65results about How to "Accurate classification effect" patented technology

The invention provides a magnetic resonance image feature extraction and classification method based on deep learning, comprising: S1, taking a magnetic resonance image, and performing pretreatment operation and feature mapping operation on the magnetic resonance image; S2, constructing a multilayer convolutional neural network including an input layer, a plurality of convolutional layers, at least one pooling layer/lower sampling layer and a fully connected layer, wherein the convolutional layers and the pooling layer/lower sampling layer are successively alternatively arranged between the input layer and the fully connected layer, and the convolutional layers are one more than the pooling layer/lower sampling layer; S3, employing the multilayer convolutional neural network constructed in Step 2 to extract features of the magnetic resonance image; and S4, inputting feature vectors outputted in Step 3 into a Softmax classifier, and determining the disease attribute of the magnetic resonance image. The magnetic resonance image feature extraction and classification method can automatically obtain highly distinguishable features/feature combinations based on the nonlinear mapping of the multilayer convolutional neural network, and continuously optimize a network structure to obtain better classification effects.

The invention relates to the field of medical image processing and pattern recognition, and provides a computer-assisted lump detecting method based on a mammary gland magnetic resonance image. The computer-assisted lump detecting method based on the mammary gland magnetic resonance image aims at solving the problems that in the prior art, the lump partition effect is not good, and the accuracy, the sensitivity and the specificity in a classification test are not high. The computer-assisted lump detecting method includes the following steps: S1, extracting an interest area from the mammary gland magnetic resonance image; S2, extracting an initial lump area from the interest area in a separated mode, and determining the contour line of the initial lump area; S3, calculating the weight distribution of characteristic parameters of the initial lump area; S4, selecting the characteristic parameters, with the weight coefficients larger than a standard weight coefficient, of the initial lump area, and carrying out training classifying to obtain optimized characteristic parameters; S5, inputting the optimized characteristic parameters into a classifier, analyzing the optimized characteristic parameters with a support vector machine classification method, determining a final lump area, and displaying the final lump area for a user. The detecting method has the good lump partition effect, the accuracy, the sensitivity and the specificity in the classification test are effectively improved, the detecting result serves as a second opinion to be provided for a radiologist, and the misdiagnosis rate and the missed diagnosis rate of the radiologist can be effectively reduced.

The invention relates to a fault arc recognition method based on extraction current features. The fault arc recognition method based on extraction current features includes the steps: training a BP neural network by using the wavelet energy corresponding to the loop current in the normal state and the fault arc state to obtain a BP classification network based on current features, and performing fault arc recognition by using the BP classification network. The fault arc recognition method based on extraction current features effectively integrates the wavelet decomposition technology with theBP neural network intelligent algorithm, fully utilizes the advantages of wavelet decomposition to reflect the time-frequency domain change of the signal, and combines with the intelligent and accurate classification effect of the BP neural network to realize fast and accurate arc detection and improve the current fault arc detection level, so as to preferably guarantee the safe and reliable operation of power equipment, bring better benefits to the development of the power industry, and has very important practical significance for the safe, reliable and stable operation of the power grid.

The invention discloses a hyperspectral image classification method based on edge preservation and a graph cut model. The hyperspectral image classification method comprises the following steps that S1, hyperspectral images to be classified are inputted; S2, the image elements of the corresponding coordinate positions of the original hyperspectral images are extracted to form a reference data sample set; S3, a supervised classification training sample set is selected; and the rest reference data samples act as a test sample set; S4, pixel level image classification operation is performed so that a probability membership distribution graph of each corresponding class is acquired; S5, filtering is performed so that the optimized class probability membership distribution graph is acquired; S6, all the ground targets are extracted: the optimized class probability membership distribution graph is cut by using the graph cut model so that the cut result of each class is acquired; and the final tag result is acquired from the cut result of each class by using the merging rule and the maximum posterior probability estimation; and S7, the final classification graph is outputted. A new strategy for area tagging is provided so that the hyperspectral image classification accuracy can be effectively enhanced.

