89results about How to "Improve optimization speed" patented technology

The invention discloses a head attitude tracking device and method based on incomplete characteristics. Normalized processing on a video stream acquired in real time is carried out by a head attitudedetection system based on computer vision, then whether a face portion exists is determined through characteristic confirmation by a model after CNN training, if face detection fails, and the incomplete facial characteristic part is completed by a trained confrontation network model; the position of the face is tracked in the video, the facial characteristic point position is marked by a sparsely-constrained model, the PnP algorithm is utilized to acquire Pitch, Yaw, Roll three angle values [thetax thetay thetaz] of the head attitude, and the angle values are then sent to a follow-up holder. The method is advantaged in that the accurate head position and attitude information can be provided, and high timeliness and good robustness are realized.

The invention relates to an electromyographic signal gait recognition method based on particle swarm optimization and a support vector machine. A particle swarm optimization algorithm is utilized to optimize a penalty parameter and a kernel function parameter of the support vector machine so that the performance of the support vector machine can be optimized, and effective recognition and classification are achieved. Firstly, wavelet modulus maximum denoising is carried out on collected lower limb electromyographic signals; secondly, time domain feature extraction is conducted on the electromyographic signals after denoising is carried out to obtain feature samples; thirdly, parameter optimization is carried out on the support vector machine by means of the particle swarm optimization algorithm to obtain a set of optimal parameters with minimal errors, and a classifier is constructed; at last, a feature sample set of the electromyographic signals is input to the classifier, and then classification and recognition are conducted on gait states. According to the method, both accuracy and adaptivity of classification are taken into consideration, the computational process is simple and efficient, and the method has broad application prospects in the field of lower limb motion state recognition.

The invention discloses a robot scene self-adaptive pose estimation method based on RGB-D camera. scene two-dimensional color image information of adjacent frames and space depth information corresponding to pixels of the two-dimensional color image are acquired based on a RGB-D camera; when the two-dimensional color image feature points are sufficient, ORB operators are adopted to extract features, the matching strategy provided by the invention is adopted to carry out accurate matching, and three-dimensional pose estimation is solved based on the pose estimation algorithm of the matching feature points; when the feature points are insufficient, the improved ICP algorithm provided by the invention is adopted to solve the three-dimensional pose estimation; then, a complete switching criterion is designed to fuse the two pose estimation methods; finally, pose estimation obtained through the two methods is optimized through a light beam adjustment algorithm, and finally smooth and accurate three-dimensional pose estimation is obtained. The three-dimensional pose estimation algorithm has the outstanding advantages of high robustness, high precision, small calculation amount, adaptability to different scenes and the like.

The invention discloses a multilayered liquid crystal display weight optimization device based on visual system characteristics. The multilayered liquid crystal display weight optimization device comprises a spatial light modulator array, a polarizing film array, backlighting equipment, and a control terminal, which are sequentially arranged in the human eye sight direction, wherein the spatial light modulator array is used for modulating transmittance of polarized light accessing the human eyes; the polarizing film array is used for modulating the bias light accessing the spatial light modulator array to be the polarized light; the backlighting equipment is used for providing backlight with uniform luminance to the spatial light modulator array; the control terminal is used for controlling the transmittance of RGB (Red, Green and Blue) channels of each pixel on the spatial light modulator array, so as to realize near-eye three-dimensional display. The invention further discloses a multilayered liquid crystal display weight optimization method based on the visual system characteristics. The method comprises the following steps: putting forward that the central area and the marginal area of a reconstructed light field possess different weight ranges, meanwhile combining with focused convergence conflict resolution conditions, computing the final weight angle, and enabling the optimization result to be closer to the truth. The multilayered liquid crystal display weight optimization device based on the visual system characteristics and the multilayered liquid crystal display weight optimization method based on the visual system characteristics disclosed by the invention obviously increase the optimization speed, and enhance the real-time performance of the optimization algorithm.

