Patents
Literature
57results about How to "Reduce the difficulty of optimization" patented technology
Efficacy Topic
Property
Owner
Technical Advancement
Application Domain
Technology Topic
Technology Field Word
Patent Country/Region
Patent Type
Patent Status
Application Year
Inventor
Weak supervision fine-grained image recognition method based on visual self-attention mechanism
ActiveCN111539469AImprove robustnessEnhanced Difficulty of OptimizationCharacter and pattern recognitionNeural architecturesFeature learningVisual perception
The invention discloses a weak supervision fine-grained image recognition method based on a visual self-attention mechanism. The method involves a student-model module, a teacher-model module and a classification-model module. The student-model and the teacher-model are combined through a Teacher-Student loop feedback mechanism based on Pairwise Approach sorting learning so as to form a self-attention region recommendation network, so that the relation between discriminative region positioning and fine-grained feature learning is enhanced, the discriminative region in the fine-grained image can still be accurately detected under the condition of lacking a target bounding box and a part marking point, and the recognition accuracy is promoted to be remarkably improved; meanwhile, a convolution layer is shared by the three modules, namely, the student-model, the teacher-model and the class-model, so that the model storage space is effectively compressed, the calculation cost is reduced, the method meets the real-time recognition task requirement, and the method is suitable for a large-scale real scene; and besides, a dynamic weight distribution mechanism is adopted in multi-task jointlearning to reduce the amount of artificially set hyper-parameters and enhance the robustness of the model, finally, the whole model is trained and learned in an end-to-end single-stage mode, and thenetwork optimization difficulty is reduced.
Owner:SOUTHEAST UNIV
Photovoltaic power prediction method based on deep convolution nerve network
ActiveCN108549929AImprove generalization abilityReduce the difficulty of optimizationForecastingNeural architecturesSupport vector machineNerve network
The invention discloses a photovoltaic power prediction method based on a deep convolution nerve network; the method comprises the following steps: using a variation modal decomposition algorithm to carry out modal decomposition for an obtained history photovoltaic power sequence, and decomposing the sequence into a plurality of frequency components and a remainder component; respectively arranging the components into data of a two dimensional format; using the frequency components of the two dimensional format as the input of a multichannel deep convolution nerve network model, predicting andoutputting a frequency component predicted value sum; using a single-channel deep convolution nerve network model to extract high order features of the remainder component in the two dimensional format, using the extracted high order features and meteorology data as the input of a support vector machine model, and predicting and outputting a remainder component predicted value; adding the frequency component predicted value sum with the remainder component predicted value, thus obtaining a photovoltaic power prediction result at a to-be-predicted moment. The method can obviously improve the photovoltaic power prediction precision, and can effectively guide the power grid in scheduling, thus ensuring the power system to stably and safely operate.
Owner:HOHAI UNIV
Improved particle swarm optimization-based mobile robot path planning method
InactiveCN107065876AIncrease diversityAvoid stagnationArtificial lifePosition/course control in two dimensionsArtificial intelligencePlanning method
The invention discloses an improved particle swarm optimization-based mobile robot path planning method. According to the method, a global optimal update strategy of disturbance is implemented in RDSAPSO (random-disturbance self-adaptive particle swarm optimization), small random disturbance is added to a global optimal position in a population, so that non-stagnation attributes in the RDSAPSO are obtained; an adaptive strategy is adopted to finely adjust three control parameters of particles in the RDSAPSO, so that global search capability and local search capability can be better balanced in the RDSAPSO, and a convergence guarantee parameter selection principle is provided for the RDSAPSO; and the constraint conditions of a path planning problem are solved through using a feasibility rule, and therefore, optimization difficulty can be decreased, and diversity can be decreased; and the developed RDSAPSO is utilized, so that an RDSAPSO-based path planning method framework is completed. With the method of the invention adopted, an effective method is provided for the path planning problem of a mobile robot.
