32results about How to "Reduce the impact of classification" patented technology

The invention discloses a multi-class Chinese text classification method fusing global and local features. The method comprises the following specific steps: obtaining text data and preprocessing thetext data; carrying out vectorization representation on the preprocessed text; obtaining context semantic features of the text by using a bidirectional long-short-term memory network with an attentionmechanism; extracting global semantic features with local semantic information on the basis of the context semantic features by utilizing a wide convolutional neural network; and inputting the finalfeature vector into a classification layer to realize text classification. According to the method, global semantic features are captured first and then local semantic features are captured step by step, so that text features can be better extracted, and the text classification precision is further improved. The method well solves the problems that in the prior art, in long text and multi-class Chinese text classification, semantic key feature extraction is difficult, and the classification effect is poor.

The present invention discloses a polarization SAR ground object classification method based on a self-step learning convolutional neural network, in order to mainly solve the problems that the priorart has low accuracy in classifying complex ground object scenes and is heavily affected by noise. The implementation scheme comprises: 1, obtaining a polarization scattering matrix S and a pseudo color RGB image under the Pauli basis from original complete polarization SAR data; 2, constructing a three-dimensional matrix to form a sample set for each pixel, and constructing a training sample setand a test sample set; 3, constructing a convolutional neural network and training the convolutional neural network based on self-step learning to accelerate network convergence and improve the generalization ability of the network; and 4, classifying the test samples by using the trained convolutional neural network to obtain a final complete polarization SAR ground object classification result.According to the method disclosed by the present invention, accuracy for classifying the target ground objects of complex ground object scenes in the polarization SAR image is improved, and the methodcan be used for feature classification and target recognition.

The invention discloses an intelligent steel cord conveyer belt defect identification method and an intelligent steel cord conveyer belt defect identification system. The identification method includes the following steps: (1) electromagnetic loading; (2) defect signal acquisition; (3) feature extraction; (4) training sample obtainment; (5) class priority determination; (6) multi-class model establishment; (7) multi-class model training; (8) real-time signal acquisition and synchronous class: electromagnetic detection units are adopted for real-time detection, detected signals are synchronously inputted into a data processor, features are extracted and then sent into established multi-class models, and the defect class of a detected conveyer belt is automatically outputted. The identification system comprises an electromagnetic loader, a plurality of electromagnetic detection units, the data processor and an upper computer, the data processor can automatically output the defect class of the detected conveyer belt, and the upper computer bidirectionally communicates with the data processor. The design of the invention is reasonable, the invention is easy to operate and convenient to put into practice, moreover, the using effect is good, the practical value is high, the reliability of conveyer belt defect detection is enhanced, and the efficiency of defect identification is increased.

The invention relates to a cloud computing fault data detection method and system, and belongs to the field of cloud fault detection. The cloud computing fault data detection method comprises: a cloudcomputing fault training data processing step: processing the data in a cloud computing fault training data set to obtain a degree of membership of each cloud computing fault training data and the fault feature weight of each fault category; a to-be-detected cloud computing data fault category judgment step: judging the fault category of the to-be-detected cloud computing data according to the processing result of the cloud computing fault training data in combination with a cloud computing fault training data set expansion rule; and a cloud computing fault training data set expansion step: adding the to-be-detected cloud computing data satisfying the cloud computing fault training data set expansion rule and the category thereof to a fault training data set. By adoption of the cloud computing fault data detection method and system, the fault training data model is perfected beneficially, and new faults are identified.

The invention relates to an object classification method and device based on artificial intelligence, a computer readable storage medium and computer equipment, and the method comprises the steps: obtaining a to-be-processed image which comprises a target detection object; segmenting a target detection object image of the target detection object from the to-be-processed image; inputting the targetdetection object image into the feature object prediction model to obtain a feature object segmentation map of a feature object in the target detection object image; obtaining quantitative feature information of the target detection object according to the target detection object image and the feature object segmentation image; and classifying the target detection object image according to the quantitative feature information to obtain category information of the target detection object in the to-be-processed image. According to the scheme provided by the invention, unnecessary image data inthe to-be-processed image can be effectively reduced, the influence of the unnecessary image data in the to-be-processed image on object classification is reduced, and the classification accuracy of the target detection object in the to-be-processed image is improved.

The invention discloses a local node mining method based on distributed data stream processing. The local node mining method comprises the following steps: S1, selecting a data sequence; S2, establishing a big data stream classification model; S3, a local node mining method; and S4, a center node mining method based on the unbalanced data stream. According to the local node mining method, the problem of local node mining in power grid equipment online monitoring data stream distributed processing is solved based on the data stream cleaning method for the power grid equipment from the perspective of large data stream distributed processing.

