40results about How to "Strong parallel computing ability" patented technology

The invention discloses a lightweight quick response aerial remote sensing inertial-stabilized platform controlling system used for carrying out initiative controlling upon triaxial frames of stabilized platforms. The system comprises a DSP module, an FPGA module, a PWM power driving module, brushless DC torque motors, a stabilized platform frame, a measuring sensor, and an interface circuit. Angular position information, angular rate information and attitude information of the frame output by the measuring sensor are acquired by the DSP module and the FPGA module through the interface circuit in real time. PID controlling algorithm is executed by the DSP module and the FPGA module, and 18 PWM signals for controlling the three brushless DC torque motors are generated. The signals are converted into voltage signals by the PWM power driving module, and the brushless DC torque motors are driven, and the frame is driven to rotate, such that the initiative controlling upon the triaxial frame is realized. According to the invention, an independent mode and a combined mode are realized, and a frame soft lock function and an independent course configuration function under the combined mode are realized. Therefore, the operation of the platform is more reliable and intelligent.

The invention relates to a memory data warehouse query processing implementation method of a database oriented integrated machine. The steps of the memory data warehouse query processing implementation method comprise constructing a memory data ware house storage model; constructing a memory data warehouse integrated machine distributed storage model; adopting data updating strategy for a high performance computing server cluster: When the high performance computing server cluster having insufficient memory capacity, adopting a round-robin queue update strategy to eliminate the oldest data and to update to the latest data; and realizing OLAP query processing of the memory data warehouse integrated machine. The memory data warehouse query processing implementation method can improve the memory database integrated machine asymmetric storage and computing resource utilization, improve overall, performance of memory OLAP and can further process different multiple queried stages on the memory database integrated machine with pipelining parallelism and can improve system OLAP query throughput performance. The memory data warehouse query processing implementation method is suitable for a memory OLAP application scene of the memory oriented data warehouse integrated machine, and can adapt to the memory OLAP performance acceleration demand under the memory database integrated machine asymmetric hardware architecture.

The invention discloses an image encryption method based on a splicing model and a hyper-chaotic system, and relates to the field of image encryption. The hyper-chaotic system and the splicing model of a DNA calculation are introduced to the process of image encryption. Firstly, a pixel value is disintegrated into four integers ranging from zero to three by using a mode of converting decimalism to quaternion; secondly, the obtained four integers are coded by using basic groups of A, G, C and T respectively, an original gray level image can be disassembled into four parts, and operations can be conducted on the four parts respectively; in addition, the method takes full advantage of a basic idea of the splicing model to form a splicing operation, and the splicing operation is applied to pixel values of a scrambling image. According to a simulated result and safety analysis, not only are large key space and high sensibility provided, but also an exhaustive attack, a statistical attack and a difference attack can be resisted, and the encryption effect is good.

The invention discloses a GPU and cascade Hash-based quick image SIFT feature matching method and belongs to the field of computer vision. The method comprises the steps of establishing a three-stage exchange mechanism of a GPU, a memory and a hard disk according to size limitation of a global display memory of the GPU and size limitation of the memory; by adopting an improved GPU parallel reduction method, performing Hash mapping of two different encoding lengths on all SIFT feature points of an image; proposing and using an upper triangular matrix block reading method; performing coarse screening through a local sensitive Hash algorithm; performing fine screening on candidate points by calculating a Hamming distance; and finally finding a most suitable matching point by calculating Euclidean distances between the screened points and to-be-matched points. By utilizing an asynchronous parallel method of a CPU and the GPU, matching result data of an image pair is copied to the memory from the display memory of the GPU and stored in the disk during GPU operation; and the matching time of the SIFT feature points is greatly shortened and the feature matching also can be finished in a short time under the condition of massive data.

