546 results about "Robustification" patented technology

The invention relates to the technical field of equipment fault monitoring, and particularly discloses an equipment fault warning and state monitoring method. The equipment fault warning and state monitoring method is characterized by including two processes of model building and model operation; the process of model building includes the steps of acquiring training data, performing data preprocessing operation on the training data, then adopting a nonparametric learning algorithm to select a memory matrix, training a residual generator and acquiring a residual threshold of each parameter; the process of model operation includes the steps of acquiring real-time data, performing data preprocessing operation on the real-time data, then calculating all parameter residuals of the real-time data, analyzing the residuals to judge whether the equipment state is normal or not, and further positioning fault causes. The method has the advantages of data-driven method based universality, robustness and high adaptive capability, the shortcoming that warning results are difficult to analyze and explain is avoided, and additionally accuracy and efficiency of fault warning are both improved due to introduction of the nonparametric learning algorithm.

Provided is a method and apparatus for support recovery of jointly sparse signals from a plurality of snapshots, thereby enhancing a capability for reconstructing a support in a variety of circumstances, by providing enhanced robustness against noise and perturbation, and/or enhanced computational efficiency. The method may include partial support recovery using a compressed sensing-multiple measurement vector (CS-MMV) scheme; and a complementary support recovery and sparsity level estimation. The complementary support recovery may use subspace information extracted from the plurality of snapshots and partial support information. The total number of elements in the partial support and in the complementary support may be equal to the sparsity level.

The invention discloses a real-time monitoring method for the energy consumption of a public building based on data mining, belonging to the technical field of building energy saving. The method disclosed by the invention comprises the following steps: S1. establishing a building energy consumption mode judgment tree; S2. collecting building energy consumption data in real time; and S3. judging whether the current building energy consumption data are energy consumption abnormal points or not, carrying out the mode matching on the current building energy consumption data and the building energy consumption mode judgment tree and judging whether the current building energy consumption data are isolated points or not. In the method disclosed by the invention, the specific energy consumption mode of the building is identified by carrying out the cluster analysis on the historical energy consumption data; the building energy consumption mode judgment tree is obtained by classifying the data; the mode matching is carried out on the energy consumption data which are dynamically collected in the real-time monitoring course for the energy consumption of the building; and the isolated point analysis is carried out on the energy consumption data and the historical data which have the same mode, thereby judging whether the current building energy consumption data are abnormal or not. The method disclosed by the invention has the characteristics of good real-time characteristic, generality and robustness.

The invention discloses a single detection method for a multi-direction scene based on the fully convolutional network. A fully convolutional single detection network model is constructed, end-to-endtraining can be carried out via single network needless of multi-step processing, a multi-scale feature extraction layer is combined with a text box prediction layer to detect multi-direction naturalscene characters in different sizes, length-width ratios and resolutions, a polygonal enclosure box is combined with characters to introduce less background interference, and a final text detection result can be obtained via simple non-maximal-value inhibition operations. Compared with the prior art, the detection method is simple and effective in structure, improves the accuracy, detection speedand robustness, and is high in practical application value.

The invention relates to a photovoltaic array fault diagnosis and early warning method comprising the following steps: combining an Elman nerve network optimized by a non-linear least square method and a decision tree with experience knowledge so as to form a fault diagnosis model; collecting present photovoltaic array operation data and meteorology data, and computing errors when compared with historical normal state data; using the fault diagnosis model to obtain the corresponding fault type and credibility when the error is bigger than a threshold; finally integrally evaluating so as to obtain the final fault type credibility, and selectively carrying out fault early warning according to the credibility values; updating a fault knowledge base according to the field actual measurement conditions. The method combines the LM-Elman nerve network and the decision tree with experience knowledge so as to built the fault diagnosis model, thus improving the history data sensitivity, providing better prediction effect when compared with a BP network, and improving the network convergence speed and training precision; the experience knowledge is supplemented, thus providing stronger robustness; the method can timely detect and diagnose, thus reducing fault incidence rate, and ensuring the photovoltaic power station to stably work.

The invention provides a business card identification method and device, and the method comprises the steps: obtaining a to-be-identified business card image; inputting the business card image into acharacter detection model to obtain each text line area, wherein the character detection model learns and obtain a corresponding relation between image features and each text line area; and inputtingeach text line area into the character recognition model to obtain business card information corresponding to each text line area. According to the method, the text line areas in the business card image can be identified through the text detection model based on deep learning, the robustness is high, the influence of low-quality and noise data on text extraction can be reduced, and therefore the universality and the application space of the method are improved. Moreover, the text line areas are subjected to end-to-end recognition based on the deep learning character recognition model, single word segmentation is not needed, higher accuracy is achieved, higher recognition capability is achieved for various complex changes, and the universality and the recognition effect of the method are improved.