The invention discloses a method of wheel torque distribution for a multi-axle drive electric vehicle based on online optimization of driving energy. The method comprises the following steps that thevehicle parameters are obtained and required torque difference values between a left side vehicle body and a right side vehicle body are obtained, after a total torque difference value is applied to the left side vehicle body or the right side vehicle body separately, whether the required torque of the one-side vehicle body is larger than the maximum torque which all drive motors of the one-side vehicle body can output is judged, according to an objective function and a constraint condition, the initial optimization of data is carried out, and first distribution of drive torque of each wheel is carried out; the slip ratio of each drive wheel is calculated, and a fitting coefficient is obtained by fitting a characteristic curve of an electric drive system; and combined with the fitting coefficient, according to the following optimal objection function, the data optimization is carried out once again, and drive torque of each wheel at the optimal performance of a whole vehicle is obtained.

The invention discloses a demand side response-based microgrid optimal unit and time-of-use electricity price optimization method, which comprises the following steps: (1) acquiring supply and demandhistorical data of a microgrid, including internal load data of the microgrid, a supply situation of distributed regenerable energy and basic information of all types of energy supply units; (2) analyzing outputs of all energy supply systems by using the historical data so as to analyze a supply and demand situation in the microgrid; (3) setting an optimized target function subjected to distributed optimization; (4) performing optimized solving by using a particle swarm search algorithm and an interior point method. According to the demand side response-based microgrid optimal unit and time-of-use electricity price optimization method disclosed by the invention, the supply situation of the regenerable energy of the microgrid and a demand side response under the excitation of time-of-use electricity price are organically combined, and an optimal unit combination of the microgrid and a time-of-use electricity price strategy are subjected to the distributed optimization through the particle swarm search algorithm and the interior point method, so that the problems of extremely high operation cost of the microgrid and too large lack of internal loads are effectively relieved, and the electricity use demand of a user can be well met.

The invention discloses a planetary exploration landing trajectory comprehensive optimization method, and relates to a planetary landing trajectory optimization method, belonging to the field of deepspace exploration. The method is implemented by the following steps: establishing a detector landing dynamics equation; calculating the probability of collision between a detector and planetary surface obstacles; setting up a planetary exploration landing trajectory comprehensive optimization problem by taking the uncertainty of fuel consumption performance and planetary landing as well as the complex terrain conditions of the obstacles existing on the planet surface into account; solving the planetary exploration landing trajectory comprehensive optimization problem, and completing the comprehensive optimization of the planetary exploration landing trajectory by taking the uncertainty of fuel consumption performance and planetary landing as well as the complex terrain conditions of the obstacles existing on the planet surface into account so as to realize safe and accurate landing. In addition, in the calculation of the collision probability, the description of the degree of threat tothe detector under the condition of uncertainty is more accurate. By simplifying a collision probability formula, the calculation amount is reduced and the optimization speed is accelerated under thepremise of guaranteeing accuracy.

The invention discloses a handover parameter optimization method. According to a measurement value reported by user equipment (UE), a network side obtains a cumulative density function CDF (i) of corresponding measured values of a cell handover region, and according to the area of a region enclosed by the cumulative density function curve CDF (i) of corresponding measured values of the cell handover region and a threshold value curve RLFthrd (i) of the measured values, the network side determines the degree of the cell fault when the cell has mobility fault; and according to the fault degree of the cell, the network side revises corresponding handover parameters so as to decrease the area of the region enclosed by the curve RLFthrd (i) and the curve CDF (i). The invention further discloses a handover parameter optimization system correspondingly. On the basis of the statistic result reported by the UE, the parameters needed to be revised and the revision degree of the parameters are determined, thus the optimization speed can be increased, and simultaneously the self-optimization performance can also be improved.

