88results about How to "Efficient and accurate classification" patented technology

Described are computer-based methods and apparatuses, including computer program products, for generating code on a network. Input files including information files and schema files are utilized to generate platform independent runtime files. The processing of the runtime files generates one or more business service applications. The runtime files map data between a standard format and an internal format.

The invention discloses a website classifying method. The website classifying method comprises the following steps: obtaining multidimensional attributes of a website and representing the multidimensional attributes by utilizing a set; carrying out self-coding characteristic learning for the set that represents the multidimensional attributes; carrying out website clustering learning by utilizing a self-coding learning result to obtain a support vector machine (SVM) used for carrying out website classifying; a step S104: while classifying any unmarked website, firstly carrying out a step S101 and a step S102 to obtain the self-coding learning result corresponding to the web site; and then, inputting the structure into the SVM obtained in the step S103, and finally carrying out website classifying to obtain the category of the website. The website classifying method disclosed by the invention can efficiently and accurately classify the website according to the industry category, and also can quickly detect a fishing webpage with malicious characteristics. A way of multidimensional attribute description is adopted, so that convenience and universality of the system are increased; and moreover, the system has extremely strong stability.

Described are computer-based methods and apparatuses, including computer program products, for generating code on a network. Input files including information files and schema files are utilized to generate platform independent runtime files. The processing of the runtime files generates one or more business service applications. The runtime files map data between a standard format and an internal format.

The invention discloses a network user classification method and device. The network user classification method comprises the following steps of: generating behavior characteristic vectors of users based on the network behaviors of the users; clustering based on the behavior characteristic vectors of the users so as to construct a user-topic matrix; determining the webpage category to which each user belongs on a basis of the user-topic matrix, wherein the network behaviors of the users comprise network access behaviors and search behaviors, and each element value of the user-topic matrix respectively represents the correlation degree between each user and each topic. According to the method and device disclosed by the invention, on a basis of various types of data of the user search behaviors and access behaviors, overall user behaviors are covered within a smaller calculation cost by the adoption of a user-topic model, so that the user information is sufficient to use, and a more accurate user classification result is faster to obtain.

The invention discloses a deep learning traffic classification method based on combination of time-space characteristics, and mainly solves the problem of low detection accuracy in the prior art. According to the implementation scheme, the method comprises the following steps: 1) collecting and marking original flow load data; 2) generating a preprocessed flow graph set based on the original flowload data; 3) training a deep learning model based on space-time characteristic combination by using the flow graph set; 4) verifying the trained deep learning model by using newly acquired and generated flow data, and deploying the model as a flow classifier at a real network node after the model is qualified; and 5) analyzing, classifying and labeling the traffic in the real network environment.According to the model constructed by the invention, the space-time characteristics of the traffic data are utilized, the traffic classification accuracy is improved, resources occupied by the classifier are reduced, the traffic classification requirement in the current network environment can be met, and the method can be applied to network edge nodes to realize encrypted traffic identificationand malicious traffic detection.

The present disclosure relates to a color classification system that accurately classifies objects in digital images based on color. In particular, in one or more embodiments, the color classification system utilizes a multidimensional color space and one or more color mappings to match objects to colors. Indeed, the color classification system can accurately and efficiently detect the color of an object utilizing one or more color similarity regions generated in the multidimensional color space.

The invention relates to a method for identifying and classifying epileptiform discharges, a ystem, a device thereof and a medium. The method comprises the following steps: acquiring a plurality of original multi-lead EEG data; respectively pre-processing each original multi-lead EEG data to obtain a plurality of lead models corresponding to each original multi-lead EEG data; employing a power spectral density method and a scale-dependent Lyapunov exponential method to extract features of each lead model respectively, and obtaining a plurality of feature index sets corresponding to each of thelead models one by one; constructing a target random forest classifier according to all the feature index sets; and perofmirng identification and classification on the EEG signals to-be-detected according to the target random forest classifier, and obtaining the test results. The method can make up for the deficiency of a traditional nonlinear signal processing method in the digitized EEG signalanalysis, realizes the classification of the normal human brain electrical signal and the epileptiform discharges, and has high recognition and classification accuracy.