The invention discloses a sorting method of a ground-based visible light cloud picture. The method comprises the following steps that 1, image preprocessing is performed on the ground-based visible light cloud picture to obtain standard cloud pictures, a plurality of images are selected randomly from the standard cloud pictures to be used as training samples, the rest are used as testing samples, and the number of the training samples is larger than that of the testing samples; 2, global features of the standard cloud pictures are extracted, and comprise textural features and color features, and the texture features comprise gray level co-occurrence matrixes and Tamura features; 3, a bag of words model is built on basis of SIFT (Scale-Invariant Feature Transform) feature descriptors, and local features of the standard cloud pictures are extracted; 4, the global features obtained in the step 2 and the local features obtained in the step 3 are linearly fused, and a limitation learning machine model is built for the training samples to obtain a cloud picture classifier; 5, sorting is performed on the testing samples by using the cloud picture classifier, and a final sorting result is obtained. The sorting is more accurate by using the sorting method of the ground-based visible light cloud picture.

Provided is a method for effectively segmenting a hyperspectral oil-spill image. The method comprises steps of: defining an initial level set function and other related functions; acquiring a new fitting item in combination with a Fisher criterion; constructing an edge stop function to obtain a new length item; performing improvement in combination with an end member extraction algorithm; introducing a level set regular item to prevent reinitialization of the level set function; minimizing an energy function to obtain an Euler-Lagrange equation; setting parameters; selecting a display band and an initial contour; displaying a segmentation result graph; calculating various segmentation precision evaluation indexes; comparing and evaluating the accuracy of the segmentation results. The method can classify a target area in a simulated hyperspectral image and a real hyperspectral image, and effectively segments the hyperspectral oil-spill image with boundary blur and noise, improves the segmentation accuracy of the hyperspectral image, obtain a more accurate classification effect, makes the parameter change more stable, makes the contour curve more accurate, obtains the continuous and closed boundary contour, and has higher precision of segmentation.

The present invention provides a mutual information based parallel feature selection method for document classification, which comprises: a, selecting samples and performing classification; b, solving TF-ID values of words; c, generating an initialized data set D = { x1, x2, ..., xN }; d, carrying out distributed calculating and evenly distributing all sub data sets to m calculation nodes; e, establishing sets, wherein S = phi and V = { X1, X2,..., XM }; f, calculating joint probability distribution and conditional probability distribution; g, calculating mutual information; h, selecting a feature variable; i, determining if the number is enough; and i, performing document classification. According to the parallel feature selection method for document classification, which is provided by the present invention, Rayleigh entropy based mutual information is used for measuring correlation between the feature variable and a class variable, so that the finally selected feature variable can further represent a document classification feature, a classification effect is more accurate, and a classification result is better than a result obtained by using a common feature selection method. The selection method has advantageous effects, and is suitable for promotion and application.

The invention discloses a user behavior fine classification method and system of a mobile application private encryption protocol. The method comprises the following steps: 1) collecting the flow of amobile application, and then identifying the private encryption protocol flow from the collected flow according to set private encryption protocol characteristics; 2) collecting flow data of a set user behavior category from the identified private encryption protocol flow, and labeling the flow data; 3) generating a training set, a verification set and a test set according to the flow data collected and marked in the step 2); 4) performing feature extraction on the flow data in the three sets, and converting the flow data into feature vectors; 5) setting hyper-parameters of the selected classifier, and training the selected classifier; 6) verifying the training classifier by utilizing the feature vectors corresponding to the verification set, and 7) classifying the classifier on the testset, and if the set standard is met, classifying the mobile application traffic to be processed by utilizing the classifier.

The embodiment of the invention provides a webpage text classification method and device and a webpage text identification method and device. The webpage text classification method comprises the steps of collecting text data in a webpage; carrying out word segmentation on the text data and obtaining basic segmentation words; calculating first attribute values and second attribute values of the basic segmentation words; calculating feature values of the basic segmentation words according to the first attribute values and second attribute values; screening feature segmentation words from the basic segmentation words according to the feature values; calculating weights corresponding to the feature segmentation words; and taking the weights as the feature vectors of the corresponding feature segmentation words and training a classification model through adoption of the feature vectors. According to the embodiment of the invention, the feature extraction objectivity and accuracy are effectively ensured; the influences of features on the classification are taken into consideration; the webpage text classification accuracy is improved; and a user can obtain valid information timely and accurately from massive texts.