Owner:NORTHWESTERN POLYTECHNICAL UNIV
Voice quality optimization method and device
ActiveCN105828356AMeet the continuityMeet needsWireless communicationBatch processingQuality optimization
The invention discloses a voice quality optimization method comprising the following steps: counting poor-quality cells; comprehensively analyzing the cause of each poor-quality cell based on a variety of data sources; clustering the causes of the poor-quality cells; and merging the poor-quality cells applicable to similar solutions based on the cause cluster, and carrying out batch-processing for the same cause. The invention further discloses a voice quality optimization device. By adopting the technical scheme of the invention, the problem that there is no systematic analysis means for traditional poor-quality problem processing and the efficiency is low is solved, poor-quality cells with voice quality problems in a mobile communication network can be counted out, the causes of the problems can be analyzed quickly and positioned accurately, a wireless network can be optimized quickly, and the voice quality of the mobile network is improved.
Owner:ZUNYI BRANCH OF CHINA MOBILE GRP GUIZHOU COMPANY
Combined X-band five-bit phase shifter based on active and inactive structures
The invention discloses a combined X-band five-bit phase shifter based on active and inactive structures, and mainly solves the problems of large gain error and low phase shift precision in the priorart. The phase shifter comprises filters (1), an active balun (3) and switches (4); multiple filters and switches are respectively provided and are alternately connected; the last switch is connectedwith an input end of the active balun; an output end of the active balun is connected with an orthogonal signal generator (2) and is used for generating positive and negative in-phase signals and positive and negative orthogonal signals; and an output end of the orthogonal signal generator is connected with a one-in-four circuit (5), and is used for selecting one path of signals from four paths ofsignals, namely the positive and negative in-phase signals and the positive and negative orthogonal signals to output, thereby implementing transformation of the signals among four quadrants. The phase shifter has the advantages of small gain error and high phase shift precision, and can be used in a radio-frequency integrated circuit in which a high-precision phase shifter is needed by a radio-frequency microwave phased array receiver.
Owner:XIDIAN UNIV
Self-shielding gradient coil of nuclear magnetic resonance spectrometer and design method thereof
The invention discloses a self-shielding gradient coil of a nuclear magnetic resonance spectrometer and a design method thereof. The invention is characterized in that, the self-shielding gradient coil comprises a main coil and a shielding coil which is coaxially arranged on the outer side of the main coil. The windings of the main coil are composed of a plurality of stages of main coil upper windings and main coil lower windings which are symmetrically arranged up and down. The centers of winding wire troughs, the circle radiuses of winding coils and turns per coil, corresponding to the maincoil upper windings and the main coil lower windings, are symmetrically arranged up and down in the one-to-one mode. The windings of the main coil are wound in the winding wire troughs and are arranged in a single-layer winding mode in the axial direction. Each stage of main coil winding corresponds to an independent driving current channel. According to the self-shielding gradient coil and the design method thereof, the series single-channel circuit design is changed into the matrix multi-channel circuit design, and the coil structure is simplified. The optimization difficulty of a target field is reduced. The probability that the actually generated gradient field is deformed and the linearity is not pure due to the winding translation and the interlayer dislocation is reduced.
Owner:ウーハン ジョンケ ニウジン マグネティック レゾナンス テクノロジー カンパニー リミテッド
Method and device for optimizing boiler based on least square support vector machine (LSSVM) combustion model
ActiveCN103150581AReduce the difficulty of optimizationImprove real-time performanceCharacter and pattern recognitionReference sampleCombustion
The invention provides a method for optimizing a boiler based on a least square support vector machine (LSSVM) combustion model. The method comprises the following steps of: acquiring an initial training sample set; acquiring LSSVM model parameters and a first characteristic matrix; acquiring a sample closest to a check sample from the training sample set as a reference sample, and performing row and column exchange on the first characteristic matrix to obtain a second characteristic matrix; calculating the inverse of the second characteristic matrix; acquiring partitioning parameters of the second characteristic matrix; acquiring a third characteristic matrix of an LSSVM model; calculating the inverse of the third characteristic matrix; acquiring the corrected LSSVM model parameters; assigning the third characteristic matrix to the first characteristic matrix, and assigning the inverse of the third characteristic matrix to the inverse of the first characteristic matrix; acquiring the combustion efficiency of the boiler by using the corrected LSSVM model; and performing combustion optimal control on a corrected combustion model. The invention also provides a device for optimizing the boiler based on the LSSVM combustion model. By the method and the device, the optimization difficulty of the combustion of the boiler in a power station can be reduced, and the real-time performance of optimization is improved.
