399results about How to "Improve learning ability" patented technology

The invention discloses an image segmentation method and system based on a convolutional neural network, and the method comprises the following steps of collecting a preset number of sample images, carrying out the normalization and data enhancement of the sample images, and obtaining the training sample data; training a U-Net convolutional neural network model with a residual block through the training sample data to obtain a trained U-Net convolutional neural network model; carrying out the pixel normalization processing on to-be-segmented images which are the same as the sample images; inputting the image to be segmented after pixel normalization processing into the trained U-Net convolutional neural network model, and finally obtaining a segmented image. The method provided by the invention has the relatively higher segmentation precision, and an end-to-end segmentation mode is adopted, so that the relatively higher segmentation efficiency is achieved.

The invention discloses an intensive learning based urban intersection passing method for a driverless vehicle. The method includes a step 1 of collecting vehicle continuous running state informationand position information through a photographing method, the vehicle continuous running state information and position information including speed, lateral speed and acceleration value, longitudinal speed and acceleration value, traveling track curvature value, accelerator opening degree and brake pedal pressure; a second step of obtaining characteristic motion track and the velocity quantity of actual data through clustering; a step 3 of processing original data by an exponential weighting moving average method; a step 4 of realizing the interaction passing method by utilizing an NQL algorithm. The NQL algorithm of the invention is obviously superior to a Q learning algorithm in learning ability when handling complex intersection scenes and a better training effect can be achieved in shorter training time with less training data.

The invention discloses a deeper layer network constructing method used for gender identification or age estimation based on human face. The method includes a step (101) dividing all training pictures into a plurality of groups; (102) extracting high layer features of a group of pictures based on a convolution neural network and thereby obtaining a first matrix composed of the high layer feature vectors, and extracting low layer and global features of the same group of the training images based on an artificial neural network and thereby obtaining a second matrix composed of the low layer feature vectors, obtaining a group of gender identification or age estimation results based on the extract first matrix, the second matrix and the defined judgment formula, wherein the values of a first weight matrix W1, a second weight matrix w2, an offset matrix b and an adjusting weight beta in the defined judgment formula are updated by utilizing an error back propagation algorithm and the final values of the parameters are obtained and the network construction is completed. Judgment of age and gender of a human face is performed based on the judgment formula determined according to the values of the parameters when the network construction is completed.

The invention discloses a face image super-resolution reconstruction method based on an attribute description generative adversarial network, and belongs to the field of digital image / video signal processing. The method is characterized in that a training stage comprises three parts of training sample preparation, network structure design and network training; the network structure design adopts agenerative adversarial network framework and is composed of a generative network and a discriminant network. The generation network comprises a face attribute coding and decoding module and a super-resolution reconstruction module; the discrimination network comprises an attribute classification module, an adversarial module and a perception module; wherein the network training process is carriedout in a mode that a generative adversarial network of a generative adversarial network framework and a discriminant network are alternately subjected to adversarial training; and a reconstruction stage: taking the LR face image and the attribute description information as input, and realizing image coding, attribute adding, image decoding and image reconstruction through a trained generation network. According to the invention, the enhancement of the face information of the low-resolution face image can be completed, and the accuracy of low-resolution face recognition can be improved.

The invention relates to a virtual learning environment micro-expression recognition and interaction method based on a double-flow convolutional neural network, and the method comprises the followingsteps: S1, carrying out the preprocessing of micro-expression data: carrying out the Euler video amplification of a micro-expression video, extracting an image sequence, carrying out the face positioning of the image sequence, and carrying out the cutting of the image sequence, and obtaining the RGB data of a micro-expression; extracting optical flow information from the data amplified by the Euler video to obtain an optical flow image of the micro-expression; s2, dividing the preprocessed data into a training set and a test set, and constructing a double-flow convolutional neural network by using a transfer learning method so as to learn space and time domain information of micro expressions; s3, carrying out maximum value fusion on the output of the double-flow convolutional neural network to enhance the recognition accuracy and obtain a final micro-expression recognition model; and S4, creating a virtual learning environment interaction system by using the micro-expression recognition model, and obtaining a user face image sequence through Kinect to carry out a micro-expression recognition task.