The invention provides an SAR image target classification method based on a deep convolutional neural network. The SAR image target classification method is used for improving SAR image target classification precision. The method comprises the following implementation steps: obtaining a training sample set and a test sample set which comprise SAR target images; removing background clutters of eachSAR image in the training sample set and the test sample set; constructing a deep convolutional neural network model containing an Enhanced-SE layer transformed by a sigmoid activation function to form; training the deep convolutional neural network model; and classifying the test sample set by using the trained deep convolutional neural network model. According to the method, when background clutters in the SAR target image are removed through the morphological closed operation method, the edge gap of the target area is fused, the internal defect of the target area is filled, and the shape features of the target area are effectively reserved; Athe Enhanced-SE layer is formed by modifying the sigmoid function, the deep convolutional network is inhibited from automatically extracting redundant features, and the SAR image target classification precision is improved.

The invention relates to a multi-feature fusion overhead pedestrian detection method based on aggregated channel features and a gray level co-occurrence matrix. The method comprises extracting ACF features of a plurality of aggregation channels in a sample training set, obtaining aggregation channel feature vectors and gray level co-occurrence matrix feature vectors, sending the two vectors to a soft cascade Adaboost classifier for training, and obtaining classifier 1 and classifier 2; reading an image to be measured, extracting ACF features of the image to be measured, and obtaining an aggregation channel feature vector; sending feature vectors of aggregation channels into a classifier to classify, and obtaining candidate coordinates and target windows. The eigenvector of gray level co-occurrence matrix is obtained and sent to classifier 2 to eliminate background interference, and the output result of the final target is obtained. As that color, the gradient direction histogram, the gradient and the texture feature are fused, the background similar to the human head is filter out, the missed detection and the false detection rate of the classifier are effectively reduced, and thedetection performance of the pedestrian overlooking when a plurality of interference backgrounds exist is improved, and the method is stable, reliable and efficient, and has strong practical application value.

The invention discloses an anti-attack defense method based on a neuron activation rate. The method comprises the following steps: 1), obtaining an image data set, carrying out the preprocessing, and constructing a training set and a test set; 2) constructing a classifier model of a CNN network structure; 3) constructing a dichotomy discriminator D, and judging the output N of the classifier model about the input book x; 4) constructing an overall loss function according to the classifier model and the loss function of the dichotomy discriminator D, and determining an overall optimization target; carrying out joint training on the overall model by adopting the training set, and when the overall loss function is converged, stopping training and storing the model; and 5) inputting a test set to test the trained classifier model, if the classification accuracy is lower than a threshold value, modifying the lambda of the overall loss function, and training again according to the step 4) until the classification accuracy is higher than the threshold value. The defense method provided by the invention has good applicability and accuracy, and has a good defense effect on attack resistance.

The invention provides an electroencephalogram recognition method and system based on graph convolution and a gating circulation unit. The method comprises the steps of: obtaining electroencephalogram data, and performing the preprocessing of the electroencephalogram data, wherein the preprocessing includes filtering, benchmark correction, mean value removal, electrooculogram artifact elimination and normalization; and inputting the preprocessed electroencephalogram data into a plurality of graph convolutional networks for relatively independent feature extraction, stacking features extracted by the graph convolutional networks into a feature matrix, inputting the feature matrix into a gating circulation unit for classification and recognition, and outputting a recognition result. According to the invention, the recognition rate of the electroencephalogram signals can be effectively increased.

The invention discloses an electrical load type method. An improved decision tree classifier or a composite optimization decision tree classifier is adopted to carry out electrical load type identification, input characteristics of the classifier comprise starting current characteristics of an electrical load and steady-state current spectrum characteristics of the electrical load, and the starting current characteristics comprise the starting process time, a starting current maximum value and the starting current maximum value time; all input characteristics of different electrical load typescan fall into corresponding intermittent input characteristic overlapping areas; or when part of the input features fall out of the effective intervals of the corresponding interval input features and the other input features fall into the corresponding interval input feature overlapping areas, electrical load type identification is completed, and the generalization ability and the identificationaccuracy are high.

The invention discloses a garbage recognition and classification method based on deep learning, and the method comprises the following steps: 1, carrying out the preliminary classification of garbage;2, respectively acquiring the preliminarily classified garbage data based on a data acquisition module. Through preliminary classification of garbage, metal and nonmetal in the garbage are separated,recognition is easier during classification and recognition, the influence of the nonmetal garbage on metal garbage classification is reduced, and then recyclable metal garbage classification is moreaccurate. Through data detection of the non-metal garbage, kitchen garbage and household garbage contained in the non-metal garbage can be conveniently recognized and separated, then only recyclablenon-metal garbage such as plastic, glass and paper is left, and when the recyclable non-metal garbage is classified, the influence of the kitchen garbage and the household garbage is reduced as well;and thus, the garbage can be identified and classified more accurately based on deep learning.

