35results about How to "Guaranteed computing power" patented technology

The invention relates to a water environment analysis method based on a geographic information system and a Delft3D-FLOW calculation engine. The water environment analysis method comprises the steps that a software system framework is built through a GIS platform; a Delft3D mesh file format is analyzed; a configuration file MDF file format of the Delft3D-FLOW and meanings of multiple parameters in the configuration file MDF file format are analyzed, and a human-computer interaction configuration interface is developed for configuring the multiple parameters under the GIS platform; a model is built, configuration tasks of a plurality of environmental elements are completed through GIS operation; a configuration result is output as a file format which can be identified by the mathematical calculation engine, and the calculation engine is called to conduct calculation; a calculation result file format is analyzed, a file is read under the GIS platform and processed into a raster image with space coordinates through the image processing technology, and the raster image is loaded into the GIS platform for display. The water environment analysis method combines the mathematical model with spatial and geographic information, thereby improving analysis accuracy. The invention further provides a water environment analysis system based on the geographic information system and the Delft3D-FLOW calculation engine.

The invention discloses a method for enhancing sharpness of color image with color constancy. The method includes the following steps: based on the red, green and blue three wave bands of an image, calculating a brightness image; setting a neighborhood radius, calculating mean value and standard difference of the brightness image in each pixel neighborhood, obtaining an neighborhood mean value image and a neighborhood standard difference image; based on the neighborhood mean value image and the neighborhood standard difference image, conducting adaptive calculation on weight of each pixel point; based on the neighborhood mean value image and the weights, recalculating a new brightness image; in accordance with the theory of color constancy, based on the new brightness image and an original brightness image, conducting same ratio enhancement on the three wave bands of the original image, obtaining the image with enhanced sharpness with color constancy. According to the invention, the method can greatly increase image sharpness, reduce color distortion as much as possible, maintain color constancy, and effectively increase visual effects of the image.

The invention is applicable to the technical field of computers, and provides a coprocessor, and a method and a system for matrix operation acceleration. The coprocessor comprises a controller, a register bank, a multiplier-adder MAC array, and a memory. The controller parses a vector expansion instruction, and calls and sends data to be calculated to the register bank according to a parsing result. The register bank transmits the data to be calculated to the MAC array and transmits an operation result to the memory, the operation result being obtained by matrix operation of the MAC array on the data to be calculated, and the data to be calculated and / or the operation result are stored in the memory. In the process, the MAC array performs matrix operation according to the vector expansioninstruction, and the vector expansion instruction is generated by a main processor according to a preset function, ensuring computational power of a general microprocessor. The register bank transmitsand stores all kinds of data, and a data storage channel is independent of a receiving channel of the vector expansion instruction, improving data throughput of a general-purpose microprocessor in anoperation process.

The invention discloses a flexibly configurable neural network computing unit, a computing array and a construction method thereof. The neural network computing unit includes: a configurable storage module, a configurable control module, and a multiply-add computing module capable of time-division multiplexing; a configurable storage module The modules include: feature map data buffer, step size data buffer and weight data buffer; the configurable control module includes: counter module and state machine module; the multiplication and addition calculation module includes: multiplier and accumulator. The invention can support any type of convolution calculation, and supports parallel calculation of multi-size convolution kernels, fully explores the flexibility and data reusability of convolutional neural network calculation units, greatly reduces system power consumption caused by data migration, and improves The computational efficiency of the system.

The invention discloses a CNN processing method and device, which are used for realizing the module operation in the CNN model on the ASIC, and reducing the operation cost while ensuring the calculation performance of the CNN. An embodiment of the present invention provides a CNN processing method, including: obtaining the intensive type of the first module in the CNN model; if the intensive type of the first module is computationally intensive, deploying the first module to a dedicated integration on the circuit ASIC, and obtain the multiple computing resources of the ASIC occupied by the first module; merge the same computing resources among the multiple computing resources of the ASIC occupied by the first module, and obtain the deployment The first module on the ASIC after completing the combination of computing resources; running the first module deployed on the ASIC after completing the combining of computing resources.