Owner:ELECTRIC POWER RES INST OF GUANGDONG POWER GRID
Cell apoptosis algorithm-based dynamic recovery system and method of power distribution network
ActiveCN108923428AReduce the difficulty of optimizationSolve the problem that is prone to the curse of dimensionalityBiomolecular computersAc network circuit arrangementsPower gridComputer science
The invention provides a cell apoptosis algorithm-based dynamic recovery system and method of a power distribution network. The cell apoptosis algorithm-based dynamic recovery system comprises an information acquisition agent module, a load agent module, a repair team agent module, a data processing agent module, a restorability estimation agent module, a candidate restoring tree agent module, a synergic agent module and a control center agent module. According to the cell physiological property, the network reconstitution and repair problem is decomposed to a plurality of continuous steps, the optimization difficulty of each step is reduced, and the problem of optimization of a traditional method for many times to easily cause dimension difficulty is solved; according to a restorability estimation value, a recovery mode of a candidate restoring tree at each time frame is determined, a recovery target which is the more important to the stage is determined according to each mode, and the problems of optimizing complexity of the recovery process and long recovery time caused by multi-target optimizing are simplified during the recovery process; and fault recovery and repair are coordinated to consider a recovery scheme, the recovery efficiency is improved, the economic loss is reduced, and the cell apoptosis algorithm-based dynamic recovery system is applicable to the problem oflarge-area power failure of a multi-fault large-power grid system.
Owner:NORTHEASTERN UNIV
Tramcar timing energy-saving operation optimization method and system based on double-layer algorithm
ActiveCN111680413ALittle changeFull-line timing energy-saving optimization effect is goodData processing applicationsDesign optimisation/simulationCapacitanceIn vehicle
The invention discloses a tramcar timing energy-saving operation optimization method and a tramcar timing energy-saving operation optimization system based on a double-layer algorithm. The tramcar timing energy-saving operation optimization method comprises the following steps that a tramcar kinematics model, a vehicle-mounted super capacitor bank model and an energy consumption calculation modelare established; a speed curve is combined and generated according to the speed command given value; tramcar single-interval operation optimization is carried out based on a multi-objective genetic algorithm, and a Pareto leading edge curve about time-energy consumption is generated; whole-line timing energy-saving operation optimization of the tramcar is carried out based on a dynamic programmingalgorithm, and the total operation energy consumption is reduced on the basis that the total operation time of the whole line is not changed. According to the method, the running energy consumption of the train section can be effectively reduced, the optimal configuration of the train running timetable can be realized, and the train running cost is reduced.
Owner:NANJING UNIV OF SCI & TECH
Control arm assembly
ActiveCN104149568AReduce swing angleReduce the difficulty of optimizationResilient suspensionsPivoted suspension armsVehicle frameStress conditions
The invention relates to a control arm assembly which is applied to a vehicle independent suspension structure. The control arm assembly comprises two connecting arms and three spherical hinge structures, the two connecting arms form a V shape, and the three spherical hinge structures are arranged on the end portions of the two connecting arms respectively. One end portions of the two connecting arms are together connected to the first spherical hinge structure of the three spherical hinge structures, the other end portions of the two connecting arms are connected to the second spherical hinge structure and the third spherical hinge structure of the three spherical hinge structures respectively, the first spherical hinge structure is used for being connected with an axle wheel edge steering knuckle of a vehicle, and the second spherical hinge structure and the third spherical hinge structure are used for being connected with a vehicle frame of the vehicle. The control arm assembly can avoid movement interference between tires and a suspension and the vehicle frame in the vertical jumping process and the steering process of the tires as much as possible, a longer control arm structure also can be adopted to reduce the tilt angle of the control arm when the tires jump, the optimization difficulty of the control arm is reduced, and the stress condition of the control arm is improved.
