129 results about "Experimental validation" patented technology

The invention discloses an image classification method based on MapReduce, on Hadoop platform, firstly, using the MapReduce frame for parallel extracting image SIFT characteristic; using the MapReduce frame for sparse coding for extracted SIFT characteristic of each image and obtaining the corresponding sparse vector of the image, generating the sparse characteristic of the image; then, based on the sparse characteristic of the image, using MapReduce frame for training the decision-making tree, generating the random forest aiming at the image characteristic set; using MapReduce combined with the random forest for classified counting each image. Through experimental verification, the image classification method based on MapReduce can obviously raise the classification speed while guaranteeing not lower than single platform classification precision.

The scalable digital predistortion system provides a behavioral model that can be used to model and compensate for the nonlinear distortions of power amplifiers and transmitters. The predistorter and update algorithms make the model/DPD scalable in terms of signal bandwidth and average power, allowing for low complexity update following changes in the signal's bandwidth and/or power level. Experimental validation carried on a 300 Watt Doherty power amplifier shows that the scalable model and the predistorter architecture achieve performance similar to their conventional counterpart. However, the present model/predistorter requires the update of up to 50% fewer coefficients than the conventional model/predistorter.

The invention discloses a hardware based network simulation system. The system comprises a network simulation unit and multiple simulation nodes, wherein the network simulation unit adjusts the topological graph of the simulation nodes in a simulation network according to users, and configures characteristic parameters of a simulation link connected to the simulation nodes; and based on the adjusted network topological graph and the configured link characteristic parameters, the simulation nodes implements simulated experiments on the network protocols or server program via the simulation link. The invention further discloses a simulation method based on the network simulation system. Compared with a traditional network experiment platform, a network simulator computer (or embedded equipment) is connected with multiple simulation node computers, a simulation control system of the network simulator computer is operated to realize network topology change and configuration of the characteristic parameters of the simulation link, and the network protocol can be verified under different network topologies and link characteristic parameters via fewer common computers.

Methods and systems for improving the sensitivity of a variety of conductivity sensing devices, in particular capacitively-coupled contactless conductivity detectors. A parallel inductor is added to the conductivity sensor. The sensor with the parallel inductor is operated at a resonant frequency of the equivalent circuit model. At the resonant frequency, parasitic capacitances that are either in series or in parallel with the conductance (and possibly a series resistance) is substantially removed from the equivalent circuit, leaving a purely resistive impedance. An appreciably higher sensor sensitivity results. Experimental verification shows that sensitivity improvements of the order of 10,000-fold are possible. Examples of detecting particulates with high precision by application of the apparatus and methods of operation are described.

The invention discloses a semi-supervised random forest classification method based on Spark, which utilizes a random forests algorithm to employ replacement sampling on a training data set and column attributes, so that randomness is added in both row and column directions to ensure decision-making tree diversity, and to avoid tree pruning; in addition, a category is determined in a voting method, and accuracy is greatly improved. Accordingly, the random forests algorithm does not need to perform dimensionality reduction in processing high-dimensional data samples, and has sound effects for a sparse vector as well as a dense vector. According to the verification of a plurality of sets of experiments, the semi-supervised learning algorithm reduces a classification model error rate mean value, and improves calculating performance.

The invention discloses a data dimension reduction method based on neighborhood rule. The method includes the following steps: firstly, a spherical neighborhood of present sample points is established by using the geometric spherical-modelling theory and all the sample points contained in the spherical neighborhood are adopted as candidate neighbor points, thus not only preserving the effectivity of the dimension reduction capability when data sets are sparse but also getting the advantages of low-sensitivity to isolated points and good stability of the preserved topological structure; then a data relevance matrix more matching semantics can be obtained by relevance measurement based on route clusters to update the candidate neighbor points in the spherical neighborhood and optimize the regular neighborhood space of the present sample points, thus improving the phenomenon that the dimension reduction of sample sets provided with folded curved faces is apt to suffer the integrated-structure distortion in case of heterogeneous data distribution. The experiments on different sample sets demonstrate that the method provided by the invention is available and effective.

The utility model relates to an active control test platform and a method for vibration of a near space aircraft model. The test platform leads an aircraft model distributed and stuck with a piezoelectric sensor and an actuator to be hung on an aluminum alloy outer framework, and is connected with a function signal generator of a power amplifier to drive an exciter fixed on the aluminum alloy outer framework and lead the aircraft model connected with the exciter to generate a vibration response; and a vibration response signal is collected by the piezoelectric sensor, and is outputted on the power amplifier and acted on the piezoelectric actuator through the operation of a computer inserted with a data acquisition and output card. The active control test platform has the characteristics of convenient composition, simple structure, and good expandability. The method can provide experimental verification realization means for the active control method for the vibration of the near space aircraft model, and provide technical realization supports for exploring the further practical applications of relevant control theories and methods.

