo tree is created, and the above process is repeated. Finally, the optimal unbalance quantity andits corresponding blade sequence are obtained by searching multiple Monte Carlo trees, and the rotor blade scheduling is optimized. The invention can further optimize the blade scheduling problem while ensuring good convergence, and the algorithm has good stability.

The invention discloses a grating subdividing and direction judging method. A microprocessor is adopted for grating subdividing and direction judging, related square signals are output by a grating sensor, the grating sensor respectively outputs the square signals A and the square signals B, the phase difference of the square signals A and the square signals B is 90 degrees, the square signals A and the square signals B are connected with two interrupt input ports of the microprocessor respectively, and rising edge interrupt and falling edge interrupt are arranged in the microprocessor set with an interrupt service program. When the square signals A enter an interrupt state, the interrupt service program reads the state of the square signals B and judges the xor value of state values of the square signals A and the square signals B; when the square signals B enter an interrupt state, the interrupt service program reads the state of the square signals A and judges the xor value of the state values of the square signals A and the square signals B. According to the grating subdividing and direction judging method, the output square signals A and the square signals B are subdivided into four parts in the mode of double interrupt (rising edge triggering and falling edge triggering), the processing and calculating time of the interrupt service program is greatly shortened, calculation efficiency is improved, and the grating subdividing and direction judging method is high in real-time performance, good in stability and high in interference-resisting capacity.

The invention discloses a rapid prediction method for heat conduction-radiation coupling heat transfer characteristics of a periodic structure composite material at a high temperature, and the method comprises the steps: decomposing a to-be-solved temperature field into a macroscopic average field and a mesoscopic temperature fluctuation through a multi-scale progressive analysis method, respectively carrying out the calculation, and finally reconstructing the macroscopic average field and the mesoscopic temperature fluctuation into a complete temperature field. The calculation process comprises the steps of carrying out grid division on a macroscopic prediction model and a characterization unit under a microscopic scale, solving a periodic vector function in the characterization unit under the microscopic scale, calculating macroscopic equivalent physical property parameters, solving a macroscopic scale heat conduction-radiation coupling heat transfer equation, and finally reconstructing a multi-scale temperature field. The multi-scale model established by the method can accurately calculate the temperature field of the periodic structure composite material, and can significantly improve the prediction rate of the high-temperature heat transfer characteristic of the composite material.

The invention relates to an electric energy data acquiring and processing system based on IEC62056. The electric energy data acquiring and processing system comprises a main station server, at least one electric energy data acquiring and processing terminal and at least one electric energy metering terminal, wherein each electric energy data acquiring and processing terminal comprises an acquiring module, a communication module, an interface module and a calculating module, the acquiring module is used for data interaction with the electric energy metering terminal via the IEC62056 so as to acquire data information outputted by the electric energy metering terminal; the communication module is connected with the acquiring module and used for providing real-time communication services; the communication module adopts the IEC62056 protocol; the interface module is connected with the communication module and used for receiving data information outputted by the communication module and outputting the same to the outside, and the calculating module is connected with the interface module and used for calculating and outputting the data information outputted by the interface module to the outside. The electric energy data acquiring and processing system based on the IEC62056 has the advantages of high expansibility and data processing capacity.

The embodiment of the invention provides a CRC computing method and device. The method comprises the steps that target input data is acquired; the target input data is sequentially divided into n target data sequences with the same sequence length; n processors are utilized to perform cyclic redundancy check (CRC) computing on the n target data sequences to obtain corresponding n first CRC values; target sequence influence coefficients corresponding to all the first CRC values are acquired; Galois domain multiplication is performed on all the first CRC values and the corresponding target sequence influence coefficients to obtain n second CRC values; and XOR operation is performed on the n second CRC values to obtain a target CRC value to realize CRC computing on the target input data. Through the embodiment, the number of times of XOR operation in the CRC computing process can be reduced, so that the computing rate is increased; and besides, the number of times of thread synchronization is reduced, so that the computing rate is further increased.