57results about How to "Guaranteed classification effect" patented technology

The invention provides a method for transient stability assessment and real-time control of an electric system. According to the method, technologies concerning knowledge discovery are introduced to realize the stability assessment and the control of the electric system; and on the basis of establishing a transient stability assessment model, the corresponding relation between the system running condition and the overall stability of the electric system is found out, the inherent running discipline of the electric system are grasped, and the fast assessment of the stable running state of the electric system is realized, therefore, the actual running management of the electric system can be guided. The method has the advantages that timely and effective stability control scheme can be established for the running state which loses stability after great disturbance, and the safe and the stable running of the system can be ensured at lowest price.

The invention provides a small-sample face recognition method which relates to a method for face recognition by means of electronic equipment. According to the face recognition method, a single-layer multi-scale convolutional neural network structure is used for performing multi-characteristic fusion and classification. The method comprises the steps of preprocessing a face image; extracting a multi-characteristic graph of the face image; extracting a DWT low-frequency subband graph, an Sobel edge characteristic graph and an LBP pattern characteristic graph; performing multi-characteristic fusion by means of the single-layer multi-scale convolutional neural network structure; and predicating a classification result by means of a Softmax classifier, thereby realizing face recognition. The small-sample face recognition method overcomes a defect of low classification recognition rate caused by a classification method in settling a small-sample face recognition problem.

The invention discloses a feature visualization method and system for a convolutional neural network model based on sparse attention, and the method comprises the steps: carrying out the feature extraction of an input color image, and outputting a multichannel feature map; performing weighted adjustment on the feature map by using pixel-level attention; adopting a cross entropy loss function as aclassification loss function, carrying out L1 regularization constraint on pixel-level attention, improving the classification loss function, and training the feature map after weighted adjustment toobtain a classification result; superposing the adjusted feature map with an originally input color image to obtain visual display of important features of the color image so as to give visual explanation of a classification result. The system comprises a feature extraction module, an attention module, a classification module and a feature visualization module. According to the invention, while the image classification accuracy is improved, the most important feature region of the image is visually displayed by using features.

The invention discloses a motor imagery identification method and system fusing a CNN-BiLSTM model and probability cooperation, belonging to the field of man-machine interfaces. The invention aims tosolve the technical problem of how to effectively extract features and how to quickly realize classification, and the method adopts the technical scheme that (1) the method captures and extracts time-space depth features in an EEG signal by merging a CNN model of a BiLSTM network, and inputs captured and extracted time-space depth features into a ProCRC classifier for classification, and uses testset data pairs to construct the CNN-BiLSTM model for performance evaluation to achieve intent recognition. The method comprises the following specific steps: S1, acquiring an electroencephalogram signal; S2, constructing a deep neural network; S3, establishing a classifier; and S4, testing and evaluating the model. The system comprises an electroencephalogram signal acquisition unit, a deep neural network construction unit, a classifier construction unit and a model test and evaluation unit.

The invention discloses an image computer-aided diagnosis method for multi-sequence nuclear magnetic resonance images, belongs to the field of computer-aided diagnosis, and relates to a computer-aided diagnosis method for multi-sequence image processing, texture feature extraction, classification and decision fusion of magnetic resonance imaging (MRI)-based T1WI, T2WI, an arterial phase, a portal vein phase and an equilibrium phase. According to the method, five sequences of the T1WI, the T2WI, the arterial phase, the portal vein phase and the equilibrium phase of MRI are integrated under a digital image processing and mode identification framework, and the image computer-aided diagnosis is completed by means of a neural network, a voting mechanism and a decision-making tree according to three levels of region of interest (ROI) processing, multi-sequence MRI classification and individual classification. By the method, multi-parameter, multi-sequence and multidirectional imaging is provided, and a combined classifier can select a sequence having the optimal distinguishing performance from the five sequences according to different stages of an anomaly structure to serve as the classification attribute of the corresponding stage. The image computer-aided diagnosis method has the advantages of rich information, clear levels and high classification accuracy.

The embodiment of the invention provides an image classification method and system based on incremental learning, and a computer medium, and the method comprises the steps: selecting old category dataof an old classification model and new incremental data, and constructing an incremental learning data set; constructing an incremental learning new classification model; inputting an incremental learning data set to the incremental learning new classification model, and performing incremental learning training under the constraint of an incremental learning loss function to obtain a trained incremental learning new classification model; and inputting a to-be-classified image to the trained incremental learning new classification model, and carrying out image classification to obtain an imageclassification result. According to the method, learning is carried out on the basis of the old category data together with the new category data, the similarity of the mapping vectors of the old category can be kept consistent in the incremental learning process under the constraint of the incremental learning loss function, and then the classification performance of the test data of the old category is kept while the information of the new category is learned.