The invention discloses a PSO-based electric quantity forecasting method, which belongs to the technical field of electric quantity forecasting. The forecasting method of electricity quantity under the policy of cut-off and limited production based on BP model. Firstly, the input data are analyzed and processed. Then, taking historical electricity consumption as independent variable and historicalelectricity consumption as dependent variable to train samples, using PSO algorithm to optimize the weights and thresholds of BP neural network, calculating the prediction accuracy of different parameters, and obtaining the weights and thresholds of BP model with high prediction accuracy; Finally, the BP neural network model is forecasted, the optimized parameters of particle swarm optimization algorithm and the forecasting samples are input to the forecasting model, and the forecasting value is obtained. BP neural network algorithm is optimize by PSO, Considering the influence of air qualityindex, meteorological factors and the output factors of main production stopping and limiting products on electricity consumption, the eigenvector of electricity consumption is studied and trained, and the forecasting effect is proved to be ideal by experiments. A new way of forecasting electricity consumption under the influence of production stopping and limiting policy is provided.

This invention relates to one digital code camera to imbed photos into high height water mark in digital image water mark technique, which adds the device and water mark imbed device to get high data in the camera to use digital water mark technique and imbeds the data height information by water mark type into digital images as future validation base to identification digital image height.

The invention provides a whole-exome sequencing data analysis method. The method comprises the following steps of 1) quality control of sequencing data; 2) genome mapping of the sequencing data; 3) seeking of high-confidence genome mutation by the sequencing data; and 4) annotation of mutation sites. According to the method, the analysis of large-scale data is finished through simple parameter submitting, wherein the analysis of the large-scale data comprises quality detection of original data, data denoising and genome mapping of sequencing read; an upstream part takes over original sequencing data of a lower machine; the analysis of the sequencing data is finished through a parameter automated submitting and analysis module; and candidate pathogenic mutation sites and related genes are output, thereby providing a basis for later experimental verification.

A method for screening out super-accumulated plants from flowers and plants includes applying preliminary screening experiment to select out flowers and plants with high tolerance and high accumulation property and then applying pot culturing and water planting experiment to verify that flowers and plants have character of super-accumulated plant and have repair potentiality of plant.

The invention provides a fine-grained recognition method based on reinforcement learning and cross bilinear features, aiming at solving the problem that an area with the best discrimination capabilityof a fine-grained image is difficult to mine. An actor-Critic strategy is used to mine the most attention-grabbing areas of an image. An Actor module is responsible for generating top M candidate areas with the best discrimination capability. A Critic module evaluates the state value of the action by utilizing the cross bilinear characteristic; and then calculates a reward value of the action under the current state by utilizing a sorting one-type reward, further obtains a value advantage, feeds the value advantage back to the Actor module, updates the output of the region with the most attention, and finally predicts the fine-grained category by using the region with the most discrimination capability in combination with the original image characteristics. According to the method, the region with the most attention of the fine-grained image can be better mined. It is verified by experiments that the recognition accuracy of the present invention on the CUB-200-2011 public data set isimproved compared with the existing methods, and the high fine-grain recognition accuracy rate is achieved respectively.

The present invention discloses a method for calculating the word semantic relevancy. According to word semantic information in the how-net and expression characteristics of the how-net for word semantic, semantic relations in the how-net are extracted and a semantic relation graph based on the how-net is constructed; expansion of the semantic relation graph is carried out; and finally, based on the semantic relation graph, calculation of the word semantic relevancy is carried out. The method has the beneficial effects that based on the semantic relation graph, graph theory knowledge is combined with information in the semantic relation graph, the method for calculating the word semantic relevancy on the basis of semantic relation graph is disclosed and the method is verified to be effective and feasible by experiments.

The invention discloses a crack damage monitoring and strain field measuring method based on an FBG (Fiber Bragg Grating) sensor array spectrum. The method comprises the following steps: 1, selectinga test specimen to be measured; 2, performing structural mechanics analysis on the test specimen to be measured; 3, sticking a sensor to the test specimen; 4, performing a loading experiment on the test specimen stuck with the FBG sensor to determine an experimental system; 5, performing analysis on a series of obtained spectra to explore the corresponding relation between the characteristic parameters of the spectra and the initial positions of cracks as well as the extension conditions of the cracks; and 6, comparing the stress strain and the distance curve obtained by the FBG sensor with finite element simulation results. By virtue of the steps, crack damage monitoring and crack tip stress field measurement are realized, the corresponding relation between crack damages and spectra is achieved, and the feasibility of the method for measuring the crack tip strain field can be verified through experiments.