Owner:XUZHOU HEAVY MASCH CO LTD
Robot relief grinding water turbine fixed guide vane tool
ActiveCN222570383UReduce the number of timesImprove work efficiencyHydro energy generationGrinding work supportsWater turbineControl engineering
The utility model belongs to the technical field of water turbine machining, and particularly relates to a robot relief grinding water turbine fixed guide vane tool which comprises a first adjusting mechanism and a second adjusting mechanism, the first adjusting mechanism and the second adjusting mechanism each comprise two bases, adjusting plates are vertically installed on the two bases, sliding supports are connected to the adjusting plates, and the sliding supports are connected to the first adjusting mechanism and the second adjusting mechanism. A bearing plate is connected between the two sliding supports, two adjusting bottom plates are symmetrically arranged on the bearing plate, and a vertically-arranged movable baffle is further arranged on the outer side of the first adjusting mechanism. The tool is suitable for robot relief grinding pumped storage and installation of the fixed guide vane of the mixed-flow water turbine. And through limiting of the tool, the number of times of manual teaching of the robot tool path is effectively reduced, and the working efficiency of automatic relief grinding of the guide vane is improved. Through limiting of the tool, the focusing accuracy of scanning of the three-dimensional camera is improved, then the optimization difficulty after the planning relief grinding path is generated is reduced, and the overall machining quality of the robot for relief grinding of the fixed guide vane is improved.
Owner:DONGFANG ELECTRIC MACHINERY
Air route generation method and device, equipment and storage medium
ActiveCN112444248AImprove smoothnessImprove convenienceNavigational calculation instrumentsPosition/course control in three dimensionsLocal optimumSimulation
The embodiment of the invention discloses an air route generation method and device, equipment and a storage medium, and the air route generation method comprises the steps: obtaining waypoints set bya user and flight parameters of an unmanned plane, and sequentially connecting the waypoints through a straight line to obtain an initial air route; determining the flight speed of the waypoints in the initial route by taking a preset speed constraint curve, a target speed in the flight parameters and the maximum acceleration as constraints; and taking the target speed, the maximum acceleration and a preset route deviation value as constraints, and adopting a preset smoothing strategy to carry out nonlinear optimization on the initial route after the flight speed is determined to obtain a target route. According to the embodiment of the invention, the initial route after the waypoint flight speed is distributed is used as the initial value of nonlinear optimization so that the optimization difficulty is reduced, the optimization result is prevented from falling into local optimum, the target route with good smoothness and fitting degree can be obtained, and the efficiency and the effect of executing plant protection operation according to the target route are ensured.
Owner:GUANGZHOU XAIRCRAFT TECH CO LTD
Aircraft trajectory control method fused with Gaussian process
PendingCN114200959AEfficientReliablePosition/course control in three dimensionsData setFlight vehicle
The invention discloses an aircraft trajectory control method fused with a Gaussian process. The aircraft trajectory control method comprises the following three parts: three-dimensional data collection, Gaussian process modeling and model prediction control. According to the method, model predictive control is used as a master control algorithm, meanwhile, a Gaussian process is fused to predict the process of the next moment, the model predictive control fused with the Gaussian process is combined with the motion speed track of the four-rotor aircraft, and the method has certain high efficiency, accuracy and real-time performance. And the trajectory tracking error of the aircraft can be effectively reduced. According to the aircraft trajectory model predictive control method, the optimization difficulty in the model predictive control process is reduced by sampling the data set in the training sample, and the stability of model predictive control is improved.
Owner:XIAN UNIV OF TECH
Optical fiber probe and length-adjustable variable-focus optical fiber OCT device based on beam expanding
ActiveCN112426128AWith adjustable focus positionEnable high-resolution imagingSurgeryDiagnostics using spectroscopyHigh resolution imagingBeam source
The invention provides an optical fiber probe and a length-adjustable variable-focus optical fiber OCT device based on beam expanding. The OCT device comprises a broadband light source, a laser lightsource, a first coupler, a reference arm light path structure, a sample arm light path structure and a spectral analysis component; the broadband light source is used for emitting a source light beam,and the source light beam is divided into a first light beam and a second light beam by the first coupler; the laser light source is used for emitting pump light; the first light beam enters the reference arm light path structure, and the second light beam is coupled with the pump light emitted by the laser light source and enters the optical fiber probe of the sample arm light path structure, sothat the focus of the optical fiber OCT device is adjustable; and the spectral analysis component is used for receiving and analyzing the first light beam reflected back from the reference arm lightpath structure and the second light beam reflected back from the sample arm light path structure. The variable-focus optical fiber OCT device adopts an all-fiber probe structure as a sample arm, is small and exquisite, is easy to package, has the characteristic of adjustable focus, and can realize high-resolution imaging beyond the fixed depth-of-field range of a traditional probe.