The invention discloses a large-scale network information propagation calculation method and device based on a distributed platform. The method comprises the following steps of: constructing a configuration network according to network parameters; Constructing of information propagation model, selecting to set parameters of a network and an information spreading model according to requirements; Configuring information format, Individual attributes, Message transfer function, The method comprises the following steps: carrying out iterative computation on a message fusion function, a message processing function, an input network, an information format, an individual attribute, a message transfer function, the message fusion function and the message processing function through a parallel distributed platform, and screening out nodes with propagation activities in a network containing propagation records and connected edges of the nodes to obtain the propagation records and outputting thepropagation records. The method has the advantages of being high in parallel computing capacity, high in user customization capacity, good in propagation computing universality and adaptability and simple and efficient in development process, the development difficulty of parallel distributed computing can be greatly reduced, and the problem of parallel distributed storage does not need to be additionally considered.

The invention relates to the field of image encryption, and designs an image encryption method based on a chaotic system and an insertion-deletion model. According to the method, the Hamming distance, the Hamming inverse distance and the Hamming feeding distance are introduced into the generating process of initial values of the chaotic system; in addition, when the method is used for encrypting or decrypting an image, positions of image pixel values are scrambled through the insertion-deletion model; finally, XOR operation of a DNA sequence is applied to diffusing the image pixel values. It can be obtained from simulation results and safety analysis that the method has the good encryption effect and can resist various attacks of an intruder. According to the problem emphatically solved, the basic thought of the insertion-deletion model in DNA calculation is applied to image encryption.

The invention discloses a verification platform for a control and protection algorithm of a high voltage direct current power transmission system. The verification platform is composed of a power system real-time emulator, an emulator workstation, a PXI industrial computer and a PXI upper computer. The verification platform utilizes the real-time emulator to replace a miniature small-power actual high voltage direct current power transmission system, and overcomes the limitation that the theoretical physics is difficult to change the parameters and topological structure of the system and a short trouble test is dangerous to perform. The platform selects the PXI industrial computer to serve as the implementing platform of the high voltage direct current control and protection algorithm to replace an expensive direct current control and protection deice; meanwhile the PXI industrial computer has the advantages of being easy to develop algorithm and call the parameters online. The platform plays a very important role in the control and protection algorithm verification and parameter setting of the high voltage direct current power transmission system.

The invention relates to a light terminal 3D cloud rendering method. A 3D program is deployed in a cloud server cluster according to a preset rule, the cloud server cluster receives a request and allocates the request when a user terminal initiates the request, the 3D program can execute rendering operation on the cloud server cluster and return a rendering result to the user terminal, and the user terminal judges the rendering result and performs corresponding processing. According to the invention, the cloud server cluster is applied; the method is based on extremely high hardware configuration and massive computing resources of a cloud rendering server, a flexible elastic computing capability and a super-strong parallel computing capability are provided, graphs and instructions can be quickly processed, a plurality of 3D programs can be rendered in parallel, and high-performance host-level high-definition image quality and extremely smooth experience can be obtained by all terminalequipment without downloading while the user cost is reduced; the cloud rendering server can carry out load balancing to build a cluster to process a more complex large 3D program, the problem that acertain server fails and cannot work is not worried about, and the cloud rendering server is safer and depends on the spectrum.

The invention discloses a method for structuring ultrasound long axis image quickly with high fidelity. The method is implemented by using a parallel computing architecture based on a graphics processor to perform efficient management to three-dimensional IVUS data with large data volume by quadtree data structure, performing recursion traverse to the quadtree and computing the intersecting point coordinate of the plane on which the long axis image is located and the quadtree data node, decomposing the long axis image into a series of small planes, then sending to the graphics processor for mapping and interpolation computing. The invention solves the problem that the data volume is too large to be loaded display memory of the graphics processor, improves utilization efficiency of texture cache of the graphics processor, and enhances computation speed and image quality by computing the three-dimensional spline interpolation by using single instruction multidata vector dot product of the graphics processor. The invention can achieve above 30 frame per second in structuring long axis image with high fidelity on common middle / high-grade display card for large scale IVUS image data of 512*512*4000*16bit.