The invention provides a protein composite identification method based on a random walking model. Interaction data and false-negative or false-positive noisy data truly existing on a protein network are forecasted through the random walking algorithm. On the protein interaction network obtained after false-negative data and false-positive noisy data are removed, protein composites with the biological significance are identified through a H-index graph model, the semantic similarity between the protein composites is calculated according to a GO body, and the identified protein composites are finally determined. According to the protein composite identification method based on the random walking model, the algorithm is insensitive to input parameters, and the effectiveness of the provided algorithm is verified through experiments.

The invention relates to the field of subway channel monitoring, in particular to a laser reflection sheet achieving method suitable for positioning and deformation analysis in a channel. The method comprises steps of material analysis; channel deformation analysis reflection sheet processing; point cloud data processing; reflection sheet coding scanning; and experiment verification, wherein the reflection sheet coding scanning comprises a binary system hash code and the binary system hash code comprises forward-direction compiling of the binary system hash code and inverse-direction recognition of the binary system hash code. The beneficial effects are that by using a dot matrix to scan the binary system hash code, through total point cloud data reflection intensity, effective reflectionpoint recognition is scanned and screened, so point cloud plane fitting and plane projection are effectively reflected; and through point cloud coordinate conversion, certain processing is performed on the point cloud data, and through a specific coding rule, transcoding is performed on codes included in the point cloud information, so information like the city where a settlement observation pointis located, the channel number and the point number is obtained, and automatic recognition and reverse-direction designing of channel deformation reflection sheets are achieved.

The behavioral model and predistorter for modeling and reducing nonlinear effects in power amplifiers addresses the model size estimation problem. The GMP model is replaced by the hybrid memory polynomial/envelope memory polynomial (HMEM) model within a twin nonlinear two-box structure to reduce the number of variables involved in the model size estimation problem, without compromising model accuracy and digital predistorter performance. A sequential approach is presented to efficiently estimate the model size. Experimental validation is carried out to evaluate the performance of the size estimation and the accuracy of the HMEM-based twin-nonlinear two-box model with respect to that of the GMP-based twin-nonlinear two-box model.

A method of generating parameters to guide a spreading process of a three dimensional printer may include the following steps: determining one or more properties of an actual powder; generating a virtual powder model which mimics the actual powder; performing one or more virtual spreading simulations; experimentally validating virtual spreading; and using advanced regression techniques to generate spreading process map from a few virtual spreading simulations.

The invention provides a group abnormal behavior detection method based on an air monitoring platform. Firstly, light flow vectors of feature points are appropriately corrected by estimating depth information of an image to reduce a target movement speed estimation error caused by the perspective phenomenon, then the light flow vectors of the feature points are clustered, and target detection under a moving camera is achieved by combining a background movement consistency law. Abnormal behaviors are detected by adopting a double-Gauss mixed model, and model parameters are solved by using an expectation maximization algorithm. Finally, misjudgment is verified by adopting a time queue mechanism, space coordinates of the abnormal feature points are clustered by means of a simplified agglomerative hierarchical clustering algorithm, the isolated abnormal feature points are removed, and abnormal groups are marked. The validity of the method is verified by experiments in multiple scenes.

The invention provides a panoramic street view privacy protection method based on aggregation channel features, and relates to the field of computer vision. The method comprises the following steps that multichannel features are extracted, a classifier is trained and enhanced, feature combinations suitable for faces and license plates are obtained through experimental verification and detection of the faces and the license plates is unified; 10 channels in total including LUV, gradient magnitude and 6 gradient directions of histograms are obtained in the multichannel features through the experiment to act as a feature set of the optimal street view face detection and license plate detection computation speed and final classification accuracy; the channel features of scales near the scale are estimated according to the existing scale channel features; a target is self-adaptively blurred according to the size of the detection target; at least 2 weak classifiers are trained and finally a robust strong classifier is formed, and an improved rapid enhancing decision tree is used as the classifier to perform pruning in the early stage of training so that computation speed can be greatly enhanced; and the functions of background batch processing, click blurring and rapid deblurring are additionally arranged for plotting and other special application backgrounds.

A study of several key conditioner design parameters has been conducted. The purpose was to improve conditioner performance by considering factors such as wafer defects, pad life, and conditioner life. For this study, several key conditioner design parameters such as diamond type, diamond size, diamond shape, diamond concentration and distribution, were selected to determine their effect on CMP performance and process stability. Experimental validations were conducted. Conditioner specifications were matched to each specific CMP environment (intended application) in order to improve process stability and CMP performance particularly for emerging technology nodes. Several conditioner designs were developed and run successfully in the field. Significant planarity improvement for a 300 mm CMP process was achieved in accordance with one embodiment, and an increase of pad life and wafer polish rate was simultaneously achieved with another embodiment.