The invention provides a low-loss power distribution network optimization and reconfiguration method based on a genetic algorithm. According to the method, firstly, original progenitor chromosome populations are generated and encoded; whether lonely islands exist in the original progenitor chromosome populations or not is judged; chromosomes with the lonely islands are removed, and progenitor chromosome populations are obtained; the adaptation degrees of each chromosome in the progenitor chromosome populations are calculated, and sequencing is carried out according to the values of the adaptation degrees; whether the adaptation degrees meet design requirements or not is judged; if the adaptation degrees do not meet the design requirements, duplication, intersection and variation of the chromosome populations are carried out; the obtained chromosome populations are subjected to Elitism processing, i.e., before next iteration, the chromosomes with the best adaptation degree in the iterated progenitor chromosome populations are put into the next iteration; the iteration is repeated until iteration stopping conditions are met; and the optimum value is output. The low-loss power distribution network optimization and reconfiguration method based on the genetic algorithm provided by the invention has the advantages that Elitism is introduced; convergent results are more accurate; the problem of inconsistent convergent results of the genetic algorithm is solved; and the optimization efficiency is high.

The invention is suitable for unmanned planes, and provides a three-dimensional environment modeling method for an unmanned plane. The method comprises the steps: obtaining the point features and line features of key image frames, and carrying out the division of a pattern region; selecting one frame closest to the key image frames as the target key frame, searching a matching point and line in the target key frame, employing an epipolar geometry triangle to obtain 3D points and lines, and completing the generation and updating of face features through combination with the pattern region; enabling the obtained multilayer features to be effectively combined in a three-dimensional space, wherein the multilayer features comprise the point features, line features and face features; and employing dual quaterion to achieve the unified description of the multilayer features. The invention also provides a three-dimensional environment modeling system for the unmanned plane. The method and system can further provide basic research results for the research and development of novel dedicated dual quaterion resolving hardware, can improve the optimization speed, also can reduce the wrong matching rate, and improve the precision and robustness of a map.

The invention discloses a soft measurement method for the key parameters in the edible fungus fermentation production process, and the method is used for solving the problem of online estimation of the key biochemical quantity which is difficult to measure in real time in the edible fungus fermentation process. The method comprises the following steps: firstly, by analyzing the process mechanism of an edible fungus fermentation process, selecting a proper auxiliary variable and establishing a training sample database according to the historical tank batch data; constructing a least squares support vector machine soft measurement training sample database by combining the main variables and the auxiliary variables of the current to-be-predicted tank batch fermentation process with the main variables and the historical auxiliary variables of the historical fermentation process, and constructing a soft measurement model corresponding to the soft measurement training sample database, and then optimizing the normalization parameter gamma and the kernel parameter sigma 2 in the soft measurement model by means of a grey wolf algorithm, and establishing a least squares support vector machine soft measurement model based on the grey-wolf optimization, and finally obtaining a corresponding key biochemical parameter prediction value. According to the method, the grey-wolf optimization algorithm for simulating the grey wolf behaviors is adopted, the structure is simple, the parameter setting is small, the global searching capability is high, and the gradient information is not considered.

The invention discloses a method for optimizing a freeform surface optical system in combination with a surface form and a field-of-view optimization strategy. The method uses the respective coefficients of a Zernike standard polynomial representing an optical system wave aberration as an evaluation basis, develops a optimization process by a successive approximation strategy combining surface form optimization and field of view optimization, and comprises steps of: firstly, selecting an initial small field of view, searching out the maximum item of system aberration targetedly in each optimization step according to an established XY polynomial and a polynomial relation model, selecting a corresponding XY polynomial item in the freeform surface as a new-added variable to optimize and balance the aberration, and synchronously adjusting the weight of each field of view according to the wave aberration rms values of respective fields of view of the system; after a structure satisfying a performance index is optimized and acquired within a small field of view range, gradually expanding the field of view and repeating the optimization steps until optical system structure parameters within all field of view range are obtained. The method has a large field of view, fast optimization speed, aberration pertinence and guidance.