The invention discloses a fault diagnosis method of a high-voltage circuit breaker based on a depth belief network, which comprises the following steps: step 1, selecting a data sample required by anexperiment, and dividing the unified standardized sample data into a test sample and a training sample according to a specific proportion; Step 2: building and initializing the DBN deep belief networkfault diagnosis model; Step 3, inputting a large number of unlabeled samples or unlabeled samples in the pre-training set from the bottom of the model, and pre-training the RBM in the model by usinglayer-by-layer unsupervised greedy learning; Step 4: the whole model being fine-tuned by genetic algorithm; Step 5, the fault diagnosis model of the high-voltage circuit breaker obtained by training being classified to the fault samples of the test set in step 1, so as to obtain the fault classification result, and the diagnosis accuracy rate of the model being counted. The invention discloses a fault diagnosis method of a high-voltage circuit breaker based on a depth belief network, which can train a large amount of data samples to realize the fault diagnosis function of the high-voltage circuit breaker.

The invention discloses a rockburst grade predicting method based on an information vector machine to mainly solve the problem that in the current underground engineering construction process, the rockburst geological disaster prediction effect is not good. A rockburst evaluation index and a grading standard are selected, and domestic and overseas great deep rock engineering rockburst instances are widely collected to establish an abundant training sample library. A cross validation strategy is utilized for training an IVM model with the superior statistics mode recognition performance, accordingly, a nonlinear mapping relation between the rockburst evaluation index and the rockburst grade is established, model initial parameter setting and the training sample library are adjusted according to the training result, and the IVM model predicting the rockburst grade is finally established. By means of the method to predict the rockburst grade, the complex mechanical analysis or calculation is not needed, only the input feature vector of a sample to be predicted needs to be input in the prediction model, and then the prediction value of the rockburst grade can be obtained. The method is economical, efficient and high in prediction precision and has good engineering application prospects.

The invention provides a dairy cow individual recognition method based on a deep convolutional neural network, and relates to an image recognition method in image data processing. According to the dairy cow individual recognition method, the dairy cow individual can be effectively recognized by extracting characteristics by virtue of a convolutional neural network in deep learning and combining the characteristics with textural features of dairy cows. The dairy cow individual recognition method comprises the following steps: collecting dairy cow data; preprocessing a training set and a test set; designing the convolutional neural network; training the convolutional neural network; generating a recognition model; and recognizing the dairy cow individual by virtue of the recognition model. By virtue of the dairy cow individual recognition method, the defects that an existing algorithm for processing dairy cow images by virtue of an image processing technique is single, the stripe characteristics of the dairy cows are not adequately and well combined with an image processing technique and a mode recognition technique, and therefore, the recognition rate of the daily cows is low are overcome.

The invention discloses a rolling bearing fault diagnosis method based on a sparse encoder and a support vector machine. A deep learning and autonomous cognition method based on a stacked sparse automatic encoder is adopted, essential characteristics of input data are automatically extracted from simplicity to complexity and from a low level to a high level, rich information hidden in known data is automatically dug out, the deep learning is adopted for extracting the characteristics and features learnt by two layers are integrated together to form input of the support vector machine, and through classification by the support vector machine, the working state and the fault type of the rolling bearing can be judged. The method of the invention can improve the fault characteristic extraction efficiency and the accuracy.

For accurately measuring the power consumption, the invention provides a power consumption prediction method for a system. The method comprises S1 data acquisition: acquiring the power utilization information of residents, industrial users and business users through an intelligent ammeter, acquiring the influence factor information through an external system, and generating a history table for the users at the same time; S2 analysis of power utilization rules for the users: performing fitting analysis of the acquired power utilization information of the users by means of combination with the history table, and obtaining the power utilization rules for the users through analysis; S3 analysis of influence factors: setting a load fluctuation range according to the user type, and performing analysis of influence factors for the time point surpassing the fluctuation range to obtain the influence value of each influence factor for the power consumption of the users; and S4 power consumption prediction for a system: predicting the future short-term power consumption for the system by means of combination of analysis of power utilization rules and analysis of influence factors. The prediction method shows a brand new development direction for short-term power consumption prediction in future, and can be widely applied to the prediction field with great significance.