Owner:FOSHAN UNIVERSITY
Boiler optimization method and device based on least squares support vector machine combustion model
ActiveCN103150581BReduce the difficulty of optimizationImprove real-time performanceCharacter and pattern recognitionReference sampleCombustion
The invention provides a method for optimizing a boiler based on a least square support vector machine (LSSVM) combustion model. The method comprises the following steps of: acquiring an initial training sample set; acquiring LSSVM model parameters and a first characteristic matrix; acquiring a sample closest to a check sample from the training sample set as a reference sample, and performing row and column exchange on the first characteristic matrix to obtain a second characteristic matrix; calculating the inverse of the second characteristic matrix; acquiring partitioning parameters of the second characteristic matrix; acquiring a third characteristic matrix of an LSSVM model; calculating the inverse of the third characteristic matrix; acquiring the corrected LSSVM model parameters; assigning the third characteristic matrix to the first characteristic matrix, and assigning the inverse of the third characteristic matrix to the inverse of the first characteristic matrix; acquiring the combustion efficiency of the boiler by using the corrected LSSVM model; and performing combustion optimal control on a corrected combustion model. The invention also provides a device for optimizing the boiler based on the LSSVM combustion model. By the method and the device, the optimization difficulty of the combustion of the boiler in a power station can be reduced, and the real-time performance of optimization is improved.
Owner:ELECTRIC POWER RES INST OF GUANGDONG POWER GRID
Horizon automatic interpretation method based on minimum seismic waveform unit classification
The invention provides ahorizon automatic interpretation method based on minimum seismic waveform unit classification. The horizon automatic interpretation method includes the steps that 1, seismic data of a target interval are subjected to continuous enhancement pretreatment; 2, coarse-mesh horizon interpretation is carried out on target horizon; 3, a preliminary interpretation result obtained inthe step 2 is used as a control point of horizon time, and top and bottom time windows containing a plurality of complete seismic waveforms are obtained by shifting upwards and downwards; 4, seismicdata in the time windows are divided into minimum waveform units; 5, characteristic values of the minimum seismic waveform units are extracted; 6, a minimum seismic waveform unit classification modelof a random forest is constructed; and 7, a horizon automatic interpretation result of the whole area is optimized. According to the horizon automatic interpretation method, the difficulty of optimizing a horizon automatic recognition algorithm is lowered, meanwhile professional constraints are imposed, the reliability of the horizon automatic interpretation result is effectively improved, and good application prospects are achieved.
Owner:CHINA PETROLEUM & CHEM CORP +1
Remote sensing rotation detection network based on context feature adaptive weighted fusion
ActiveCN119445354ASolve the boundary problemimprove performanceCharacter and pattern recognitionNeural learning methodsEngineeringSelf adaptive
The invention discloses a remote sensing rotation detection network based on context feature adaptive weighted fusion. The method mainly comprises the following steps: (1) designing a one-dimensional Gaussian angle label encoder (GALE) to encode angles in an OBB label, and proposing angle classification loss (ACL); (2) GALE, ACL and YOLOv5 are combined, and a rotating target detector baseline GLR-YOLO is constructed; (3) designing a double-branch context adaptive modeling module (DBCAM) to perform context feature modeling, conversion and fusion so as to reconstruct a feature extraction network; and (4) providing an adaptive global feature aggregation and reweighting module (AGFAR), constructing an adaptive detection head (ADH) in combination with the detection head, and solving the problem of output interlayer feature inconsistency. The method is mainly used in the field of remote sensing image target detection, and has important research value in the fields of building early warning, traffic monitoring, national defense safety and the like.
Owner:SICHUAN UNIV
Graph optimization method and mask preparation method
ActiveCN110119063AReduce widthIncrease widthOriginals for photomechanical treatmentGraphicsAlgorithm
The invention discloses a graph optimization method. The graph optimization method comprises the following steps: providing a to-be-optimized graph, wherein the to-be-optimized graph comprises a firstgraph and a second graph and the first graph and the second graph are different in size; carrying out OPC processing on a first graph with low adjustability; performing OPC processing on the second graph; and meanwhile, adaptively adjusting the first graph. Thus, the width of the final first graph is greater than the minimum size of the mask on the basis of ensuring that the final first graph isrelatively accurate, the condition that the width of the final first graph is close to the minimum size of the mask is avoided, and the graph optimization difficulty is also reduced. Therefore, the quality of the mask plate can be improved through the mask preparation.