The embodiment of the invention provides a classification method and device for multiple rounds of conversations. The classification method comprises the steps: obtaining multiple rounds of conversations between a target user and a robot; respectively inputting the user questions of each round of conversation in the multiple rounds of conversations into a first feature extraction model, and respectively outputting a first feature vector corresponding to each round of conversation through the first feature extraction model; according to the sequence of each round of conversation, adopting a self-attention mechanism for the first feature vector corresponding to the conversation before each round of conversation, and generating a second feature vector corresponding to each round of conversation; inputting the behavior characteristics of the preset historical behaviors of the target user into a second characteristic extraction model, and outputting a third characteristic vector through thesecond characteristic extraction model; and determining the category of the multi-round conversation at least according to the second feature vector and the third feature vector corresponding to eachround of conversation. The classification method can guarantee the effect of classifying multiple rounds of conversations.

The invention relates to the technical field of garbage recycling, in particular to a garbage classification recycling treatment device. The garbage classification recycling treatment device comprisesa primary separation mechanism, a secondary separation mechanism and a screening mechanism. The primary separation mechanism comprises a separation box, a bag breaking assembly and a winnowing assembly. The screening mechanism comprises two conveying belts; a second limiting assembly is disposed at the upper end of each belt; the structures of the second limiting assembly are same as those of first limiting assemblies; a second screening assembly is disposed at the tail end of each conveying belt; and the structures of the second screening assemblies are same as those of first screening assemblies. Through repeated classification and separation, metals can be recycled, solid and liquid classification and separation of various garbage can be further performed, the garbage classification effect can be guaranteed through multi-step separation, the functions of the device can be added by additionally arranging the bag breaking assembly, and thus the device is more convenient to use.

The embodiment of the invention provides a network traffic classification method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a target classification model trained based on a loss function; classifying traffic data to be classified based on the target classification model; wherein the loss function is determined based on a weight parameter of atraining sample in the network traffic data. According to the method of the invention, the loss function is determined through the training sample weight parameters based on the classification model,and the to-be-classified flow data is classified through the target classification model trained by the loss function, so that the problems of sample imbalance and positive and negative sample proportion imbalance during model training are solved, the classification performance of a subclass data set is ensured, and accuracy of classification results is improved.

The invention discloses an online comment automatic reply method based on deep semantic matching, which is used for finding out online comments closest to input comment semantics in a database by combining sentence vector cosine similarity and multi-dimensional emotion matching degree. The method specifically comprises the steps that feature words of different topics are obtained through Canopy +Kmeans clustering, and on this basis, topic feature word expansion is conducted through a topic model CorEx based on priori knowledge. Meanwhile, a BERT-BiLSTM emotion analysis model is constructed, and multi-dimensional emotion analysis is carried out on online comments according to theme feature words obtained through clustering and by means of dependency syntactic analysis. Online comments arematched with the most similar semantics in the database by combining sentence vector cosine similarity and a multi-dimensional sentiment analysis result, data enhancement EDA operation is performed onmerchant replies of the comments, and sentences with the highest sentence vector cosine similarity are selected as automatic reply contents. According to the invention, automatic reply online comments can be provided for merchants conveniently, efficiently and accurately.

The invention relates to the field of information collection and application, in particular to an image compression method and device and a storage medium. The method comprises the following steps: inputting an image sample with a category attribute into a pre-constructed joint learning model; and carrying out joint training on the first initial deep learning model, the second initial deep learning model and the third initial deep learning model of the joint learning model to obtain a compression model comprising an image compression sub-model, an image reconstruction sub-model and a categoryrecognition sub-model. Practical application, and inputting the target image to be compressed into the compression model obtained by training, performing compression processing by utilizing the imagecompression sub-model of the compression model, and obtaining the target compression image corresponding to the target image according to the output of the image compression sub-model. The image compression method and device can realize image compression on the premise of maintaining the original image classification performance.

The invention provides a text classification method, a text classification device, a storage medium and electronic equipment, and relates to the technical field of computers. The text classification method comprises the steps: obtaining to-be-classified text data, and converting the to-be-classified text data into multiple types of coding features; carrying out the common coding on the multiple types of coding features to obtain a common classification vector of the to-be-classified text data; performing classification processing on the common classification vector to obtain a primary classification result corresponding to the common classification vector; and in combination with a keyword matching mode, classifying the primary classification result again to obtain a category label corresponding to the to-be-classified text data. According to the embodiment of the invention, the text classification accuracy can be improved.

The invention relates to an improved active power distribution network large-scale scene analysis method, and the method comprises the following steps: obtaining historical data which comprises wind power data, photovoltaic data and load data; generating a large number of scenes on the basis of the probability density distribution function of the wind power data, the photovoltaic data and the load data; and taking the maximum ratio LS of the inter-class distance L1 to the intra-class distance L2 as a selection basis of a k value, and performing scene reduction on the generated scene by using a k-means clustering algorithm to obtain a typical planning scene. Compared with the prior art, the method has the advantages that the scene reduction part is improved, and the optimal value of the k value is found according to the ratio LS of the inter-class distance L1 to the intra-class distance L2, so that the optimal clustering number can be obtained, the classification effect can be guaranteed, and the integrating degree with original data can also be guaranteed.