The invention discloses a continuous identity authentication method based on touch screen slip behavior characteristics. The method comprises the following steps: analyzing touch screen slip operation behaviors generated when a user operates touch screen equipment; classifying touch screen slip operations into four operation modes according to touch screen slip directions, extracting behavior characteristics under each operation mode, establishing a user identity model under each operation mode based on the behavior characteristics, and performing continuous authentication on the identity of the user of the touch screen equipment by use of a window average method. According to the method, the touch screen slip behaviors do not need to be memorized or carried, behavior data collection can be finished in a daily touch screen equipment use process of the user without cooperation of the user, and non-invasive initiative identity authentication can be realized; in addition, a method of respectively performing modeling and window authentication on different types of touch screen operations is adopted, so that the stability of the authentication model can be ensured, the touch screen behavior characteristics of the user can be better embodied, and the robustness and fault tolerance of continuous identity authentication are obviously improved.

Disclosed is a deep long-term and short-term memory recurrent neural network acoustic model establishing method based on selective attention principles. According to the deep long-term and short-term memory recurrent neural network acoustic model establishing method based on the selective attention principles, attention gate units are added inside a deep long-term and short-term memory recurrent neural network acoustic model to represent instantaneous function change of auditory cortex neurons; the gate units are different in other gate units in that the other gate units are in one-to-one correspondence with time series, while the attention gate units represent short-term plasticity effects and accordingly have intervals in the time series; through the neural network acoustic model obtained by training mass voice data containing Cross-talk noise, robustness feature extraction of the Cross-talk noise and establishment of robust acoustic models can be achieved; the aim of improving the robustness of the acoustic models can be achieve by inhibiting influence of non-target flow on feature extraction. The deep long-term and short-term memory recurrent neural network acoustic model establishing method based on the selective attention principles can be widely applied to multiple voice recognition-related machine learning fields of speaker recognition, keyword recognition, man-machine interaction and the like.

The invention discloses a retinal vessel segmentation method based on combination of deep learning and a traditional method and relates to the fields of computer vision and mode recognition. According to the method, two grayscale images are both used as training samples of a network, corresponding data amplification, including elastic deformation, smooth filtering, etc., is done against the problem of less retinal image data, and wide applicability of the method is improved. According to the method, an FCN-HNED retinal vessel segmentation deep network is constructed, an autonomous learning process is realized to a great extent through the network, convolutional features of a whole image can be shared, feature redundancy can be reduced, the category of multiple pixels can be recovered from the abstract features, a CLAHE graph and a gauss matched filtering graph of the retinal vessel image are input into the network, an obtained vessel segmentation graph is subjected to weighted average, and therefore a better and more intact retinal vessel segmentation probability graph is obtained. Through the processing mode, the robustness and accuracy of vessel segmentation are improved to a great extent.

A fault analysis method includes: using a matrix of two sets of microgrid power distribution networks to analyze and solve a fault current, and for various types of faults of the distributed power distribution system, obtaining appropriate boundary conditions to calculate a variety of different types of single or simultaneous fault currents of load points. The present invention may be further applied to a situation where a bus or impedance or parallel loop is added. The present invention has good robustness and execution speed, and requires small memory space for calculation of analysis and identification of a power flow fault of the distributed power distribution system, and may be actually applied to an instrument control system for identification and analysis of a fault of a large-scalemicrogrid distribution system.

The invention belongs to the technical field of image processing, discloses a method for identifying and positioning power transmission line insulators in unmanned aerial vehicle aerial images, and solves the problems in the prior art that the detection precision of an identification algorithm of the insulators is not high, the robustness is low, and the identification algorithm is easy to be affected by sample number. A group of Gabor wavelet basis with different sizes and different directions and training sample images are taken as convolutions so as to form a group of characteristic vectors which accurately describe sample image texture characteristics. A random forest machine learning algorithm with a semi-supervised learning mode is used to train sample data sets of the known category and the unknown category so as to obtain an insulator identification model. Through the mode from left to right and from top to bottom, a detection window with the same size as the training sample traverses the input images with different sizes. The detection window combining the identification model detects and positions the positions of the insulators in the input images with different sizes. And finally the accurate positions of the insulators in the input image with the original size are determined by using a non-maximum inhibition method.

The invention provides a pedestrian re-identification method based on multi-scale feature cutting and fusion, particularly provides pedestrian re-identification network training based on multi-scale depth feature cutting and fusion and a pedestrian re-identification method based on the network, and performs pedestrian re-identification through multi-scale global descriptor extraction and local descriptor extraction. The extraction of the global descriptor is to carry out average pooling and feature fusion on feature maps of different layers of the deep network, and the extraction of the localdescriptor is to horizontally divide the feature map of the deepest layer of the deep network into a plurality of blocks and respectively extract the local descriptors corresponding to the feature maps. In the training process, a minimum smooth cross entropy cost function and a difficult sample sampling triple cost function are used as the target training network parameters. By adopting the technical scheme of the invention, the problem of feature mismatching caused by factors such as pedestrian posture change and camera color cast in pedestrian re-identification can be solved, and the influence caused by background can be eliminated, so that the robustness and precision of pedestrian re-identification are improved.

