112results about How to "Avoid redundant calculations" patented technology

The invention provides a method for detecting an X-ray mammary gland lesion image based on feature pyramid network under transfer learning. The method comprises the steps: 1, establishing a source domain data set and a target domain data set; 2, establishing a deformable convolution residual network layer by a deformable convolution and extended residual network module; 3, establishing a multi-scale feature extraction sub-network based on a feature pyramid structure through a feature map up-sampling and feature fusion method in combination with a deformable convolution residual network layer;4, establishing a deformable pooling sub-network sensitive to the focus position; 5, establishing a post-processing network layer to optimize a prediction result and a loss function; and 6, migratingthe training model to a small sample molybdenum target X-ray mammary gland focus detection task so as to improve the detection precision of the network model on the focus in the small sample image. According to the method, a transfer learning strategy is combined to realize focus image processing in a small sample medical image.

The present invention discloses a license plate recognition system by utilizing spatial-temporal search-space reduction and method thereof. The license plate recognition system can selects a pixel with character edge feature from the inputted image and a plurality of candidate regions from the image; and detect if the candidate region of the inputted image has been appeared in a preceding image. Then the recognition system in the present invention separates the candidate region into a plurality of blocks, calculates the repeated block based on a current image and a preceding image, and determines whether the candidate region is repeated according to the repeated block. Thus, by discarding the repeated regions, the license plate recognition system in the present invention can avoid redundant calculation and improve the efficiency and performance for recognition of real plate character in the image.

This invention relates to the field of multi-objective workflow optimization. Certain exemplary embodiments of the invention are applicable in cases where workflow descriptions contain choice variables relating for instance to the selection of a specific service provider out of several service providers that provide similar services, to the selection of human workers, or to the selection between alternative subworkflows. A binding represents a combination of choices, binding the choice variables to specific values. Bindings induce specific cost and/or quality properties to the workflow, a binding being Pareto-optimal if no other binding exists that is at least as good for every cost and/or quality property and better for at least one property. Certain exemplary embodiments relate to a system and/or computer-implemented method for computing an approximation of the set of Pareto-optimal bindings such that the computed approximation satisfies specified minimum precision requirements

The invention discloses a five-phase permanent magnet synchronous motor model prediction torque control method, and belongs to multi-phase motor alternating current control system design and manufacture in the field of power electronics and power drive. The main steps are as follows: first a sector where a dead-beat voltage vector is is calculated, and only four voltage vectors (two medium vectors plus two large vectors) in the sector are selected as an input set of model prediction torque control, thereby realizing real-time updating; and then a prediction model is utilized to perform prediction calculation on the four voltage vectors in the input set, and combined with an objective function, an optical voltage vector is selected as an input of a system at the next moment. The method provided by the invention adopts model prediction torque control, has excellent dynamic response performance, and is relatively small in calculation amount, digital realization is relatively easy, and the pressure of a digital chip is relieved.

The invention discloses a deep neural network multi-path reasoning acceleration method for edge intelligent application, comprising the following steps of: firstly analyzing the classification capability and calculation cost of quitting a branch outlet in advance of each layer of a deep neural network, and selecting an outlet combination with the maximum reasoning benefit to be added to an original model; setting a threshold unit between the exit and a trunk layer, training, and judging whether the task can exit from the current exit or not; compressing intermediate feature data for a task which cannot exit in advance at a terminal layer but must be transmitted to an edge layer; and finally, monitoring and analyzing the network load and the computing capabilities of the terminal and the edge equipment on line in an edge computing environment, cutting a multi-path model by taking the minimized reasoning time delay as a target, and respectively deploying model blocks in a terminal layerand an edge layer to finally form a multi-path reasoning acceleration framework. According to the method, the reasoning flexibility can be improved, the accuracy is ensured, the total reasoning time delay is reduced, and the real-time and high-precision requirements of edge intelligent application are met.

The invention relates to a motion vector estimation method which comprises the following steps of: detecting a motion state of a current encoding macro block and carrying out motion estimation by using a first search mode if the motion state is a first motion state; and carrying out motion estimation by using a second search mode if the motion state is a second motion state. By analyzing context environments of the encoding macro block and using different search modes aiming at the macro blocks with different motion states, the invention effectively avoids redundant computations in the macro block motion vector search process under the static or simple motion state, reduces the computation amount, can save 45 percent of computation amount for a sequence of a simple state compared with a hexagonal search, and save 21 percent of computation amount for a sequence of a complex state compared with the hexagonal search.