The invention provides an air conditioning system sensor fault diagnosis method based on a wavelet neural network. The method is characterized by comprising the steps that comprehensive analysis and diagnosis are conducted for defining an air conditioner state control, a diagnosis condition set and a decision event set; sensor fault classification and characteristic description of an air conditioner system are used for classifying fault types; sensor fault diagnosis of the air conditioning system is used for generating a fault diagnosis flow; a network structure and a primary function are determined; selection and pretreatment of training samples are conducted; and an alarming threshold is set. The method has good diagnosis effect on abrupt changing faults such as bias faults and completefaults. Drift biases can be detected as long as the drift distance exceeds the alarming threshold. The method also has good fault detection effect on precision decrease faults of sensors.

The invention discloses a composite fault diagnosis method and device based on a deep decoupling convolutional neural network. The method comprises the following steps: 1, data set collection and calibration: respectively collecting vibration acceleration signals under single and composite fault experiment conditions of a rotary machine to obtain a plurality of samples, respectively forming a training set and a test set, and respectively setting category labels; 2, building a one-dimensional deep convolutional neural network model; 3, constructing a decoupling classifier with a multi-label output characteristic under a keras framework; 4, training a network model, and finally obtaining an optimal deep decoupling convolutional neural network model; and 5, performing intelligent diagnosis and classification on the composite fault, and outputting to obtain a real-time diagnosis result of the composite fault. According to the method, on the premise that only a single fault signal is used for training the deep decoupling network model, each single fault feature in the composite fault signal is extracted, and decoupling and classification of the composite faults are achieved through a decoupling classifier.

The invention discloses a planning method of an intelligent vehicle autonomous running dynamic trajectory and belongs to the technical field of intelligent transportation. The planning method of the intelligent vehicle autonomous running dynamic trajectory comprises (1) a step of generating a trajectory of an intelligent vehicle through searching for an optimum solution of a mathematical model by means of an optimum control method, (2) a step of extracting features of the trajectory generated in the first step, and (3) a step of judging whether the features meet presupposed constraint conditions or not; if the features meet the presupposed constraint conditions, regarding the trajectory as a final output trajectory; if the features do not meet the presupposed constraint conditions, adjusting weight W<1>, W<2>, and W<3> according to presupposed weight adjustment rules and then returning to the first step. The invention further discloses a planning system of the intelligent vehicle autonomous running dynamic trajectory. The planning system of the intelligent vehicle autonomous running dynamic trajectory comprises a trajectory generation module, a feature extraction module, an estimation module, and a weight adjustment module. Compared with the prior art, the planning method of the intelligent vehicle autonomous running dynamic trajectory and the system of the same have a stronger learning ability and a better adaptability to the unknown environment.

The invention provides a three-dimensional reconstruction method based on a convolution neural network, belonging to the technical field of computer vision. Three-dimensional features are more accurate and retrieval accuracy is higher. Compared with the popular feature extraction network, the network learning ability of the method is stronger, and the feature information of the object point cloudextracted through the network is richer. The algorithm has good real-time performance, whether modeling, feature extraction, database retrieval or final model registration can be completed in a shorttime. Moreover, the accuracy of the network model proposed in this method is better than that of other models based on depth learning, which shows that the network structure can learn the data distribution rules directly from the three-dimensional point cloud. Compared with the traditional feature extraction method, the proposed method based on convolution neural network can significantly reduce the computational time, and the Euclidean distance retrieval algorithm can also improve the retrieval efficiency.

The invention relates to a micro-expression recognition method based on a space-time appearance movement attention network, and the method comprises the following steps: carrying out the preprocessingof a micro-expression sample, and obtaining an original image sequence and an optical flow sequence with a fixed number of frames; constructing a space-time appearance motion network which comprisesa space-time appearance network STAN and a space-time motion network STMN, designing the STAN and the STMN by adopting a CNN-LSTM structure, learning spatial features of micro-expressions by using a CNN model, and learning time features of the micro-expressions by using an LSTM model; introducing hierarchical convolution attention mechanisms into CNN models of an STAN and an STMN, applying a multi-scale kernel space attention mechanism to a low-level network, applying a global double-pooling channel attention mechanism to a high-level network, and respectively obtaining an STAN network added with the attention mechanism and an STMN network added with the attention mechanism; inputting the original image sequence into the STAN network added with the attention mechanism to be trained, inputting the optical flow sequence into the STMN network added with the attention mechanism to be trained, integrating output results of the original image sequence and the optical flow sequence through the feature cascade SVM to achieve a micro-expression recognition task, and improving the accuracy of micro-expression recognition.