The invention discloses a time series data nearest-neighbor classifying method based on subsection orthogonal polynomial decomposition. The time series data nearest-neighbor classifying method includes dividing a time sequence into subsequences comprising complete fluctuation trends on the basis of time sequence coded identification turning points; extracting Chebyshev coefficient as subsequence features by means of a first type Chebyshev polynomial decomposition subsequences, and constructing subsequence feature vectors; finally in the nearest-neighbor classifier, classifying by the dynamic planning algorithm based on local mode matching as distance metric function. Classifying accuracy and efficiency are superior to other nearest-neighbor classifiers to the great extent, and the time series data nearest-neighbor classifying method plays an important role in daily activity of people and industrial production, such as in applications of banking transactions, traffic control, air quality and temperature monitoring, industrial process monitoring, medical diagnosis and the like, massive sampling data or high-speed dynamic data can be classified and predicted, abnormalities can be detected and online modes are identified.

The invention discloses a classification method and system of emotions of news readers. The classification method comprises the following steps: acquiring a news text and a comment text as well as word characteristic information from target linguistic data; fusing the word characteristic information and converting the word characteristic information into available linguistic data with a corresponding format of a maximum entropy model; dividing the available linguistic data into training linguistic data and testing linguistic data according to a pre-set rule, and dividing the training linguistic data into marked samples and unmarked samples; training the marked samples to obtain a maximum entropy model; classifying emotion classes of the unmarked samples by using the maximum entropy model to obtain posterior probability of each emotion class corresponding to the unmarked sample; carrying out emotion class marking on the unmarked samples with the preset quantity and maximum uncertainty of the posterior probability to form new marked samples, and updating the current marked samples and unmarked samples; and circulating the last step until all the unmarked samples are marked. The classification method and system can be used for efficiently classifying the emotions of the news readers when the scale of marking the linguistic data is relatively small.

The invention discloses a deep learning intestinal polyp segmentation method based on multi-scale information and a parallel attention mechanism, and the method comprises the steps: extracting features from finer granularity in a mode of building a branch during coding, and recalibrating a feature response through an improved compression excitation module; on the basis of pooling of the cavity space pyramid, further extracting and fusing features by establishing the relation between branches, multi-scale features of the intestinal tract and the polyp can be more accurately extracted and distinguished, so the problem that wrinkles of the intestinal tract wall are often misjudged as a polyp area during segmentation is well solved; during decoding, abandoning shallow features, refining the deep features, and further establishing a boundary relationship by using an attention mechanism, so a polyp boundary can be segmented more accurately on the basis of shortening training time.

The invention discloses a new non-interactive K-nearest neighbor classification method under privacy protection, invention relates to the vector classification field of K nearest neighbor classification algorithm under privacy protection. The steps are as follows: 1. The client end encrypts the training data in the training data set composed of a plurality of training data with labels by a vectorhomomorphic encryption method to obtain a ciphertext data set and an intermediate matrix, and uploads the ciphertext data set and the intermediate matrix to the cloud; 2. That client end receives theplaintext vector group to be classified and encrypt the plaintext vector group to obtain the ciphertext vector group, and uploads the ciphertext vector group to the cloud end; 3. According to that ciphertext data set and the intermediate matrix, the cloud end calculates the similarity between each ciphertext vector in the ciphertext vector set and all ciphertext data contain in the ciphertext dataset, obtains the classification result set of the ciphertext vector set according to the nearest neighbor classification algorithm, and sends the classification result set to the client end. The invention greatly improves the efficiency and security of encryption, realizes non-interactive technology, achieves real outsourcing calculation, and reduces the calculation pressure of the client.

The embodiment of the invention discloses an electromagnetic signal classification method and a device. The method comprises the following steps: constructing a scattering network through wavelet scattering transformation; scattering features of electromagnetic signals in the electromagnetic signal set are extracted by using the constructed scattering network; and obtaining a feature sample set, training a support vector machine (SVM) classification model by using the feature sample set, extracting scattering features of electromagnetic signals to be classified by using the constructed scattering network, and inputting the extracted scattering features into the trained SVM classification model to obtain a classification result. According to the method of the invention, the scattering network and the support vector machine are organically combined, the structure of the convolutional neural network is reserved, and a filter obtained through data learning in the convolutional neural network is replaced by the pre-constructed wavelet filter, so that the calculation complexity is greatly reduced; through wavelet cascade operation, noise interference in the signal classification processcan be effectively overcome, and efficient and accurate electromagnetic signal classification is achieved.