The invention discloses an ant colony optimization processing method of large-scale multi-target intelligent moving route selection; after data of NTSP target logistics delivery addresses, distances between every two addresses, and M price of cost for passing through each route are obtained, a route planning unit is solved by ant colony optimization technology so as to obtain a specific walking route for intelligent mobile-agent delivery, and the route is outputted to an executive mechanism for realization. When the method is used to solve the problem of large-scale multi-target intelligent moving route selection, the invention has good optimization performance, and has the advantages of parallelism, self-organization, strong robustness, and the like, and the obtained solutions are large in quantity, high in quality, and have strong approximation capability to the real Pareto solution set; the obtained solution set has uniform distribution; the calculation speed is high. The inventioncan be used in intelligent processing units of route planning systems in fields of logistics distribution, intelligent traffic, internet, robots, etc.

The invention relates to an uncalibration machine vision-based intelligent navigation control system and an uncalibration machine vision-based intelligent navigation control method, and belongs to the technical field of automation and detection. In order to overcome the disadvantage that whether a technical effect is good or bad depends on uncalibration parameters in a conventional scheme, in vision system-based mobile robot real-time obstacle avoidance and a navigation control method in the technical scheme of the invention, an image is automatically acquired and analyzed for the purpose of realizing the control of a mobile robot platform; the image is processed rapidly from image feedback information obtained directly by utilizing the principal of machine vision; and feedback information is given in a time as short as possible for participating in the generation of a control decision so as to form the position closed-loop control of an end effector of the mobile robot. The scheme improves the adaptability and the work efficiency of a robot, effectively maintains the speed and the precision in an image processing process, enhances the robustness and the stability of a robot control system, and reduces cost input and energy consumption in an implementation process of the technical scheme.

The present invention discloses an external parameter calibration method for a 3D camera group, which can solve the problems of the inconvenient operation, low precision and poor robustness in the existing camera calibration technology. The method comprises the following steps: S1 carrying out calibration of internal parameters for a 3D camera group; S2 carrying out calibration of the external reference for the camera pair; and S3 using the pedestrian detection to calibrate the external parameters of the camera group. The method disclosed by the present invention has a simple operation, high precision and good robustness, has obvious advantages compared with the prior art, and is suitable for market promotion.

The invention discloses a method and device for using a neural network to correct white balance of images, and belongs to the technical field of digital image processing. The method comprises the steps of 1, collecting an original image; 2, calculating parameter values of color temperature of an ambient light source; 3, preprocessing the original image and an image to be processed to obtain a preprocessed image and a second preprocessed image which are used for training respectively; 4, constructing a convolutional neural network model; 5, training the convolutional neural network model; 6, utilizing the convolutional neural network to calculate and obtain a red channel gain value (gainR) and a blue channel gain value (gainB); 7, utilizing the red channel gain value (gainR) and the blue channel gain value (gainB) to conduct white balance correction on the image to be processed, and obtaining a white balance image after correction. The method solves the technical problem that white balance easily loses efficiency in the prior art, the calculation speed of an algorithm is effectively increased, the accuracy is greatly improved, and the model is very good in robustness.

The invention discloses a distributed unmanned ship cluster division and formation control method, which comprises the following steps: S1, performing environmental perception on a target region by adopting a method of combining a Bayesian network and a body in an information fusion technology; S2, determining the number of required unmanned ships according to the characteristic of a task and a navigation environment, dividing unnamed clusters, and determining feasible formation patterns; S3, selecting an optimal task formation pattern or a secondary optimal task formation pattern from the feasible formation patterns according to an optimization index and a performance index of the task; S4, performing simulation and testing: performing modeling simulation according to the selected task formation pattern and the navigation environment, and testing the robustness and the fault tolerance of the task formation pattern, then ending. According to the method, a collaborative formation pattern of the unmanned ship clusters can be efficiently and flexibly changed, and the method has the advantages of safety, stability and reliability.

The invention provides a T-S fuzzy model-based flexible spacecraft multi-objective integrated control method. According to the method, a T-S fuzzy dynamic model of a flexible spacecraft is established, and the universal approximation of the T-S fuzzy model of the spacecraft is proved; the uncertainty of the inertia of the spacecraft which is caused by relative movement of flexible parts, and various kinds of space disturbance torques are considered; a control-performance LMI description and multi-objective integrated LMI method is adopted; and a robust H-infinity state feedback controller which can make a closed-loop system meet pole constraints and control input constraints is designed based on the T-S fuzzy model of the spacecraft. As indicated by numerical value simulation results, the designed state feedback control system has the advantages of short dynamic adjustment time, fast response, small overshoot and high steady-state accuracy, and can effectively inhibit the vibration of the flexible parts caused by attitude variation, and has high robustness and adaptability for the uncertainty of the model of the spacecraft.