The invention relates to a line loss prediction method of a BP neural network optimized by a genetic algorithm, belonging to the technical field of line loss prediction and machine learning of a powernetwork. At first, the method obtains line characteristic parameters through the power network line, and establishes a prediction model of a BP neural network for the characteristic parameters. Then,the individual length is determined according to the weights and thresholds in the topology of BP neural network, and the individual is selected, crossed and mutated by genetic algorithm with real number coding. Finally, the convergence condition is judged and the optimal individual is selected. Then the BP neural network is initialized and trained with the variable learning rate momentum BP algorithm until the network converges. The genetic algorithm is used to optimize the BP neural network algorithm to predict the line loss. The method has the advantages of improved prediction accuracy, less calculation time and enhanced stability. Therefore, the method has certain research significance.

The invention relates to an electrocardio-signal emotion recognition method based on deep learning. The electrocardio-signal emotion recognition method comprises the following steps that under variousemotion picture inducing conditions, electrocardio data of a testee under different emotions is acquired through the chest or the wrist, the acquired electrocardio data is segmented into electrocardio signals fixed in length, and corresponding labels are made for the electrocardio data obtained under different emotions; the acquired electrocardio data is preprocessed, baseline drift removal basedon wavelet transformation is firstly performed, and noise removal is performed; data standardization is performed, normalization processing is conducted on noise-removed electrocardio data; a model training stage is performed.

The invention relates to a full convolution network based remote-sensing image land cover classifying method. The method comprises the following steps: S1, performing data enhancement on a data set with limited data quantity; and generating a training set of which the data quantity and quality meet the training requirement; S2, combining the improved full convolution network FCN4s and the improvedU type full convolution network U-NetBN; and building a remote-sensing image land cover classifying model; S4, minimizing the cross entropy loss as the decrease of random gradient; learning the optimal parameters of the model to obtain the trained remote-sensing image land cover classifying model; and S4, performing pixel class classifying prediction on the predicted remote-sensing image throughthe trained remote image land cover classifying model. According to the method, the properties of the FCN full convolution network and the U-Net full convolution network are combined, so that the remote-sensing image land cover classifying performance can be improved.

The invention relates to an X-ray weld defect detection method based on a convolutional neural network, and the method comprises the following steps that a weld image data set containing different types of weld defects is built, and all weld images in the data set are marked with weld labels; an AF-RCNN model is established, wherein the AF-RCNN model comprises a backbone network module, a region generation module and a target classification and position regression module; the backbone network module adopts a residual network (ResNet) and a feature pyramid network (FPN) structure, an efficient convolution attention module is introduced between the residual network (ResNet) and the feature pyramid network (FPN) to enhance the learning ability of the network for unobvious defects and small target features, and a CIOU loss function is introduced to enhance the positioning ability of aiming frames; and an AF-RCNN model is trained by using the established data set, wherein the AF-RCNN model is used for classifying and positioning weld defects. The accuracy of all defects reaches 94% or above, and the detection speed is 11.65 FPS.

The invention discloses an LSTM and SAE-based telephone voice emotion analysis and recognition method; according to the method, voice information is subjected to a series of preprocessing of sampling,quantifying, pre-emphasis, framing, windowing and the like, and then fast fourier transformation is carried out on the voice information to obtain frequency domain characteristics of the voice information, and next, the voice characteristic parameters MFCC are extracted. By adoption of the method, an LSTM+ SAE neural network model is established to train the extracted MFCC characteristic parameters to obtain depth characteristic information of voice signals; the classification accuracy is obtained by combining a full-connection layer and a softmax regression algorithm, and model training is completed; and finally, the MFCC characteristic parameters to be tested are input into a trained model, emotion analysis is carried out on the telephone voice, and the emotion of a speaker is judged.