The invention discloses a classification optimization method used for equivalent model parameters of electrical power system and based on trace sensitivity. Firstly, electrical power systems considering load dynamics are divided into groups and gathered, and a system dynamic equivalent model structure and a parameter initial value are determined; dynamic responses of an original system and an equivalent system are calculated under a presupposed fault with an implicit trapezoidal integration method; trace sensitivity of active power, reactive power and boundary node voltage of all call wires to an equivalent model parameter are calculated according to a jacobian matrix output in a calculation process; the trace sensitivity is subjected to normalization processing, average values of the sensitivity are sorted by size, and a static parameter and a dynamic parameter are divided; the static parameter is subjected to least squares optimization according to a steady state time interval response curve and the trace sensitivity, and then a dynamic response time interval response curve is subjected to predicting correction according to the static parameter correction quantity and the trace sensitivity, and the dynamic parameter is optimized; and an equivalent model optimal parameter which meets the requirement of the accuracy can be obtained after a little optimization iteration.

The invention relates to the technology field of Internet of Things, and particularly discloses a network optimization method based on LTE (Long Term Evolution) wireless and ultra-low-power wide-areanetwork (LPWAN) Internet of Things wireless convergence. The method comprises the following steps of: introducing a token mode to simplify a complex protocol; adopting a hardware decoding mode to improve the speed of a network protocol conversion; using a caching mechanism to avoid data fragments in the LPWAN network to improve the schedulability; establishing a routing stress model for LPWAN multi-hop communication based on a traditional radio frequency model; regarding the communication route between a source node and all multicast nodes as a multicast tree, wherein the multicast tree with the smallest integrated routing pressure is the optimal path under the constraint condition; determining the optimal number of communication nodes; and establishing a set of access routing alternativepaths between the multicast source nodes and all multicast nodes. According to the inherent relationship between the LTE wireless and the ultra-low-power wide-area network (LPWAN) Internet of Things wireless system, the software/hardware method is used to improve the network convergence tightness, and a wireless network routing node pressure model is established, thereby optimizing the pressure ofrouting nodes, improving the response speed of an important data network, and significantly improving the working efficiency of a wireless network.

The invention discloses a control method based on a prediction model and a controller. The prediction model on which the control method is based is a normalized prediction model, the normalized prediction model is a functional relationship between variables having no unit and ranging from 0 to 1, according to a predicted value of the measurement quantity of a controlled object normalized at the second time, when the control amount increment normalized at the second time is optimized, a numerical range of each coefficient of an established constraint equation is relatively stable, reduction ofthe number of iterations of the optimization algorithm is facilitated, so the optimization process takes less time, the optimization speed is improved, real-time control of the underlying controller can be realized, and thereby the control effect of the underlying controller on the controlled object can be optimized.

The invention discloses a deep learning hyper-parameter tuning improvement method based on Bayesian optimization. The method comprises the steps that a fast optimal Latin square experimental design algorithm is used to generate an initial point set, and the generation rate of effective evaluation points is improved; parallel computing is used to acquire the response of the initial point set in a computer, and the response of multiple points on an objective function is calculated at the same time to construct a data set; and a Bayesian optimization iterative process is initiated, and a parallelcomputing method is used in the computing process to accelerate the whole optimization process. According to the invention, the defects of long time consumption and large performance fluctuation of atraditional Bayesian optimization algorithm can be effectively solved; the optimization speed is accelerated through parallel computing; and the optimization efficiency is significantly improved.

The invention relates to an electromagnetic mechanism static characteristic optimization method based on an adaptive weight multi-objective differential evolution algorithm, which comprises the following steps: S1, determining parameters to be optimized and static characteristic related indexes of an electromagnetic system; S2, determining the upper limit and the lower limit of each static characteristic optimization parameter according to the product material and the processing technology of the electromagnetic mechanism, and simultaneously determining an additional constraint index related to the static characteristic optimization parameter; S3, obtaining an initial population; S4: obtaining the Pareto solution set distribution of the optimization objective function; S5, selecting different mutation strategies and crossing strategies according to the number of iteration times of the current optimization parameter population and the Pareto dominance relationship of each optimization parameter, generating a progeny population, repeating the operation of step 4 to the whole population, and adopting a selection strategy considering the niche sorting result to control the population size; S6, obtaining the Pareto solution set distribution of the optimization objective function; S7, selecting a group of optimization parameters from the obtained groups of optimization parameters, and taking the group of optimization parameters as the optimization design parameters of the electromagnetic mechanism.