The invention relates to an intelligent obstacle avoidance system and method based on machine vision, in particular to an intelligent obstacle avoidance suitcase (trolley) based on machine vision, belongs to the technical field of intelligent robots, and aims at solving the problems that in the prior art, an intelligent following suitcase cannot avoid obstacles emergently, the automatic walking process is inconvenient and not intelligent enough. The method comprises a machine hardware drive part and a software data processing part. The machine hardware drive comprises a single-chip microcomputer, a drive board and a camera. Software data processing comprises an image acquisition module, a camera calibration module, an image processing module and a machine vision realization module, an image processing automatic identification target realizes an automatic following function, automatic obstacle avoidance and obstacle identification are realized, so that the intelligent suitcase (trolley)is more convenient, and the travel burden of people is greatly reduced.

The embodiment of the invention discloses a driving track classification method and device, electronic equipment and a storage medium. The method comprises the following steps: receiving a track log of a user driving track point; calculating track characteristics corresponding to each driving track point, track characteristics corresponding to every two adjacent driving track points and track characteristics corresponding to all the driving track points according to the track logs; according to the track characteristics corresponding to each driving track point, the track characteristics corresponding to every two adjacent driving track points and the track characteristics corresponding to all the driving track points, and the track characteristics corresponding to the sample track points,the track characteristics corresponding to every two adjacent sample track points and the track characteristics corresponding to all the sample track points which are determined in advance, determining the track type of the driving track. Therefore, the user tracks can be accurately classified, and the labor cost and the time cost can be saved.

The invention discloses a foundation cloud picture cloud classification method, and belongs to the technical field of image information processing and weather. The method comprises the following steps that (1) the textural feature, shape feature and color feature of a cloud picture are extracted to form a 21-dimensional feature vector; (2) normalization processing is carried out on each bit of the 21-dimensional feature vector; (3) a self-adaptive extreme learning machine model is built, and network training is carried out through a training sample; (4) the normalized 21-dimensional feature vector is adopted as the input of the self-adaptive extreme learning machine, and the varieties of clouds are adopted as output for cloud classification. The textural feature, shape feature and color feature of the cloud picture are comprehensively utilized, the self-adaptive extreme learning machine model based on k neighbors and the extreme learning machine is built, the foundation clouds are accurately classified, classification performance of the method is more accurate than that of an existing method, and the important application value is achieved.

The invention discloses an audio-visual combination based intelligent garbage classification-recovery method and terminal. Image data of complete images of garbage of different materials is collected,and a preprocessing convolutional neural network as well as a first sub-classification model is established and trained; sound signals of impact sounds of garbage of different materials are collected, and a second sub-classification model is established and drained; garbage data is input to the preprocessing convolutional neural network to determine whether the garbage belongs to single-type garbage or mixed-type garbage; and he image and sound data is input to the first and second sub-classification models respectively for processing to obtain a garbage type result of the two models, and a final result is obtained by weighting. The intelligent garbage classification terminal comprises a garbage can body, an automatic sensor, a classification mechanism, a garbage storage box and a controldisplay. According to the invention, digital management and control of the whole garbage treatment production process can be formed, and the garbage treatment efficiency is improved.

The invention discloses a pulmonary nodule detection method based on a convolutional neural network, and belongs to the field of deep learning image detection. After the pulmonary parenchyma is extracted, the attention of the neural network to the target area can be effectively improved, and the number of parameters can be effectively reduced. After the step of preliminary candidate nodule detection is completed, the target area close to the center of the pulmonary nodule is selected for nodule detection, the overall detection range is narrowed, a relatively accurate classification effect is obtained, and the classification performance is improved.