The invention discloses a forest smoke and fire real-time monitoring method based on robust multi-view, and belongs to the technical field of forest smoke and fire safety monitoring. The method comprises the following steps: firstly, uniformly arranging forest smoke and fire data collection modules in a forest, and each forest smoke and fire data collection module comprising a forest smoke and fire image data collector; setting a robust multi-view detection module on a background server; inputting forest smoke and fire image data to the background server through a data transmission module by the forest smoke and fire data collection modules; and finally, enabling a robust multi-view detection module located on the background server to process and detect the input forest smoke and fire image data so as to judge whether a smoke and fire accident exists in the forest or not. The robustness multi-view detection module in the method is high in robustness, the forest smoke and fire accidentscan be monitored in real time, the monitoring precision is high, and high expandability is achieved.

The invention discloses an electrical load type identification device. The device comprises an information acquisition module, an information processing module and a communication module, wherein theinformation processing module adopts an improved decision tree classifier based on a membership deviation product or a composite optimization decision tree classifier based on the membership deviationproduct and Bayesian to carry out electrical load type identification; the input characteristics of the classifier comprise the starting current characteristics of the electrical load and the steady-state current spectrum characteristics of the electrical load, and the starting current characteristics comprise the starting process time, a starting current maximum value and the starting current maximum value time. When all input characteristics of different electrical load types can fall into corresponding intermittent input characteristic overlapping areas; or when part of the input characteristics fall out of the effective intervals of the corresponding intermittent input characteristics and the other input characteristics fall into the corresponding intermittent input characteristic overlapping areas, the electrical load type identification is completed.

The invention discloses an electrical load starting operation identification device. The electrical load starting operation identification device comprises an information acquisition module, an information processing module and a communication module; the information processing module adopts an improved decision tree classifier based on the membership deviation quadratic sum or a composite optimization decision tree classifier based on the membership deviation quadratic sum and Bayesian to carry out electrical load starting operation identification; the input characteristics of the classifiercomprise starting current characteristics of the electrical load and steady-state current spectrum characteristics of the electrical load, and the starting current characteristics comprise the starting process time, a starting current maximum value and the starting current maximum value time; all input characteristics of different electrical load types can fall into corresponding intermittent input characteristic overlapping areas; or when part of the input characteristics fall out of the effective intervals of the corresponding intermittent input characteristics, and when the other input characteristics fall into the corresponding intermittent input characteristic overlapping areas, the electric load starting operation identification is completed.

The invention discloses a coupler yoke breaking identification method based on image processing, and belongs to the technical field of image detection. The objective of the invention is to solve the problem of low accuracy of an existing coupler yoke fracture detection method. The method comprises the following steps: firstly, enhancing a target image, extracting a boundary region of a supportingplate and a background by adopting a local adaptive threshold value, counting a segmentation result in the horizontal direction, searching upper and lower boundaries to obtain an accurate positioningintercepted image, then calculating a pixel variance in the column direction to obtain a variance curve, and traversing the pixel variance curve to find the starting and ending positions of the jump point of the variance curve, correspondingly expanding L pixels from the starting position in the accurate positioning interception image to serve as the starting coordinates of the suspected fault area, and intercepting a suspected fault area sub-image according to the obtained coordinates; and then based on the extracted features of the suspected fault area sub-graph, identifying the fault by using an SVM classifier. The method is mainly used for coupler yoke breakage identification.

The invention provides a classification method and device for objects in road point cloud, electronic equipment and a storage medium. The classification method comprises: collecting initial point cloud data; dividing the initial point cloud data according to a first preset division interval to obtain a point cloud profile; determining a target two-dimensional image corresponding to the point cloud profile according to the number of point clouds in each pixel included in the point cloud profile and a mapping relation between the point clouds and the gray values; slicing the target two-dimensional image along a preset direction according to a second division preset interval, and determining a pavement layer according to the number of point clouds included in each slice layer; and denoising and classifying the target two-dimensional image according to the pavement layer and the preset height distance, and determining classification information of the pavement object in the vehicle driving process. By adopting the technical scheme provided by the invention, the two-dimensional point cloud profile with a certain width is taken as a processing unit, and under the condition that the vehicle-mounted point cloud data volume is huge, the operation space is ensured, and the processing efficiency is improved.