Owner:SEMICON MFG INT (SHANGHAI) CORP +1
Multi-joint mechanism trajectory planning method, device, electronic equipment and storage medium
ActiveCN113043277BSolve the problem of joint angle exceeding the limitSolve the problem of over-limit angleProgramme-controlled manipulatorLinear motionSimulation
The present invention relates to a trajectory planning method, device, electronic equipment and storage medium of a multi-joint mechanism; the method includes: determining the first point in the movement process of the terminal of the multi-joint mechanism from the starting point A to the ending point B based on the joint space trajectory planning method; Joint trajectory planning data; based on the generalized inverse matrix method, the second joint trajectory planning data for the end of the multi-joint mechanism moving along a straight line from the starting point A to the ending point B is determined; according to the preset motion advance, based on the second joint trajectory Planning data to plan the trajectory of each joint of the multi-joint mechanism; judge in real time whether the angle value of each joint in the trajectory planning exceeds the limit value of the joint angle, and if yes, optimize the joint posture in the trajectory of each joint of the multi-joint mechanism. The invention realizes the rectilinear motion of the end of the multi-joint mechanism, solves the problem that the joint angle exceeds the limit during the motion, and can realize the purpose of optimization for various initial configurations, thereby reducing the difficulty of optimization.
Owner:ZOOMLION HEAVY IND CO LTD
Spacecraft escape trajectory intelligent planning method
ActiveCN114415730AReduce the difficulty of optimizationImprove computing efficiencyInternal combustion piston enginesPosition/course control in three dimensionsTime domainEngineering
The invention discloses a spacecraft escape trajectory intelligent planning method, which comprises the steps of splitting a spacecraft long-period pursuit game problem into a plurality of short-period pursuit game sub-problems according to a rolling time domain strategy; in each short period, according to the spacecraft orbit state at the corresponding initial moment, a spacecraft pursuit problem optimization model is constructed, the optimal pursuit parameter for pursuit of the spacecraft is predicted and evaluated by using the completely trained deep neural network, and then an escape trajectory planning index function is updated according to the real-time orbit state and the predicted pursuit information. And obtaining the current optimal escape pulse maneuvering speed increment by using a sequential quadratic programming algorithm. The problems that in the prior art, solving is difficult, solving efficiency is low and result robustness is poor when the optimal escape strategy of the spacecraft pulse pursuit game problem is solved are solved.
Owner:NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
Neural network model optimization method based on class expansion learning
ActiveCN110796260AImprove optimization effectImprove the ability to distinguishMachine learningNeural architecturesData setEngineering
The invention discloses a neural network model optimization method based on class expansion learning. The method specifically comprises the following steps: obtaining a data set which is used for training and contains various types of samples, and defining an algorithm target; extracting features of each type of pictures in the data set by using a universal model, and evaluating the error-prone degree of each type according to the distribution condition of each type of features; adding several types of data with the highest error-prone degree into a training pool, and optimizing the neural network by using the data in the training pool; after the optimization is completed, adding the remaining data with the highest error-prone degree into a training pool, expanding the categories in the training pool, and further optimizing the neural network by using the training pool on the basis of the neural network obtained by the last training; and continuously performing class expansion on the training pool until the whole data set enters the training pool to obtain a final optimized neural network model. The method is suitable for neural network model optimization based on multi-class datasets in supervised learning, and has good effect and robustness for various complex conditions.
Owner:ZHEJIANG UNIV
Quantum classical mixed image classification method based on group isovariant and delayed aggregation
ActiveCN121962777Aovercome lossGuaranteed sensitivityNeural learning methodsFeature vectorFeature fusion
The invention discloses a quantum classical mixed image classification method based on group isovariant and delayed aggregation. The method comprises the following steps: lifting a convolution module, inputting a target image, initializing a convolution kernel and rotating the convolution kernel, and carrying out convolution and stacking on the target image to obtain an output image; the group convolution module is used for converting the output image of the convolution module into an image sample and carrying out convolution and reverse remodeling operation to obtain an output image; the group pooling and flattening module is used for carrying out adaptive average pooling on an output image of the group convolution module, compressing the output image into a feature vector and obtaining an output feature vector; the group equivariant quantum feature processing module encodes the feature vectors to quantum bits according to the feature vectors output by the group pooling and flattening module; and the delay aggregation and output module measures the quantum bits, aggregates and classifies measurement results, and completes classification of target images. The method provided by the invention has strong nonlinear feature fusion capability, and can realize more accurate image classification.