The invention relates to a hand-written manchu alphabet identification system, belongs to the mode identification technology field and solves a problem that manchu alphabets only having one extracted single characteristic because of letter particularities have no noise interference prevention capability. The system is characterized by comprising a processing assembly (1202), wherein the processing assembly (1202) comprises one or more processors (1220) for instruction performing to accomplish steps of a manchu identification method. The method comprises steps that a K-neighbour method is employed to carry out classification processing on manchu alphabet characteristic values after dimension reduction, and manchu alphabets corresponding to to-be-identified manchu alphabet images are acquired. The system is advantaged in that on one hand, dimension reduction and classification calculation efficiency improvement are realized, on the other hand, the manchu alphabet characteristic values after dimension reduction are easier to distinguish, relatively good noise interference prevention capability is realized, and thereby accuracy of the manchu alphabets identified through employing the K-neighbour method according to the manchu alphabet characteristic values after dimension reduction is relatively high.

The invention discloses an automobile wheel hub classification method based on a word bag model and a support vector machine. The automobile wheel hub classification method includes the steps of 1) pretreating a given automobile wheel hub images to make the given automobile wheel hub images become a standard database for training and test; 2) extracting global texture characteristics and global color characteristics of images, 3) extracting partial characteristics of the images employing SIFT descriptors, conducting K-Means clustering for the partial characteristics to form characteristic description based on the word bag model, 4) fusing global and partial characteristics, using an extreme learning machine to learn and obtain an automobile wheel hub classifier; and 5) testing samples to be tested based on the obtained automobile wheel hub classifier to obtain final classification result. Global and partial characteristics are fused, and an extreme learning machine is used to learn an automobile wheel hub classifier. More accurate classification performance can be obtained than a conventional automobile wheel hub classification method based on KNN, SVM and BP. The accuracy in most tests reaches over 99.9%.

The invention discloses a weighted classification method for disaster information importance of blog articles based on deep learning and XGBoost algorithm. The method mainly comprises the following five steps of (1) receiving and preprocessing social media text data; (2) constructing a disaster word pre-training word vector table for converting social media text data into word vectors; (3) converting the social media text data preprocessed in the step (1) into a word vector with the dimension of d through the updated word vector table in the step (2), and performing information category classification and importance category classification on the social media text data by utilizing an XGBoost algorithm; (4) weighting the information classification result and the alarm classification resultof the social media text data to determine the importance of the contained text information. Based on social media text data, importance weighting functions are added, through dividing information categories and alarm categories of social media text data, the association degree between the information category and the alarm is established, classification is carried out by combining deep learningand an XGBoost algorithm, and the association between the information category and the alarm category is weighted to obtain a final social media text data information importance index. Experimental results show that the information importance classification effect of the social media text data is obviously improved.

The invention discloses a garbage classification method. The method comprises the steps that garbage is poured on a placement disc from a feeding port; the garbage on the placement disc is identifiedthrough an identification device; under control of a controller, the identified garbage moves by adjusting a sliding assembly, and during movement, a push plate is driven to push the garbage to a corresponding conveying groove; the garbage enters a corresponding collection box for collection through the conveying groove. According to the garbage classification method, the efficiency of classifyingthe garbage can be improved, and the device is simple in structure and convenient to apply and popularize.

The invention relates to the technical field of vehicle navigation, and discloses a road condition detection method and device, electronic equipment and a readable storage medium, and the road condition detection method comprises the steps: receiving a road condition video from video collection equipment; extracting a plurality of video frame images from the road condition video, and obtaining a plurality of optical flow density images based on the plurality of video frame images, wherein the optical flow density image is used for representing dynamic information formed by movement of each pixel point in a video frame image relative to an object; and respectively classifying the plurality of video frame images and the plurality of optical flow density images, and determining a road condition result based on the classification of the video frame images and the optical flow density images. The road condition detection method provided by the invention not only depends on the traffic flowspeed, but also can obtain a more accurate road condition result.

The invention belongs to the technical field of underwater target recognition and signal processing, and particularly relates to an underwater target recognition method based on wavelet line spectrum feature extraction, and the method comprises the steps of carrying out the spectrum analysis of a signal received by a sonar array, and obtaining the spectrum information in each frequency band; based on the obtained frequency spectrum information in each frequency band, extracting a maximum value of a line spectrum feature in a single frequency band as a feature vector of the current frequency band, and obtaining a feature vector of each frequency band; splicing or averaging the feature vectors of all the frequency bands to obtain optimized feature vectors; and taking the feature vector of the current frequency band and the optimized feature vector as a new feature vector, inputting the new feature vector into a pre-trained time delay neural network, and outputting target category information corresponding to the current frequency band as a classification result.