The invention relates to a self-adaption quantum ant colony-based query method and a self-adaption quantum ant colony-based query device for railway traffic transfer plans. The method comprises the following steps of performing permission setting and database update setting on a query machine terminal by a management console through a management machine; starting the query machine terminal, receiving a query request by a touch screen, and sending the query request to a central processing module; building a multi-mode optimal path model by the central processing module according to the query request, executing a self-adaption quantum ant colony algorithm based on the multi-mode optimal path model, and working out optimal railway traffic transfer plans under different optimal targets; and judging whether optimization is ended or not, and feeding back a computed result to the touch screen if the optimization is ended. The device comprises the management console, the management machine and the query machine terminal. Compared with the prior art, the method and the device disclosed by the invention are capable of performing high-efficiency optimization design on the optimal transfer of a complicated national railway traffic network under the multi-constraint condition, and therefore, the flexibility and the high efficiency of passenger travelling to the railway traffic transfer can be improved.

The invention discloses a crowd sensing task allocation and user recruitment model cross validation method and system. The method comprises the following steps: inputting a plurality of time-space sub-tasks obtained by dividing a crowd sensing task model according to regions and time by using a task allocation model to be subjected to cross validation, and inputting parameters of an agent agent determined by using a user recruitment model; importing the time-space subtasks into an artificial social platform in which a user agent model, an artificial social environment model and an agent interaction model are pre-established, and starting simulation calculation to generate an experimental sample; and performing data statistics on the specified evaluation indexes for the generated experimental samples to obtain an experimental result containing the specified evaluation indexes. According to the method, cross validation of the task allocation model and the user recruitment model can be realized, and the method has the advantages of high parallel computing capability, good user customization capability, simple and efficient development process, high practicability and wide application prospect.

The invention discloses a distributed large-scale complex community detection method and device based on node degree. The method comprises the following implementation steps: preparing relational network data; Loading the relational network data to a distributed file system for network segmentation, customizing a message transfer function, a message fusion function and a message processing function based on a label propagation algorithm of the node degree, and then loading the message transfer function, the message fusion function and the message processing function into the distributed file system; Configuring operation parameters of the distributed file system; And calling the distributed file system to execute community detection and outputting a community detection result. The method is suitable for community detection of a large-scale complex network, and has the advantages of high accuracy, good stability, strong parallel computing capability, good community detection universality and adaptability, and simple and efficient development process.

The invention provides an image encryption method based on the Hash function and an inserting-deleting model and relates to the field of image encryption. The Hash function and an inserting-deleting model of DNA computing are introduced to an image encryption processing. First, operating pixel values by use of the Hash function to obtain a Hash table; scrambling the pixel values according to the Hash table and then performing DNA coding to a scrambled matrix; in addition, the method makes full use of the advantages of the inserting-deleting model; the inserting-deleting model is applied in spreading the pixel values of the image. Simulation results and security analysis show that the method not only has large cipher code space and high sensitivity, but also can resist exhaustive attacks, statistical attacks, and differential attacks, etc., and has good encryption effect. The method is stressed on applying the Hash function and the inserting-deleting model in image encryption.

The invention discloses a binocular perception method capable of realizing hardware acceleration, which comprises the following steps of: firstly, leading out external trigger pins of two camera chipsof a binocular camera, and configuring registers of the camera chips to enable the camera chips to work in an external trigger mode; capturing a binocular image through a binocular camera and performing offline calibration to obtain parameters of the binocular camera; then, connecting a binocular camera to a PL part of the Zynq Ultra Scale + MPSoC to carry out synchronous triggering , and collecting a real-time binocular image generating a custom IP according to the parameters of the binocular camera, and in the PL, calling the custom IP to perform camera correction and performing binocular stereo matching on the real-time binocular image so as to generate a real-time disparity map; and finally, transmitting the real-time disparity map to a PS part of the Zynq Ultra Scale + MPSoC, carrying out data interaction with a DDR4 memory in the PS part, and meanwhile, generating a Linux kernel and a file system to manage an image cache in the DDR4. The invention can be applied to an embedded system having requirements for computing power and power consumption, and hardware acceleration can be effectively achieved.