A method for intelligently classify and managing that quality of tube yarn and its implement device are disclosed, the quality information of tubular yarn was collected on-line and its grade was determined, and then transporting the tube yarn to the corresponding winding area through different channels according to the grades, wherein the quality information is obtained by processing an image of ayarn in an instantaneous stable state after the image is collected, Grades are created by matching the grades, quality information of historical tubes of known grades and quality information of current tubes of which grades are to be determined as categories, The training sample and the test sample are determined using the 'one-to-one' classification method of the SVM, and the device comprises acollecting device for collecting quality information of the tubular yarn, an RFID system for confirming grade information of the tubular yarn and storing it in an electronic tag on the tubular yarn, and a track classifier for adjusting the traveling direction of the tubular yarn according to the grade. The method of the invention has the advantages of high accuracy and efficiency, simple device structure, high automation degree and good application prospect.

The invention belongs to the technical field of computer vision, and provides a fine-grained image classification algorithm based on discriminant learning. A new end-to-end autoregressive positioningand discriminative priori network model is provided. The new end-to-end autoregressive positioning and discriminative priori network model discriminates the size of the patch more accurately through learning and exploring, and can classify images in real time. Specifically, a multi-task discriminant learning network is designed, and the multi-task discriminant learning network comprises an autoregressive positioning sub-network and a discriminant prior sub-network, and the discriminant prior sub-network has a guide loss function and a consistency loss function to learn an autoregressive coefficient and a discriminant prior map at the same time. The autoregressive coefficient can reduce noise information in the discriminative patch, and thousands of candidate patches are filtered into the number of bits by learning the discriminative probability value by the discriminative priori map. A large number of experiments show that the proposed SDN model reaches the latest level in accuracy andefficiency.

The invention relates to an intelligent logistics classified processing device based on Internet +, and effectively solves the problems that an existing device cannot carry out omnibearing code scanning input, cannot classify the volume and the weight of logistics express items at the same time, and cannot make corresponding adjustment according to different classification standards. Visual code scanning equipment and a volume detection bin are arranged on an assembly line, width detection plates are arranged on the two sides of the volume detection bin, width contact sensors are arranged on one sides of the width detection plates, a height detection plate is arranged on the volume detection bin, a height contact sensor is arranged on one side of each height detection plate, a sorting binis arranged on one side of the volume detection bin, a sorting disc is rotationally connected into the sorting bin, and a small item conveying bin and a large item conveying bin which are fixedly connected to the sorting bin are arranged on the side face of the sorting bin, According to the device, omnibearing code scanning input of express classification is realized, the practicability and applicability are very high, the requirements of logistics classification for timeliness and high efficiency are met, and logistics classification becomes more efficient and faster.

The invention provides an intelligent recyclable garbage delivery integrated box, wherein the top of the integrated box is provided with a box cover, the top of the box cover is provided with a solarpanel; a voice guiding system, a multimedia integrated machine, a two-dimensional code scanner, a stripping blade and a throwing port are arranged at the upper part of the integrated box, and the throwing port is connected with a conveying device; the lower part of the multimedia integrated machine is sequentially provided with a waste paper throwing port, a ribbon machine, a sundries throwing port and a cap throwing port; the lower part of the integrated box is provided with a plastic bottle box body, a pop can box body, a bottle cover box body, a sundries box body and a waste paper box body;a fixing plate is installed at the top of the plastic bottle box body, and a shredding device is arranged on the fixing plate; the advantages are as follows: the box provided by the invention can guide the citizens to classify the garbage based on the methods of cloud computing, internet of things, artificial intelligence and the like, realize the return of the garbage by recycling the garbage, and develop the garbage classification consciousness; the integrated box is selectively installed according to the scene to accurately and efficiently realize garbage sorting; decision analysis is provided to the government through big data so as to transform to a resource-saving and environment-friendly society.

Embodiments of the invention disclose a hyperspectral indoor monitoring method based on deep learning, which comprises: acquiring images of bacteria in a scene to be detected through a hyperspectral camera; recognizing the images of the bacteria through an artificial neural network model to obtain bacteria recognition results; determining radiation intensity of ultraviolet of an LED ultraviolet sterilizing device according to the bacteria recognition results; controlling the LED ultraviolet sterilizing device to kill bacteria with the radiation intensity. The hyperspectral indoor monitoring method based on deep learning is suitable for killing harmful bacteria effectively and can also ensure human health safety.