The invention discloses a Web API association pattern mining method based on a graph neural network, which comprises the following steps: S1, collecting a Web API data set, converting the Web API data set into a heterogeneous graph, and inputting the heterogeneous graph into the graph neural network; s2, based on the heterogeneous graph input into the graph neural network, performing multi-channel convolution operation through the graph neural network to obtain a plurality of first graph structures; s3, splicing and stacking the plurality of first graph structures to obtain a second graph structure; and S4, inputting the second graph structure into the graph convolutional neural network to obtain a classification result, and recommending to the user according to the classification result to complete mining of the Web API association mode.

The invention discloses an information source adaptive dynamic selection-based efficient fusion identification method. The method comprises the steps of collecting data for a training sample set, performing preprocessing and feature extraction, and dividing attributes of training samples into N attribute sets; obtaining the attribute set for enabling the classification accuracy of the training sample set to be highest; and classifying targets and judging whether the reliability of classification results meets a threshold requirement or not: when the requirement is met, performing target identification on the targets; and when the requirement is not met, obtaining new classification results, performing optimization fusion on all the classification results to obtain a fused classification result, until the reliability of the classification results meets the threshold requirement or all the attribute sets are used up, and when all the attribute sets are used up, performing target identification on the targets by using the finally obtained classification result. The problems of excessively high cost, poor robustness and non local optimum of an optimization process in an information source obtaining process are solved.

The invention discloses a Hash sample balance cancer labeling method for histopathologic images, which comprises the following steps: extracting features of various types of images by using a convolutional auto-encoder, extracting image block features by using an unsupervised convolutional auto-encoder CAE, and using a supervised convolutional neural network (CNN) for a final classification task; mapping an image in a high-dimensional feature space to a low-dimensional binary space by using a Hash method, forming Hash codes for all multi-class image samples, and wherein each Hash code corresponds to a subspace in an original feature space, also called a Hash bucket; and calculating the selection proportion of the extracted samples in each hash bucket, and selecting representative samples. The invention provides an efficient resampling method HBU for the class imbalance problem in a large-scale pathological histological image classification task. The pairwise distance between the samples does not need to be calculated, and the invention has high efficiency and high expansibility.

The invention belongs to the technical field of computer networks, and discloses a DGA domain name detection method and system, a medium, equipment and a terminal. The twinborn architecture model Siam-BLA is used for fitting a domain name feature space and comprises a preprocessing layer, an embedding layer, a feature extraction layer and a similarity calculation layer, the network BLA is obtained by splitting the structure of the twinborn architecture model Siam-BLA and is used for extracting domain name features, and the module Weighted-vamp is used for carrying out exclusive similarity measurement on two domain names; d; calculating a reference vector representing each category of feature condition; and an efficient twin architecture multi-classification and unknown class identification algorithm is provided. According to the technical scheme, feature engineering and large-scale data labeling are not needed, the recognition accuracy rate reaches 98% or above, the accuracy rate of part of categories is even 100%, the classification accuracy rate in a small sample environment is improved, and the time of a twin architecture for multi-classification prediction is shortened.

The embodiment of the invention belongs to the field of artificial intelligence, and particularly relates to a text classification method, device and equipment and a storage medium, and the method comprises the following steps: obtaining a to-be-analyzed text, carrying out word segmentation on the text to form a segmented word set, and obtaining subject terms of the text according to a subject model to form a subject term set; obtaining a word embedding vector of each word in the subject word set, performing dimension reduction on the word embedding vectors, mapping the word embedding vectors to a plane, and constructing a Voronoi diagram according to mapping points on the plane; calculating a semantic distance between the non-subject terms and the subject terms, and adding the non-subject terms into a Voronoi diagram; identifying a word node type of each word in the Voronoi diagram, and calculating a semantic distance between word nodes through a corresponding algorithm according to the word node type; and inputting the semantic distance between the word nodes into a pre-constructed graph convolutional neural network to output a graph implicit vector, and carrying out text classification according to the graph implicit vector. According to the invention, the accuracy of text classification is improved.