The embodiment of the invention discloses an application program optimization method and device, electronic equipment and a storage medium. The method comprises the steps that user data of an application program are acquired, statistics is conducted on the retained user data, and the retained user data comprise retained users meeting the information pushing condition and retained users meeting theinformation non-pushing condition in at least two time periods; determining a change parameter of an application evaluation value according to the reserved user data and a pre-established relationalgraph between the number of users and the application evaluation value; and determining an optimization type of the application program according to the change parameter so as to indicate the optimization of the application program. According to the embodiment of the invention, the application program optimization efficiency and accuracy can be improved.

The invention discloses a DC receiving-end system transient load shedding control optimization method based on sensitivity analysis. The method comprises the steps of 1, acquiring a DC receiving-end system model and a parameter; 2, based on the acquired DC receiving-end system model and the parameter acquired in step 1, calculating a system transient stability margin and a sensitivity matrix in acurrent load shedding rate; 3, according to the calculated system transient stability margin and the sensitivity matrix in the current load shedding rate, establishing a mathematical model of a DC receiving-end system transient load shedding control optimization problem; and 4, for the established mathematical model of the DC receiving-end system transient load shedding control optimization problem in the step 3, solving the load shedding control optimization problem according to an approximate linear planning method based on sensitivity analysis. The DC receiving-end system transient load shedding control optimization method has advantages of effectively reducing calculation amount, improving optimizing speed, and reducing economic cost in load shedding as possible under a precondition that the system has the preset transient stability margin.

The invention discloses a method, a device and a system for simulating the process of a workflow middleware. According to a user request, a process template which is specified by the user is obtained from a database; the obtained process template is traversed according to the breadth first search algorithm, and simulating data is obtained; the simulating data is stored in a first extensible markup language (XML) file; a workflow interface is invoked, and the simulating data which is stored in the first XML file is simulated, thus obtaining process running data; the obtained process running data is stored; the stored process running data is analyzed, thus obtaining analysis results; and the analysis results are stored in a second XML file. According to the method, the device and the system for simulating the process of the workflow middleware disclosed by the invention, the process template is traversed according to the breadth first search algorithm, the obtained simulating data covers all possible paths of the process template, the simulating process can be finished automatically under the driving of a simulating strategy, and business data does not need to be input, thus being suitable for ordinary users who are not programmers.

The invention relates to a target tracking method based on pigeon intelligent optimization Kalman filtering parameters. The method is implemented by the following steps: step 1, getting a real curve; step 2, getting an original curve; step 3, initializing pigeon-inspired optimization parameters; step 4, designing a cost function; step 5, performing optimization by use of a map compass operator; step 6, performing optimization by use of a landmark operator; and step 7, getting a result and verifying the result. By using the method, the parameters of a Kalman filter can be optimized better, the status value of a target at next moment can be predicted more quickly, and effective guarantee is provided for tracking a dynamic target.

The invention belongs to the technical field of engineering blasting, and particularly relates to a method for predicting the average diameter of jet flow of a metal shaped charge liner and application thereof. The method comprises: on the basis of considering material attributes, constructing a jet flow average diameter-material attribute relational expression according to the stress state in thejet flow forming and moving process of the shaped charge liner and the action mechanism of the material attributes on the jet flow form in the jet flow forming and moving process; substituting the material attribute of the metal material into the jet flow average diameter-material attribute relational expression to obtain a jet flow average diameter-material attribute equation; and substituting the material attribute of the to-be-measured material into the jet flow average diameter-material attribute equation to obtain an average diameter prediction value of the shaped charge liner shaped charge jet flow. According to the prediction method provided by the invention, the influence of material attributes is fused into the prediction process, the problem that the energy-gathered jet flow form difference of different materials cannot be determined by a traditional theory is solved, and the accuracy of a jet flow average diameter prediction result is improved.