The invention discloses an image classification model training method and device, electronic equipment and a storage medium. An image classification model comprises a plurality of image classificationmodules, each image classification module corresponds to different image classification scenes, and a training data set is acquired; feature extraction is performed on each image sample classification label to obtain a feature image corresponding to each image sample; taking the feature image corresponding to each image sample as input data, taking the sample classification label corresponding toeach image sample as output data, and respectively training the plurality of image classification modules to obtain a plurality of trained image classification modules; and obtaining a trained imageclassification model based on the plurality of trained image classification modules. Since each trained image classification module classifies each image classification scene, each image classification module in the trained image classification model can accurately and efficiently classify the image corresponding to each image classification scene.

The invention relates to a parallel arm intelligent silkworm cocoon sorting robot and a silkworm cocoon sorting method. The parallel arm intelligent silkworm cocoon sorting robot comprises a main control system, an image collecting device, a silkworm cocoon conveying system and a parallel arm sorting manipulator, and the collecting end of the image collecting device faces silkworm cocoons conveyed by the silkworm cocoon conveying system; and the execution tail end of the parallel arm sorting manipulator is located above the silkworm cocoon conveying system, and the image collecting device and the parallel arm sorting manipulator are both electrically connected with the control system. According to the parallel arm intelligent silkworm cocoon sorting robot, the parallel arm sorting manipulator is adopted, a negative pressure type suction cup is adopted, accurate silkworm cocoon grabbing is achieved, and secondary damage to silkworm cocoons is avoided; silkworm cocoon images are acquired through an upper camera and a lower camera, so that the recognition accuracy is improved; through a constructed silkworm cocoon classification CNN model based on deep learning, accurate and efficient identification and classification of the silkworm cocoons are realized; and the control system converts the position coordinates of the silkworm cocoons and drives the parallel arm manipulator to act to grab the silkworm cocoons, so that efficient and automatic sorting of the silkworm cocoons is realized.

The invention discloses a method, a system and equipment for intelligently recycling garbage. The method comprises the following steps: acquiring RGB image and depth information of an object on a conveyor belt, and calculating initial grabbing pose of each object; then, calculating real-time grabbing pose of each object according to the initial grabbing pose of each object and the movement speed of the conveyor belt; and according to the real-time grabbing pose of each object, grabbing and placing each object by a mechanical arm. By the adoption of the method, system and equipment for intelligently recycling garbage, different types of recyclable garbage can be intelligently recognized and thrown into the garbage cans of the corresponding types, so that the mixed recyclable garbage can beefficiently and accurately classified and extracted, and the recyclable garbage classification efficiency and accuracy are greatly improved.

The invention belongs to the technical field of garbage recycling devices and particularly discloses an intelligent-classifying dry and wet garbage barrel. The intelligent-classifying dry and wet garbage barrel comprises a power supply, an outer barrel body, a dry material inner barrel body and a wet material inner barrel body. Barrel covers and cover turning mechanisms are arranged at the tops ofthe dry material inner barrel body and the wet material inner barrel body correspondingly. Driving motors and driving wheels are arranged on one sides of the barrel covers, a pre-containing groove isformed between the two driving wheels, and a moisture sensor is arranged at the bottom of the pre-containing groove. Two full-bin detecting mechanisms are arranged in a cavity. Conductive openings are formed in the bottoms of annular cavities, and conductive plates are arranged on the two sides of each conductive opening correspondingly, wherein one of the conductive plates is connected with thepower supply, and the other conductive plate is electrically connected with the corresponding cover turning mechanism. Traction wheels are arranged in the barrel covers, winding wheels are arranged onthe lower portions of the dry material inner barrel body and the wet material inner barrel body correspondingly, traction ropes are connected with the traction wheels, and the winding wheels are connected with threaded rods. Inner threaded holes are formed in the bottoms of the dry material inner barrel body and the wet material inner barrel body. By adopting the intelligent-classifying dry and wet garbage, the problems of dry and wet garbage classifying and garbage spilling can be effectively solved.

The invention discloses a Hadoop platform-based improved parallel KNN online public opinion classification algorithm. According to the algorithm, Hadoop distributed storage characters and a MapReduceprogram for designing parallel kNN are utilized to solve problems during the processing of bulk data, and test verification on classification ability and classification efficiency of a parallel kNN algorithm is carried out. Experiment results prove that the Hadoop platform-based improved parallel KNN online public opinion classification algorithm is capable of rapidly, efficiently and correctly classifying online public opinion data when being used for processing bulk online public opinion data.