The invention discloses a cloud media server system. The cloud media server system comprises a cloud media management platform and a cloud media resource pool, wherein the cloud media management platform comprises a service management module, a resource management module, an operation management module and a cloud media resource adaptation layer module; the service management module is applied to service catalogue management, service instance management, user management and portal management; the resource management module is applied to asset management, resource encapsulation, resource monitoring, resource scheduling, template management and application management; the operation management module comprises report management, charging management, maintenance management, system management and external interface management; and the cloud media resource adaptation layer module is used for finishing data interaction and virtual management based on a virtual platform and the cloud media resource pool. The cloud media server system has the advantages of ultrahigh parallel computing capability, highly-flexible expandability and accurate charging measurement; the platform resource service capability can be brought into full play; and the network operation cost of services can be lowered.

The invention discloses a small FMCW-based (frequency modulated continuous wave) SAR (synthetic aperture radar) imaging system by using an FPGA (field programmable gate array) to mainly overcome the disadvantages of large volume, heavy weight and high power consumption of the traditional SAR imaging system. The small FMCW-based SAR imaging system comprises a radar front end, an acquisition preprocessing unit, a signal processing unit and a receiving host, wherein the acquisition preprocessing unit comprises an instruction communication module, an A / D (analogue / digital) acquisition module, an FPGA control and calculation module and a data sending interface; the signal processing unit comprises a data receiving interface, an FPGA control module, a DSP (digital signal processor) calculation module and a data sending interface; the acquisition preprocessing unit and the FPGA of the signal processing unit are used for receiving back waves and inertial navigation parameters of the radar front end, feeding back the regulation instruction to the radar front end to perform distance pulse pressure, transmitting data to a DSP, receiving the calculation results of the DSP and transmitting thecalculation results of the DSP to the receiving host for display. According to the small FMCW-based SAR imaging system, one picture can be imaged in real time in one second under the airborne condition, the small FMCW-based SAR imaging system has the advantages of small size, light weight, small power consumption and good reliability and can be applied to the fields of radars, guided missile, remote sensing, etc.

The invention discloses a big data processing framework-based application of a neural network algorithm in IDS, and designs a big data technology processing framework-based application of a BP neuralnetwork classifier model in an invasion detection system. Besides a big data technology processing framework, the model furthermore applies a Kolmogorov theorem and a training function trainbfg quasi-Newton algorithm-based BP neural network algorithm.

The invention provides a secondary sound source laying method for a helicopter cabin active noise control system. The secondary sound source laying method for a helicopter cabin active noise control system is characterized in that: based on a quantum particle swarm optimization algorithm, encoding the position of a secondary sound source, taking one encoding string as a particle, performing iterative computation, and computing the fitness value of each particle during each iteration process, and performing element value updating on the particles. The secondary sound source laying method for a helicopter cabin active noise control system has memory so that each particle can still maintain the individual historical extremum of each particle when the algorithm is ended, and taking regard of the situation that the secondary sound source in a helicopter cabin has constraint, various meaningful selection schemes can be provided. Besides, the secondary sound source laying method can search the secondary sound source position in a plurality of secondary sound source working faces, and can select better laying of secondary sound source in the helicopter cabin; and the secondary sound source laying method can take regard of a plurality of frequencies at the same time, laying of secondary sound source robustness at a plurality of harmonic frequencies, that is, good noise reduction effect can be obtained when the secondary sound source is laid at the frequencies.