The invention provides a QR two-dimensional code self-adaptation binarization processing method and device based on light balance. The method mainly comprises the steps that image information of a QR two-dimensional code is collected, and the image information of the QR two-dimensional code is transformed into a gray level image; light balance processing is conducted on the gray level image of the QR two-dimensional code according to an illumination detection result; each row of pixel points in the gray level image after light balance processing is conducted are grouped according to an image gray level jumping threshold value, and the binarization threshold value of the gray levels of the pixel points of each group of each row is determined according to the maximum between-cluster variance method, and binarization processing is conducted on the gray levels of the pixel points in each group of each row according to the binarization threshold value. According to the QR two-dimensional code self-adaptation binarization processing method and device based on light balance, pre-recognition is conducted on the gray level distribution of the QR two-dimensional image, light balance processing is conducted on the QR two-dimensional image, the binarization threshold value is determined through the maximum between-cluster variance method, redundant calculations can be avoided, the binarization efficiency of the QR two-dimensional code is improved, and meanwhile the recognition accuracy of the QR two-dimensional code is ensured.

The invention discloses a UWB/MEMS combination-based UWB base station position error compensation method, relates to the technical field of integrated navigation, and aims to solve the problems that the UWB base station position measurement accuracy in the existing system has errors, is easily influenced by the environment and is difficult to accurately measure. The UWB base station position errorcompensation method comprises the following steps: 1, collecting data output by a UWB system and an MEMS system; 2, selecting the state quantity and the observed quantity of the system, and establishing a UWB/MEMS integrated navigation system state space model; 3, performing state estimation on the UWB/MEMS integrated navigation system state space model by using a derivative unscented Kalman filter; and 4, obtaining a position error of the UWB base station and gyroscope constant drift and accelerometer zero offset of MEMS inertial navigation according to a state estimation value output by thederivative unscented Kalman filter, feeding back the position error, the gyroscope constant drift and the accelerometer zero offset to the combined system, and outputting attitude, speed and positioninformation.

The invention discloses a voice recognition method and system based on a triggered non-autoregressive model. The method comprises the following steps: S11, extracting an acoustic feature sequence; S12, generating a convolution downsampling sequence; S13, generating an acoustic coding state sequence; S14, calculating probability distribution and connection time sequence loss of the prediction marks; S15, calculating the positions and the number of peaks; S16, calculating cross entropy loss by an acoustic decoder; S17, calculating a gradient according to the joint loss of the connection time sequence loss and the cross entropy loss, and performing back propagation; and S18, circularly executing the steps S12 to S17 until the training is completed. The system comprises an acoustic feature sequence extraction module, a convolution downsampling module, an acoustic encoder, a connection time sequence classification module, an acoustic decoder and a joint loss calculation module which are connected with one another in sequence. The connection time sequence classification module comprises a linear change module, a connection time sequence loss calculation module and a peak extraction module.

The invention provides a mobility managing system and method for an integrated identifier network as well as a routing method. In the integrated identifier network, the mobility managing system based on DHT (distributed hash table) is provided, so that the system has good load balance and expandability, burden on a mapping server is alleviated by a local buffer, and the time of establishing communication and time delay of data forwarding are shortened; the Overlay routine method and the mobility managing method which are based on MDP (Markov decision process) are provided, therefore the method has the characteristic of distributive computation, and can be used for balancing and reducing the complexity of time and space, and effectively avoiding a loop and abundant redundancy computation so that the methods are suitable for a large scale mobility management; and by utilizing the methods, a remuneration function is defined by combining physical layer and application layer information, thereby reducing the influence resulting from topology inconformity between an Overlay network and a physical network to a certain extent, and optimizing the update and query performance of a mobility management mechanism.

The embodiment of the invention discloses an obstacle-oriented wireless sensor network connectivity restoration method and device. By supposing that n network partitions exist in a wireless sensor network and each network partition is represented by a partition node, the whole wireless sensor network is represented by n partition nodes. The obstacle-oriented wireless sensor network connectivity restoration method comprises the steps: establishing a minimum spanning tree of the n partition nodes under an obstacle condition; allocating a relay node, including a static relay node and a movable relay node, for the established minimum spanning tree of the n partition nodes under the obstacle condition; by utilizing the allocated relay node, obtaining network data in a partition node connected with the relay node, and connecting each partition node according to the obtained network data, thereby restoring the connection of the wireless sensor network including the n network partitions. By adopting the embodiment provided by the invention, a problem of wireless sensor network connectivity restoration under the obstacle condition is solved.