The invention discloses an image texture classification method based on Weber local binary counting. Multi-scale decomposition is firstly carried out on a texture image, and a differential excitation vector characteristic and a local binary counting histogram characteristic are built; L scales of Weber local binary counting descriptor characteristics are then built; and finally, a K nearest neighbor classifier is adopted to classify the extracted L scales of Weber local binary counting characteristics for the texture image, and a classification result is acquired. The method of the invention has the beneficial effects of effectively improving the texture image classification precision, effectively being adaptive to changes of a texture image imaging condition and effectively improving the classification speed.

The invention discloses a mechanical pump small sample fault diagnosis method based on WGAN-GP-C and metric learning, which is based on one-dimensional convolution and time series data design and comprises four parts of data preprocessing, data enhancement, metric learning and model optimization. The data preprocessing realizes data self-adaptive denoising and standardization; in the data enhancement part, network and structure modification is carried out based on WGAN-GP to obtain WGAN-GP-C, so that data expansion according to categories is realized, and edge boundary information of the data is enhanced; a metric network combines a residual idea and a spatial adaptive structure to realize feature mapping, and then a KNN algorithm is utilized to realize state classification; in the model optimization part, the network is optimized by using a weight quantization idea, the KNN is realized by using a Ball-tree, and training data are deleted according to important factors, so that the algorithm performance is improved on the whole. The method has good performance under the condition of rare data, is high in practical value, and can provide thoughts for related workers maintaining mechanical pumps.

The invention provides a data classification method and system. The method comprises the steps of adding keyword lists in project data; writing data classification marks into the keyword lists; classifying the project data according to the content of the keyword lists. According to the embodiment, keyword lists can be added in project data; by writing classification marks into the keyword lists, the data can be classified according to the classification marks in the keyword lists after the keyword lists are acquired. The method is simple and highly operable, guarantees the classification effect and improves the user experience greatly.

The present invention discloses a polarization SAR image classification method based on non-stationary modeling of a polarization scattering mechanism, in order to solve the problems that the existingpolarization SAR image classification is affected by noise and has low accuracy for the mixed pixels with no obvious main scattering mechanism. The implementation steps are: initially classifying measured images; estimating the auxiliary random field according to the polarization scattering characteristics, and associating the polarization scattering characteristics with the non-stationarity; dividing the pixel point stationarity by using the auxiliary random field; calculating correlation functions for the stationary pixel points to obtain a unitary potential energy function, a data item, and a binary potential energy function; calculating the membership degree for non-stationary pixel points; constructing a posterior probability model of a fuzzy triple-recognition random field (FTDF) model by using the obtained functions, and performing classification by using the maximum posterior probability criterion; and if it is marked that the random field converges, outputting a result, and otherwise repeatedly constructing the classification model according the iterative rule until the termination iteration requirement is reached, and outputting a classification result. The method disclosed by the present invention has high detection precision and good anti-noise performance, and can be used for polarization SAR image classification.

The invention discloses a chaotic baseband wireless communication decoding method based on a genetic optimization support vector machine. The method comprises the following steps: 1, carrying out communication setting and initialization; 2, after an initial population of the genetic algorithm is generated, initializing a population algebra k to be equal to 0; 3, solving a support vector machine classification model, evaluating individual fitness, storing an optimal individual and an optimal classifier, and judging whether the optimal individual fitness reaches a target value or not or whether the number k of iterations reaches the maximum algebra or not; 4, selecting, crossing and mutating a genetic algorithm to generate a new population, and turning to the step 3; step 5, carrying out symbol decoding by using an optimal classifier; 6, if n is smaller than LF-3, enabling n=n+1, and executing the step 5, or otherwise, turning to step 7; 7, enabling Fr = Fr+1, and turning to the step 2 if Fr is less than NF being the total frame number of transmission data, otherwise, ending the process, and finishing all decoding. According to the method, direct decoding can be carried out without channel information, and the decoding process is simplified.

The invention discloses a classification network and an implementation method and device thereof, and the method comprises the steps: constructing a classification network which comprises a Stem module, a plurality of Inception modules and a plurality of IR modules based on an IRv2 algorithm, wherein a batch standardization BN layer of the classification network uses a gamma parameter; initializing the classification network, and performing sparse training on the initialized classification network to obtain a sparse gamma parameter; on the basis of the sparse gamma parameter, carrying out the network slip pruning on the Stem module, the pruning Inception module and the plurality of IR modules, wherein network slip pruning is not carried out on the non-pruning Inception module. According to the technical scheme, the classification effect of the classification network can be ensured, and meanwhile, simplification of a network structure is realized, so the training efficiency is improved, the reasoning time and memory occupation are reduced, manpower, working hours and funds are saved, and the classification network can be deployed on hardware equipment with different computing power.