The invention discloses a robust neural network control system for a micro-electro-mechanical system (MEMS) gyroscope based on sliding mode compensation and a control method of the control system. The control system comprises a given trajectory generation module, a sliding mode surface definition module, a neural network controller, a weight adaptive mechanism module, a sliding mode compensator, an MEMS gyroscope system, a proportional-differential control module, a first adder and a second adder. The control method of the control system comprises the following steps of: establishing an MEMS gyroscope kinetic model based on a sliding mode surface, designing a controller structure, and designing an updating algorithm of a radial basis function (RBF) network weight, so that the trajectory of the MEMS gyroscope is tacked. By the control method, the influence of the unknown dynamic characteristic of the MEMS gyroscope and noise interference can be compensated on line, the vibration trajectory of the MEMS gyroscope completely follows a reference trajectory, and the anti-interference robustness and reliability of the system are improved; the updating algorithm of the network weight is designed on the basis of a Lyapunov stability theory, so that the stability of a closed-loop system is ensured; and a powerful basis is provided for expanding the application range of the MEMS gyroscope.

The invention discloses a real-time tracking method of a multi-channel kernelized correlation filter. The method includes the following steps: a training step of conducting ridge regression on a previous frame of object information to acquire a filtering template; a detection step of detecting the current frame of image with the acquired filtering template and outputting a filtering response; an updating step of real-time updating the filtering template and the appearance of an object. According to the invention, the method uses kernel function to fuse multi-channel characteristics, overcomes the selection limit of the multi-channel characteristics, and transforms the problem of linear optimization of ridge regression to the problem of non-linear optimization of a higher space through the kernel function, such that the filtering template with excellent robustness is constructed, the speed of a tracking machine is greatly increased, and tracking requirements of real world can be met.

A system and method for testing robustness of a simulation model of a cyber-physical system includes computing a set of symbolic simulation traces for a simulation model for a continuous time system stored in memory, based on a discrete time simulation of given test inputs stored in memory. Simulation errors are accounted for due to at least one of numerical instabilities and numeric computations. The set of symbolic simulation traces are validated with respect to validation properties in the simulation model. Portions of the simulation model description are identified that are sources of the simulation errors.

The invention provides a wind energy converting system sliding mode control method and device based on a T-S fuzzy model. To solve the problem of the fault of an actuator in a wind energy converting system, the T-S fuzzy model is utilized for describing a nonlinear wind energy converting system with uncertain actuator fault information, the approximation accuracy of a controlled object is improved, and a good model foundation is established for sliding mode control. Meanwhile, by means of a sliding mode controller designed based on the linear matrix inequality technology, the stability of the wind energy converting system is guaranteed, and the robustness and fault tolerance of the wind energy converting system are improved. The precise tracking of the rotating speed of a power generator and the electromagnetic torque can be achieved when the uncertain actuator fault exists in the wind energy converting system, and the maximum wind energy capturing of the wind speed below the rated value is achieved, and a valuable reference scheme is provided for efficient and stable running of the wind converting system.

The invention belongs to the technical field of concrete structure damage detection, and discloses a multi-target crack recognition method based on a YOLOv3 deep learning algorithm, which comprises the following steps: importing a crack image into a YOLOv3 model, and automatically compressing the image into 416 * 416 pixel resolution; dividing the original image into S * S grids according to the scale size of the feature map by adopting an up-sampling and feature fusion mode similar to FPN; taking the cross-to-parallel ratio of the candidate box and the real box as an evaluation criterion, and; performing K-means clustering analysis on mark boxes for all crack target marking boxes of the image training set to obtain the size of a candidate box; and predicting the probability that the frame contains the target for each boundary frame through logistic regression. According to the method, the complexity of network training is simplified, and the operation cost is reduced; according to the method, the multiple targets are quickly and accurately identified, the accuracy far superior to that of other models is obtained while the target detection is quickly realized, and the method has higher robustness and generalization capability and is more suitable for an engineering application environment.

The invention belonging to the field of face recognition discloses a yolo-based face detection and face alignment method. The method comprises steps of network training and network verification. At the step of network training, a face data set is established; an image in the face data set is marked; and a face detection and alignment database is reconstructed. Therefore, problems that the efficiency is low and joint tasks can not be carried out because face detection and alignment are carried out by means of stage division during MTCNN face recognition are solved; the robustness of face recognition and the generalization ability of the network are improved; and a problem of over fitting caused by a few of samples is solved.