The invention discloses a semi-supervised deep network picture classification method based on swarm intelligence. The semi-supervised deep network picture classification method comprises the followingsteps of: 1), preprocessing a training data set and a testing data set of a digital picture, and carrying out normalization and centralization processing; 2) calculating network loss by using the training data set, specifically, 2-1) for marked data, calculating hybrid KL divergence loss between a predicted value and a real label value of the network, and 2-2) for unmarked data, calculating groupconsistency loss among a plurality of network model prediction values; 3) optimizing weight parameters of a deep network through a back propagation algorithm by using the hybrid KL divergence loss ofthe marked data and the group consistency loss of the unmarked data, and 4) classifying the test data set by using the trained deep network. The semi-supervised deep network picture classification method can realize efficient and accurate classification of the pictures.

The invention discloses an extensible network attack behavior classification method. The method comprises the following steps: carrying out data preprocessing on network flow data; extracting a new feature expression and an optimal original feature set from the multi-dimensional feature attributes of the network traffic data; obtaining model related parameters used for preliminary judgment of network behavior attack categories through classification model training; and obtaining the weight values of the known attack category and the normal behavior of the network behavior and the weight valueof the new attack category to comprehensively judge the attack category of the network behavior. According to the invention, the classification result of the network attack behaviors is optimized; anda supervised learning model and an unsupervised learning model are optimized respectively by extracting new feature expressions from multi-dimensional feature attributes of the network traffic data and selecting an optimal original feature set capable of maximally expressing data features, so that a new attack category can be effectively identified on the basis of ensuring the judgment accuracy of the known attack category.

The invention discloses an image segmentation method, an apparatus and a computer-readable storage medium. The original image is inputted to a pre-trained neural network for calculation, and each object in the first image is classified to obtain a pixel point set corresponding to each object. The invention has the effects of accurately dividing each pixel of an object to achieve accurate positioning and accurate attitude judgment.

The invention relates to the technical field of voice processing, and provides a language recognition model training method, a language recognition method and related equipment. The language recognition model training method comprises the steps: acquiring sample data including an initial voice and a target language of the initial voice; preprocessing the initial voice to obtain a spectrogram; training a language recognition model, including: extracting spatial features of a spectrogram through a convolutional neural network; extracting time sequence features of the spatial features through a recurrent neural network; performing full-connection operation on the spatial features based on the time sequence features, and predicting a language probability through a classifier; and adjusting parameters of the language identification model according to the language probability and the target language until the language identification model converges. According to the method, the language of the voice can be efficiently and accurately classified, and data support is provided for subsequent voice recognition.

The invention discloses an SAR image target classification method based on a multi-kernel scale convolutional neural network. The method comprises the following steps: 1, selecting different types of SAR images as a sample set; 2, carrying out convolution on the input SAR image in parallel by adopting a multi-scale convolution kernel in each convolution layer, and carrying out multi-scale optimization fusion on extracted multi-kernel scale features to obtain fusion features; 3, carrying out the multi-level optimization fusion of the fusion features extracted from the shallow, middle and deep convolution layers, and obtaining the final features; 4, inputting the final features into a full connection layer and a softmax classifier to obtain a prediction result, and comparing the prediction result with a real result to complete a network training process; and 5, inputting the SAR image to be classified into the trained multi-kernel scale convolutional neural network to obtain a corresponding category. According to the method, the target feature representation integrity of the SAR image can be improved, higher classification precision and classification efficiency are obtained, and the method has better engineering application value.

The invention discloses a malicious code classification method based on a deep residual network and a mixed attention mechanism. According to the method, the malicious code is visualized into an RGBA image, and the information entropy of the code is not additionally calculated, so that the defects that the malicious code image information source is single, the image feature is not obvious and the calculation amount is too large are overcome; meanwhile, a deep residual network and an attention mechanism are combined to construct a malicious code classification model, the deep residual network improves the classification precision, and meanwhile, the gradient disappearance problem is relieved through short circuit connection, so that model convergence is accelerated, and the discrimination ability of the model is improved; and a mixed attention mechanism is adopted behind each residual unit, so that more critical deep features are extracted from two dimensions of channels and spaces, and the malicious code classification accuracy is further improved.