The invention relates to a method for estimating an underwater acoustic channel impulse response function in a multipath environment. Matching pursuit is improved by using a Schmidt orthogonalization strategy, so that the residual can be reduced on the orthogonalization direction by each step of iteration in a solving process so as to avoid redundant calculation. According to the method provided by the invention, a Schmidt orthogonalization process is imported in a matching pursuit iteration strategy for the first time, a convolution matrix is constructed via probe signals or training sequences, matching selection is performed on the convolution matrix by using Schmidt orthogonalization, and the matching intensity in every matching operation of the convolution matrix is arranged in sequence to construct a dictionary; and an estimation result of underwater acoustic channel is obtained by pseudo inverse operation. Therefore, delay and amplitude attenuation parameters are obtained.

The invention discloses a method for analyzing a random response of a non-proportional damping structure under an earthquake effect. The method comprises the following specific steps: 1, acquiring a unit impulse response of the structure about any excitation; 2, based on a random vibration theory, establishing a direct convolution formula responding to power spectrum calculation; 3, performing PODfeature orthogonal decomposition on the non-uniform modulation function excited by each point; and 4, calculating an evolution frequency response matrix by using FFT, and further determining an evolution power spectrum matrix and a variance matrix of the structure response. According to the method, the response statistics is explicitly represented as the convolution form about the pulse response,and the convolution is calculated by using the FFT technology, so that the response calculation efficiency is greatly improved; the method is suitable for a linear structure, is also suitable for a nonlinear structure, and can also be used for a linear system subjected to order reduction through a modal superposition method.

The invention provides a blocking and hierarchical structure based power distribution system reliability evaluation method which comprises the following steps; establishing a reachable matrix according to a power distribution system, dividing a manual block and an automatic block according to the reachable matrix, establishing a manual block hierarchical structure and an automatic block hierarchical structure according to the manual block and the automatic block, and performing reliability analysis according to the manual block hierarchical structure and the automatic block hierarchical structure. According to the method, basic analysis units of the system are reconstructed according to blocking theory, the establishment of a fault-mode consequence list is avoided according to the minimum path theory, repeated calculation on the reliability of identical routes is avoided according to the hierarchical structures, high analysis efficiency is obtained, and the problems that the power distribution grid reliability analysis is large in calculation amount, slow in speed and low in accuracy rate are solved.

A self-adaptive power distribution network fault positioning method based on a fault current path comprises the following steps: encoding a feeder line section, an FTU and a power supply in a power distribution network, and constructing a network description matrix D; updating the connection relationship in the network description matrix D according to the actual operation topological structure ofthe current power distribution network; generating a power distribution network adaptive switching function according to the network description matrix D; when the power distribution network fails, reading current out-of-limit information uploaded by all FTUs in the power distribution network, and generating a current out-of-limit information matrix I; and establishing a fault positioning model of the power distribution network, solving the fault positioning model of the power distribution network based on an intelligent optimization algorithm, and finally determining a fault position. The method can be universally used for a traditional power distribution network and a power distribution network with distributed power supply access, rapid and reliable positioning of a single fault and aplurality of faults in the power distribution network can be realized, and switching of the operation state of the power distribution network does not influence the method provided by the invention.

The invention provides a multi-mode pattern matching method based on structure correlation against the multi-mode pattern matching problem. The redundant calculation problem can be solved. On the premise of adding a small amount of extra space based on the structure correlation of mode patterns, the overall performance of the pattern mode matching is improved. According to the method, a structureinheritance relationship between mode patterns is established mainly through structure correlation between the mode patterns, the originally mutually independent mode patterns are expressed as an integral data structure, and the data structure retains non-repeating parts which all the mode patterns need to be matched with. The mode pattern matching sequence is planned again, so that when the matching is performed in a data map, multi-time matching with the structurally repeated parts in the mode patterns can be avoided, so that the pattern mode matching time performance is greatly improved, and the goal of speeding up the calculation is achieved.

A multi-threaded graphics processor is configured to generate a tessellated iso-surface from a volumetric description using slices of values that are stored in render targets. The volumetric description can be a complex mathematic equation, a sum of metaballs, a pre-computed scalar field represented as a 3D volume texture, or a rendered volume. Slices are aligned along an axis and spaced before being intersected with the volume to determine iso-surface intersections for the x, y, and z axes. The intersections are stored in render targets and are processed by a shader program to produce a tessellated iso-surface.