The invention relates to a stream line texture mapping unit system structure. A texture mapping unit comprises a level computing unit, a task allocation unit, a four-path texture address computing unit, a first data assembly and task allocation register, an L1 texture cache, a second data assembly and task allocation register, a four-path texture data extraction and normalization unit, a third data assembly and task allocation register, a four-path texture comparator, a fourth data assembly and task allocation register, a four-path filtering unit and an output data assembly unit, which are connected in sequence. The stream line texture mapping unit system structure provided by the invention provides reference and guide functions for a GPU system for improving the texture mapping performance in a hardware acceleration mode.

The invention discloses a high-precision real-time ultrasonic image scan conversion method comprising the steps: putting coordinate conversion and an image interpolation algorithm in a scan conversion module into GPU by using a parallel calculation frame and an SIMD vector operation base on a GPU and drawing a proper agent geometric model to carry out 32 bit floating point number high-precision calculation; and smoother image quality is obtained by using an optimal high-precision bicubic interpolation algorithm. The high-precision real-time ultrasonic image scan conversion method can achieve scan conversion speed over 60 frame / second for large-scale data of 1024 sampling points on each line of 512 line data on present common middle and low level display cards, sufficiently develop and usestronger parallel computation capability supplied by the present GPU, and thereby the traditional scan conversion module is upgraded into a next-generation high-definition scan conversion module.

The invention provides a hyperspectral image target detection method and device, the method is realized based on an FPGA development board, and the method comprises the steps: obtaining a to-be-detected hyperspectral image, and forming an original matrix of the hyperspectral image; calling an IP core realized based on an FPGA development board, and calculating the hyperspectral image matrix by adopting an RX algorithm to obtain a target information result matrix; and comparing the target information result matrix with a preset judgment threshold, and if the target information result matrix is greater than the judgment threshold, determining the target information corresponding to the target information result matrix as the target information of the hyperspectral image. The complex calculation process is transplanted through the FPGA device, and compared with a traditional processor, hyperspectral image target detection can be rapidly and efficiently carried out.

The invention discloses a recommendation method for a next interest point of a social network, and the method comprises the steps: obtaining a centralized representation of a preference interest point of a user through an encoder layer of a Transform; secondly, sending the obtained centralized representation of the user preference interest points to a divided time window; then, aiming at different interest preference distributions of the same user in different time windows, endowing different weights in different time windows, and obtaining a preference interest point state of the user according to the obtained weights; finally, filtering out candidate POIs from all the POIs according to geographical factor information and merchant popularity, calculating scores of the candidate POIs based on the preference interest point state of the user, and recommending Top_k POIs to the user. According to the method, the problem that recommendation of the next interest point of the social network is difficult due to sparse data can be solved.

The invention relates to a light terminal 3D cloud rendering method. The 3D program is deployed in the cloud server cluster according to preset rules. When a user terminal initiates a request, the cloud server cluster receives and distributes the request, and the 3D program can perform rendering operations on the cloud server cluster. , and send the rendering result back to the user terminal, and the user terminal judges the rendering result and performs corresponding processing. The present invention applies cloud server clusters, based on the extremely high hardware configuration and massive computing resources of the cloud rendering server, provides flexible elastic computing capabilities and super parallel computing capabilities, can quickly process graphics and instructions, and render multiple 3D programs in parallel. While reducing user costs, all terminal devices can obtain high-performance host-level HD image quality and extremely smooth experience without downloading; cloud rendering servers can perform load balancing and build clusters to process more complex large-scale 3D programs, without worrying about a certain server It is safer and more reliable to solve the problem of failure to work.

The invention relates to a stream cipher based secure communication protocol. The communication is one-time pad, keys are owned by both parties and are synchronous, the generation speed is fast, the fault-tolerant capacity is achieved, a 128-bit original text is processed every time, digital signature and data encryption are performed on the original text, and subsequently, 256-bit processed data are sent; firstly, original text grouping and hashed value calculation are performed, secondly, the data are encrypted, and finally, data decryption is achieved. According to the stream cipher based secure communication protocol, a method is designed on the basis of a stream cipher system, a grouping encryption processing structure is used for reference, the 256-bit data can be processed every time, high encryption and decryption efficiency is achieved, a random number generator is used for generating the keys which are owned by both communication parties and are synchronous, and the occupation space of the keys is small.