Owner:NANJING UNIV OF INFORMATION SCI & TECH
Orchard multi-robot task allocation optimization method based on hierarchical path reconstruction
PendingCN121879360AReduce the difficulty of optimizationquality improvementVehicle position/course/altitude controlPosition/direction controlMathematical modelOrchard
The invention discloses an orchard multi-robot task allocation optimization method based on hierarchical path reconstruction, and the method comprises the steps: S1, building a multi-target mathematical model, optimizing the maximum completion time and total energy consumption, including robot load, speed, energy consumption and battery capacity constraints, and employing VLDIM to generate an initial solution population; s2, optimizing the task sequence of each path by using DRRM in the microscopic layer; s3, robot-level DRRM optimization, TRRM task redistribution, CRRM charging path reconstruction and environment selection are executed iteratively; and S4, splitting the longest path by using the SRRM, redistributing the longest path through the MILP1, and outputting a non-dominated solution set. Through hierarchical coding, multi-stage collaborative optimization and a special reconstruction mechanism for core constraints, the task allocation efficiency and scheme quality of a multi-robot system under complex constraints are significantly improved, and the method is suitable for agricultural automation and warehouse logistics scenes.
Owner:ZHENGZHOU UNIV
Global optimization preprocessing method for intelligent manufacturing production scheduling
InactiveCN110826013AReduce variable sizeGuarantee equivalenceForecastingManufacturing computing systemsGlobal optimizationOptimization problem
The invention discloses a global optimization preprocessing method for intelligent manufacturing production scheduling. The global optimization preprocessing method comprises the following steps: pre-solving a mixed integer linear programming problem, and reducing a search domain for the mixed integer nonlinear programming problem; the pre-solving of the mixed integer linear programming problem comprising the following steps: (1) performing variable selection based on a Fourier-Motzkin elimination method; (2) performing a redundancy constraint elimination method based on a Fourier-Motzkin elimination method; and (3) pre-solving a linear programming problem based on a Fourier-Motzkin elimination method. According to the global optimization preprocessing method for intelligent manufacturingproduction scheduling, aiming at solving the problem of solving a large-scale engineering decision optimization problem in practice, two types of model preprocessing methods with universality are adopted, so that the solving performance is effectively improved, the method can be directly applied to engineering decision problems with complex structures, and the real-time performance and the qualityof complex production scheduling decisions are greatly improved.
Owner:苏州圣典企业管理咨询有限公司
Two-stage anchor-based dynamic video abstraction method
ActiveCN113505266AReduce computationDetailed classificationVideo data clustering/classificationSpecial data processing applicationsNetwork structureEngineering
The invention relates to a two-stage anchor-based dynamic video abstraction method. The method comprises the following steps: (1) constructing an anchor-based neural network, wherein the neural network comprises an extraction network, a self-attention mechanism network, a two-stage network, namely a region recommendation network and a regression classification network; and (2) determining types and lengths of anchors in the region recommendation network and the regression classification network, completing classification of positive and negative samples, training the anchor-based neural network constructed in the step (1), fixing trained parameters, obtaining two-stage anchor-based neural network models, and performing dynamic video abstraction by using the obtained neural network models. The brand new network structure provided by the invention can process all video frames in parallel, so that the training time is reduced, the final output accuracy is effectively improved, and the network structure is relatively simple and has low requirements on computing power.