The invention provides an unmanned aerial vehicle flight state prediction method and system based on LSTM. The method of the invention comprises the following steps: step 1, constructing an action label dictionary; 2, collecting flight state information and motion data of the unmanned aerial vehicle; 3, preprocessing the collected flight state information of each unmanned aerial vehicle; 4, forming a data set in the form of a data matrix; 5, randomly dividing the data set into 70 percent of the training set and 30 percent of the verification set; obtaining the improved LSTM model of a long-short-term memory network with variance by training. 6, tuning the super parameters of the model to obtain a final model by using a verification set; 7: collecting the UAV flight state information of theUAV at the current time which needs to be predicted, and inputting the information into the final model after preprocessing and filtering to obtain the UAV flight motion prediction result. The technical proposal of the invention solves the problem that the existing control model cannot predict the flight action according to the collected flight state data.

The invention discloses a method and device for predicting the life of a satellite lithium ion battery. The method comprises the following steps: analyzing the characteristics of cycle life test data of the satellite lithium ion battery so as to obtain a fault evolution characteristic quantity; carrying out de-relax effect on capacity degradation data corresponding to the fault evolution characteristic quantity; carrying out least squares support vector machine LSSVM unsupervised learning training according to the capacity degradation data after the de-relax effect so as to complete the construction of an LSSVM model for life prediction; predicting the battery capacities of the battery in different periods through the LSSVM model, and carrying out battery failure threshold value extrapolation according to the battery capacities so as to realize the real-time prediction of the remaining useful life RUL of the lithium battery. According to the method provided by the invention, the LSSVM model for life prediction is constructed after de-relax effect is carried out on the capacity degradation data, and then life prediction is carried out through the model, so that the prediction result is correct and the problem that the process of measuring the life of the lithium ion battery is not correct is solved.

The invention aims to provide a polarization SAR image classification method based on an SSAE and an FSALS-SVM. According to the polarization SAR image classification method, a multi-implicit-strata structure of the stacked sparse automatic encoder (SSAE) is used for obtaining depth features which have the capacity for describing original data more intrinsically and are more suitable for classification, the fast sparse approximation least square-support vector machine (FSALS-SVM) which can obtain sparse solutions is used for replacing the Softmax commonly used in traditional deep learning and being combined with the SSAE, the classification accuracy of polarization SAR images is improved, the defect that a traditional polarization SAR image classification method based on pixels is greatly affected by speckle noise is overcome to a certain degree, and therefore coherence of homogeneous areas in classification result images is ensured.

The invention relates to a color-spun yarn color matching method based on a least square support vector machine. The method is characterized by comprising the following steps that reference color fibers are determined, and the corresponding color data value is acquired; a part of or all reference color fibers are selected, combined in different ways and mixed in different mixing ratios for spinning and sample preparation, and the reflectance values of all samples are measured; the reflectance value of a standard sample and the corresponding proportional relation serve as training samples, and an LS-SVM function model is trained; a corresponding sample recipe C is calculated; according to the prediction recipe C, sample making is conducted, a simulated sample is obtained, and the reflectance value of simulated sample is measured; through the comparison, the chromatic aberration between a target sample and the simulated sample is obtained, if the chromatic aberration meets a preset value, the proportion is determined as an optimum recipe, otherwise, recipe modification is conducted; the optimum recipe is input in an LS-SVM training set for training. The types of the reference color fibers adopted by the method are less, after the recipe obtained by calculation is verified and modified, the accuracy of the optimum recipe is high, color matching is simple and reliable, and the efficiency and the practicability are high.

The invention relates to a method for monitoring health of a rotary machine suitable for a working condition changing condition. The method comprises the steps that (1) a monitor model is constructed, wherein a relevance vector machine is used for fitting the function relation of health characteristic parameters and a working condition, the function relation is used as the parameters of a self-adaption threshold model, and the self-adaption threshold model is constructed; (2) a health state is monitored, wherein through test signals of the rotary machine to be detected, the constructed self-adaption threshold model is used for detecting whether test data exceed a threshold value or not, the machine is judged to be healthy if the test data do not exceed the threshold value, and otherwise, the machine is judged to break down. According to the method, the relevance vector machine is used for fitting the mean value of health characteristics and the function relation of standard deviation changing along with working condition parameters, and the method has the advantages that the relevance vector machine has strong learning capacity, the problems of the local minimum, the over-fitting and the under-fitting of a neural network can be solved, the relevance vector machine has better sparsity than a support vector machine, and obtained results are simpler and more practical. The method has the advantages of being high in monitor precision and capable of being used under rotating speed changing and load changing conditions.