The invention discloses an unstructured grid structured processing method based on array plane propulsion, and aims to solve the problems that in the unstructured grid structured processing, the localdirection is irrelevant to the application problem, the orthogonality of the triangular unstructured grids is poor, and the calculated amount is larger. According to the technical scheme, the methodis characterized by generating a two-dimensional unstructured grid; defining and initializing a key element array and a local direction array of an unstructured grid unit according to an unstructuredgrid related data structure; and determining the key elements and local directions of the unstructured grids by adopting the array plane propulsion, and outputting a key element array and a local direction array to realize the unstructured grid structured processing. By adopting the method, the local directions of the two-dimensional triangular, quadrilateral or triangular-quadrilateral hybrid unstructured grid can be efficiently generated, the orthogonal performance of the local directions is improved, and one local direction is close to the normal direction of the wall surface, so that the precision of the temperature gradient of the rocket wall surface is effectively improved, and the redundant calculation of an exhaustion method is avoided.

The invention discloses a rapid power distribution network reliability assessment method based on element transfer characteristics. The method includes the steps that the power outage influence on upstream loads by a downstream fault is considered in the adverse current transfer process of reliability indexes, and the power outage influence on downstream loads by an upstream fault is considered in the fair current transfer process. In the adverse current transfer process, the reliability indexes of a branch line are transferred to a feeder line and then transferred to a feeder line of the upper grade from the tail end of the feeder line. In the fair current transfer process, the operation time of different protective elements is included in the reliability indexes through step-by-step transfer of the protective elements. The two equivalence processes are combined to obtain the reliability indexes of all points in a system. The unreliable movement rate of the protective elements is considered, and the reliability indexes can be transferred in a whole network. The centralized and classified transfer mode is adopted, the reliability indexes can be transferred in the whole network, and the more and longer the feeder lines are, the more obvious the effect is.

The invention discloses a command and control network two-terminal reliability estimation method based on an edge expansion diagram (EED). The method comprises the following steps: a command and control network is abstracted into a topological structure diagram with nodes and edges; edge expansion is carried out on the command and control network to generate an original path function; associated edge replacement is carried out on the original path function to construct a new path function; variable sorting processing is carried out on the nodes and the edges of the topological structure diagram; the new path function is constructed into an ordered binary decision diagram (OBDD) according to a variable sorting order; and the two-terminal network reliability of the command and control network is recursively calculated from bottom to top according to the OBDD. The method provided by the invention has the advantages that the command and control network is abstracted into the topological structure diagram with the nodes and the edges, the edge expansion processing is carried out on the command and control network to generate the path function, and the OBDD is then constructed through the breadth-first search method, so that redundant calculation can be avoided through isomorphic sub-diagram identification, which achieves significant theoretical and practical application value for network two-terminal reliability analysis and researches.

The embodiment of the invention provides an image processing method and device, a computer device and a storage medium, and the method comprises the steps: obtaining a first feature map of a to-be-processed image; determining a final weight vector of the first feature map; determining a target normalization mode corresponding to the first feature map in a preset normalization set according to thefinal weight vector; and performing normalization processing on the first feature map by adopting the target normalization mode to obtain a second feature map.

The invention provides a real-time and offline business unified processing method and device based on a big data technology. The method comprises the following steps: writing business data collected in real time into a KAFKA cluster; processing and integrating the data in the KAFKA cluster based on a real-time service demand and an offline service range to obtain first detail data of the real-time service and second detail data corresponding to the offline service range; and performing analysis processing corresponding to the real-time service on the first detail data, wherein the second detail data is used for performing corresponding analysis processing on the subsequent offline service at a preset moment. According to the method and the device provided by the invention, redundant calculation of repeated access processing of partial data is avoided, and only one set of program needs to be developed to perform unified access processing on the data subjected to real-time analysis mining and offline analysis mining, so that the development time cost of an analysis mining function is reduced, and the form and the structure of the access data of the real-time service and the offline service after processing and integration are unified.

The invention provides a fast document clustering method. The method is realized by the following steps: 1, a group of key words is extracted from each document by word frequency statistics; 2, the document is expressed to be a corresponding dimensional congregation of index value, the contained key words of the document are in the characteristic space of the congregation; 3, a nerve element in a self organization mapping model is expressed as a vector in the characteristic space; 4, the documents are input in sequence, and the similarity between the documents and all nerve elements is calculated; 5, the nerve element with maximum accumulated value is the winner; the winner and neighbor nerve elements adjust weight in current document direction; 6, an individual dimension that the nerve element is matched with the input document is adjusted while the weight of other dimensions are weakened; 7, all the documents are input, and the method is over. The invention utilizes a self-organization mapping clustering model to renovate the links of document quantization expression and similarity calculation, thus the calculation efficiency is greatly improved under the condition that the number of the documents is same and the clustering quality is maintained.