The invention relates to a high-performance BW100 chip based SAR (Synthetic Aperture Radar) parallel processing method. The method comprises the following steps: an echo datum is transmitted by an FPGA (Field Programmable Gate Array) controller in a ping-pong manner and then is processed by a master BW100 chip and a slave BW100 chip, the processing result is transferred to a DDR3 (double data rate) memory through a ping-pong QDRII (quad data rate) memory to finish processing the echo number and the distance vector; the ping-pong QDRII memory takes the direction vector data from the DDR3 memory, transmits the data in a ping-pong manner to the master BW100 chip and the slave BW100 chip through the FPGA controller to process the data, and receives the processing result in a ping-pong manner, so as to finishing processing an imaging direction vector; after the distance vector and the direction vector are processed for multiple times, the whole imaging processing is completed. The invention further discloses a high-performance BW100 chip based SAR parallel processing device. The BW100 chip is taken as an SAR imaging processing core processor, and the device has the advantages of strong parallel processing capability and data communication capability and excellent overall performances, and is suitable for the SAR real-time imaging processing field.

The invention discloses a fast image SIFT feature matching method based on GPU and cascaded hash, which belongs to the field of computer vision. According to the limitation of GPU global video memory size and memory size, the present invention establishes a three-level exchange mechanism of GPU, memory, and hard disk; at the same time, an improved GPU parallel protocol method is adopted to perform two hash maps with different encoding lengths on all SIFT feature points of the image ; proposed and used the upper triangular matrix block reading method; rough screening by local sensitive hash algorithm; fine screening of candidate points by calculating Hamming distance; finally, by calculating the Euclidean distance between the screening point and the point to be matched to find the most Appropriate matching point; using the asynchronous parallel method of CPU and GPU, the matching result data of the image pair is copied from the GPU memory to the memory and saved to the disk while the GPU is operating. The present invention greatly shortens the matching time of SIFT feature points, and can complete feature matching in a short time even in the case of massive data.

The invention provides an image encryption method based on the Hash function and an inserting-deleting model and relates to the field of image encryption. The Hash function and an inserting-deleting model of DNA computing are introduced to an image encryption processing. First, operating pixel values by use of the Hash function to obtain a Hash table; scrambling the pixel values according to the Hash table and then performing DNA coding to a scrambled matrix; in addition, the method makes full use of the advantages of the inserting-deleting model; the inserting-deleting model is applied in spreading the pixel values of the image. Simulation results and security analysis show that the method not only has large cipher code space and high sensitivity, but also can resist exhaustive attacks, statistical attacks, and differential attacks, etc., and has good encryption effect. The method is stressed on applying the Hash function and the inserting-deleting model in image encryption.

The invention provides a secondary sound source laying method for a helicopter cabin active noise control system. The secondary sound source laying method for a helicopter cabin active noise control system is characterized in that: based on a quantum particle swarm optimization algorithm, encoding the position of a secondary sound source, taking one encoding string as a particle, performing iterative computation, and computing the fitness value of each particle during each iteration process, and performing element value updating on the particles. The secondary sound source laying method for a helicopter cabin active noise control system has memory so that each particle can still maintain the individual historical extremum of each particle when the algorithm is ended, and taking regard of the situation that the secondary sound source in a helicopter cabin has constraint, various meaningful selection schemes can be provided. Besides, the secondary sound source laying method can search the secondary sound source position in a plurality of secondary sound source working faces, and can select better laying of secondary sound source in the helicopter cabin; and the secondary sound source laying method can take regard of a plurality of frequencies at the same time, laying of secondary sound source robustness at a plurality of harmonic frequencies, that is, good noise reduction effect can be obtained when the secondary sound source is laid at the frequencies.