The invention discloses a UUV (unmanned underwater vehicle) dynamic planning method based on an LSTM-RNN (long short term memory-recurrent neural network), and belongs to the field of unmanned underwater vehicles. The UUV dynamic planning method includes the steps: (1) selecting a geometric model to build an obstacle environment model; (2) building a UUV dynamic planner for acquiring a data set byan ant colony algorithm; (3) designing an LSTM-RNN model for dynamic planning; (4) acquiring the data set; (5) training the LSTM-RNN by data of a training set in the data set to obtain the dynamic planner based on the LSTM-RNN; (6) inputting sonar detection information and target point information to the dynamic planner based on the LSTM-RNN to obtain the navigational direction and the navigational speed of a UUV at a next time. The method has strong learning capacity and further has quite strong generalization capacity, so that the implemented dynamic planner is applicable to complex environments. The requirement of real-time performance is met, and planned routes conform to movement characteristics of the UUV.

The invention relates to the field of ecological environment monitoring, and discloses a vegetation classification method based on a machine learning algorithm and multi-source remote sensing data fusion, which is used for efficiently realizing identification and classification of vegetation types in a target area. The method comprises the following steps: acquiring a low-altitude remote sensing image of terrestrial plants in a sample area by using an unmanned aerial vehicle, and acquiring a digital orthoimage and a digital surface model of the sample area based on the low-altitude remote sensing image; extracting elevation information of the digital surface model; acquiring an SAR image of a sample region corresponding to the aerial photography time of the unmanned aerial vehicle by utilizing satellite remote sensing; carrying out wave band and image fusion on the digital orthoimage, the elevation information and the SAR image; performing inversion model training and inversion model precision evaluation on the fused image through sample area actual measurement data and a machine learning algorithm to obtain an inversion model meeting requirements; and finally, classifying terrestrial plants in the target area based on the inversion model. The method is suitable for terrestrial plant ecological environment monitoring.

The invention relates to an image detection method and device, computer equipment and a storage medium. The method comprises the steps of obtaining at least two medical images of a to-be-detected object, and performing segmentation processing on the at least two medical images to obtain target segmentation images; wherein each target segmentation image comprises a region of interest; inputting each target segmentation image into a neural network model for feature extraction to obtain features of the region of interest; combining the features of the region of interest with features in clinicalinformation data, and performing feature selection processing on the combined features to obtain target features; wherein features in the clinical information data are features obtained after clinicaldetection is performed on the to-be-detected object; and inputting the target features into a classifier for classification, and determining the category of the region of interest. The method can improve the detection precision.

The invention discloses a wavelet neural network-based distribution network single-phase short circuit line selection method, which belongs to the technical field of distribution network protection. The method comprises the following steps: 1) db4 wavelets are selected to serve as a wavelet packet basis function to decompose transient zero-sequence current in each feeder line of the distribution network, and the sampling frequency is 10KHz; 2) the modulus maximum of the transient zero-sequence current is calculated; 3) the modulus maximum obtained in the second step and the polarity of the modulus maximum are used for training the neural network; and 4) the BP neural network after the training in the third step is applied to the distribution network with single-phase short circuit, and a fault line is determined according to a different output result. The method of fault line selection by using the wavelet neural network disclosed by the invention has good reliability and practicability, a single-phase grounding fault line can be effectively eliminated, stable operation of the distribution network system is facilitated, and an important role is played in planning and applications of a distributed power supply.

The invention discloses a deep network model constructing method, which comprises the steps of: step S1) establishing a deep network model used for facial expression identification, and initializing parameters of the deep network model, wherein the deep network model comprises a convolutional neural network used for extracting high-level features of pictures, a reconstructed network used for extracting low-level features of the pictures and a joint decision network used for identifying facial expressions; step S2) dividing all training pictures into N groups; step S3) sequentially inputting each group of the pictures into the deep network model, and training parameters in the deep network model based on a gradient descent method; step S4) regarding the parameters of the deep network modelobtained in the step S3) as initial values of model parameters, and re-dividing all the training pictures into N groups, then jumping to the step S3), and carrying out the process repeatedly until allthe trained model parameters no longer change when compared to the initial values of the model parameters. The invention further discloses a facial expression identification method and a facial expression identification system.