Owner:NANJING UNIV OF POSTS & TELECOMM
Optical system free-form surface optimization method, device and computer storage medium
ActiveCN113204113BReduce the difficulty of optimizationImprove efficiencyOptical elementsMatrix opticsImaging quality
Owner:CHANGCHUN INST OF OPTICS FINE MECHANICS & PHYSICS CHINESE ACAD OF SCI
A kind of nuclear magnetic resonance spectrometer self-shielding gradient coil and its design method
The invention discloses a self-shielding gradient coil of a nuclear magnetic resonance spectrometer and a design method thereof. The invention is characterized in that, the self-shielding gradient coil comprises a main coil and a shielding coil which is coaxially arranged on the outer side of the main coil. The windings of the main coil are composed of a plurality of stages of main coil upper windings and main coil lower windings which are symmetrically arranged up and down. The centers of winding wire troughs, the circle radiuses of winding coils and turns per coil, corresponding to the maincoil upper windings and the main coil lower windings, are symmetrically arranged up and down in the one-to-one mode. The windings of the main coil are wound in the winding wire troughs and are arranged in a single-layer winding mode in the axial direction. Each stage of main coil winding corresponds to an independent driving current channel. According to the self-shielding gradient coil and the design method thereof, the series single-channel circuit design is changed into the matrix multi-channel circuit design, and the coil structure is simplified. The optimization difficulty of a target field is reduced. The probability that the actually generated gradient field is deformed and the linearity is not pure due to the winding translation and the interlayer dislocation is reduced.
Owner:ウーハン ジョンケ ニウジン マグネティック レゾナンス テクノロジー カンパニー リミテッド
End-to-end sign language translation method and system
PendingCN113657257AEnhance feature expressionReduce the difficulty of optimizationCharacter and pattern recognitionKnowledge representationFeature extractionAlgorithm
The invention discloses an end-to-end sign language translation method, and belongs to the technical field of sign language translation. The method comprises the following steps: 1) acquiring a sign language video and a corresponding target annotation sequence and target text sequence; 2) establishing a sign language translation model; extracting visual features of the sign language video by a feature extractor and coding the same by a mask coder, and dividing a coding result into three branches for decoding; in the first branch, predicting a word boundary firstly by a boundary predictor, and then predicting an annotation sequence by an auxiliary annotation decoder in combination with an output result of the boundary predictor; in the second branch, linearly mapping the coding result and then using the coding result as the input of a CTC decoder, to generate a prediction annotation sequence; in the third branch, using the coding result as the input of a wait-k decoder, and outputting a prediction text sequence; and 3) carrying out feature extraction and coding on a sign language video to be translated by utilizing the trained sign language translation model, taking a coding result as the input of the wait-k decoder, and generating a prediction text sequence as a translation result.
Owner:ZHEJIANG UNIV
A Neural Network Model Optimization Method Based on Class Expansion Learning
ActiveCN110796260BImprove optimization effectImprove the ability to distinguishNeural architecturesMachine learningData setNetwork on
The invention discloses a neural network model optimization method based on class expansion learning. The specific steps are as follows: Obtain a data set containing samples of various categories for training, and define the algorithm goal; use a general model to extract the features of each type of picture in the data set, and evaluate the error-prone of each category according to the distribution of each type of feature Add the most error-prone types of data to the training pool, and use the data in the training pool to optimize the neural network; after the optimization is completed, add the remaining types of error-prone data to the training pool to expand the categories in the training pool , and use the training pool to further optimize the neural network on the basis of the neural network obtained in the previous training; continuously expand the training pool until the entire data set enters the training pool, and obtain the final optimized neural network model. The invention is applicable to the neural network model optimization based on multi-category data sets in supervised learning, and has better effect and robustness in the face of various complex situations.
Owner:ZHEJIANG UNIV
Positioning method of robot shovel grinding tool for movable guide vane of water turbine
The application belongs to the technical field of water turbine processing, and particularly relates to a positioning method of a robot shovel-grinding water turbine movable guide vane tooling. The positioning method comprises the following steps: installing a short shaft end positioning mechanism; installing a long shaft end positioning mechanism to position the Y-axis direction of a guide vane coordinate axis; hoisting the guide vane to make the short shaft end face of the guide vane adhere to the positioning end face of the short shaft end positioning mechanism tooling, and first perform axial coarse adjustment, then perform axial fine adjustment after the axial coarse adjustment is completed, and position the X-axis direction of the guide vane coordinate axis; position the Z-axis direction of the guide vane coordinate axis; install two jacks to support the left and right ends of a guide vane blade body profile, and adjust the guide vane to limit the rotation of the X-axis of the guide vane coordinate axis. The method can reduce the adhesion degree of the guide vane and the tooling from the level of 1mm-10mm to 0.1mm-1mm, and effectively reduces the number of manual teaching of the robot cutter track through the limiting of the tooling, and improves the working efficiency of the automatic shovel-grinding of the guide vane.
Owner:DONGFANG ELECTRIC MACHINERY