126results about How to "Strong descriptive ability" patented technology

The invention relates to a face feature extraction method based on a face feature point shape drive depth model. The method comprises the steps that the face feature point shape drive depth model is set up, N depth convolution neural networks are utilized for extracting features of N face regions divided according to the positions of face feature points to obtain the discrimination feature and the attributive feature of each region, and then all the discrimination features and the attributive features are fused to obtain features higher in descriptive ability. According to the face feature extraction method based on the face feature point shape drive depth model, the problem of robustness under change conditions of illumination, angles, expressions, shielding and the like can be well solved, and the recognition rate of face recognition under these conditions is increased.

The invention belongs to the technical field of remote sensing image target detection and identification, and in particular relates to a method for quickly detecting and identifying an airport target by using a remote sensing image based on a selective visual attention mechanism. By the method, the saliency of the original remote sensing image is analyzed by an improved attention selection model,namely graph-based visual saliency (GBVS), to acquire a saliency area, and the aim of identifying the airport target is fulfilled according to a scale-invariant feature transform (SIFT) characteristic on the area by combining a hierachical discriminant regression (HDR) tree. The method can effectively overcome the defect that the image is analyzed pixel by pixel in the conventional airport detection method. Compared with the conventional other airport detection methods, the method is characterized by high speed, high identification rate and low false alarm rate, has high noise robustness, andgreat significance and high value in practical application, and is very suitable for real-time detection under complex backgrounds in the fields of military use and civilian use.

The invention discloses a weather index and weighted LS-SVM-based power system short-term load prediction method. The method comprises the following steps of S1, obtaining a sample of original data; S2, according to the original data, calculating a comprehensive weather index; S3, performing data preprocessing on date type data and the comprehensive weather index; S4, according to an obtained dimensionless load characteristic quantity, performing gray correlation analysis between the dimensionless load characteristic quantity and a power system load, and calculating a characteristic quantity weight through a correlation degree obtained by the gray correlation analysis; and S5, building a weather index and weighted LS-SVM-based power system short-term load prediction model, performing parameter optimization by adopting a fruit fry optimization algorithm, and obtaining power system load prediction data of a to-be-predicted day through model output. The method has very good global optimization performance and few adjustment parameters, difficultly falls into local minimum and can effectively improve the power system short-term load prediction precision.

The present invention discloses a method for establishing a topological connected model of an urban underground electricity pipe network. The method comprises: step 1: collecting electricity pipe network spatial data, wherein spatial data of a topological relation model mainly originates from a map and document data; step 2: establishing an urban electricity pipe network topological relation model, and establishing an electricity pipe network base class and an element attribute information base class; step 3: establishing an electricity pipe network data model, and establishing a descriptive electricity pipe network, an electricity pipe network segment and an electricity pipe network node entity; and step 4: automatically generating an electricity pipe network entity and the electricity pipe network node entity by geometric topological relation checking, and classifying topological relation types to establish an electricity pipe network geometric topological model. According to the method for establishing the topological connected model of the urban underground electricity pipe network, the electricity pipe network data model is established after the electricity pipe network data are effectively analyzed by establishing the electricity pipe network geometric topological relation model; and a vector graph-based urban electricity pipe network topological relation model is formed by organic combination of the two models. Meanwhile, the method for establishing the topological relation can be generalized to other fields of applications of topological model-based GIS application systems.

The invention relates to a human motion emotion identification method based on Gauss feature, belongs to the field of computer mode identification, and solves the problems of low identification rate and low learning identification speed in the prior emotion identification method. The method comprises a training classifier step and an emotion identification step, wherein the training classifier step comprises a training data acquisition substep, a motion segmentation substep, a feature extraction substep and a training substep; and the emotion identification step comprises a data to be detected acquisition substep, a motion segmentation substep, a feature extraction substep and an identification substep. The method adopts the Gauss feature to describe human motion, has the advantages of strong descriptive power, low feature dimension, good Lie group structure, capability of effectively analyzing spatial structure, and the like, adopts a LogitBoost machine learning method based on a Lie group space to carry out multi-emotion identification, makes full use of the Lie group structure of the Gauss feature in a machine learning process, and has high training and identification efficiency and strong practicability.

The invention discloses a database generalization safety protection method based on abnormal and misapplication combination, which adopts a novel detection model of combining a rule-based misapplication detection technology and a machine learning-based abnormal detection technology and fuses an independent perfect database audit technology and a real-time monitoring technology. The method integrates two detection technologies, all normal accesses are judged to be illegal by a machine learning-based abnormal detection module, and abnormal accesses are judged to be illegal and not to be blocked by a rule-based misapplication detection module. The method realizes complete, strict and flexible protection of network layers of IP, protocols and ports to database application layers of access account, access base objects, access table objects, operation types, authority limits and the like, and can effectively resist attack aiming at attach of the database application layers and self vulnerability.

The invention discloses an equivalent simulation method for a grid-connected photovoltaic power generation system. In the grid-connected photovoltaic power generation system, a direct current is converted into an alternating current with the same amplitude value, the same frequency and the same phase as those of a grid voltage by a photovoltaic array through an inverter, and the alternating current is connected with a power grid, so that the photovoltaic array which is operated normally is equivalent to a constant-current source. The mathematical model of a grid-connected power generation system is represented by the following third-order dynamic differential equations which are shown in the specification, and the active response P and the reaction response Q of the grid-connected photovoltaic power generation system are shown in the specification. By the equivalent simulation method of the grid-connected photovoltaic power generation system, the steady state and transient characteristics of the photovoltaic power generation system under a grid-connected operation condition can be simulated accurately; and the equivalent simulation method can serve as a research tool of the grid-connected photovoltaic power generation system.

The invention discloses a phase encoding characteristic and multi-metric learning based vague facial image verification method. The phase encoding characteristic and multi-metric learning based vague facial image verification method comprises (1) a training phase, namely, partitioning sampling images and extracting multi-scale primary characteristics of every image block, performing fisher kernel dictionary learning through the above characteristics to generate into partitioning fisher kernel coding characteristics, performing multi-metric matrix learning on the above coding characteristics to generate a plurality of metric matrixes and obtain the metric distance after training samples are performed on multi-metric matrix projection, calculating the average metric distance and variance of positive samples and negative samples to a set and confirming a final classification threshold through a probability calculation formula of Gaussian distribution and (2) a verification phase, namely, partitioning input facial images and extracting multi-scale primary characteristics, generating partitioning fisher kernel coding characteristics, obtaining the final metric distance through the multi-metric matrix and comparing the distance and the threshold to obtain a facial image verification result. The phase encoding characteristic and multi-metric learning based vague facial image verification method has the advantages that the identification rate is high and the universality is strong.

The invention discloses a removal method and system for a bone tissue in a 3D CT (Three Dimensional Computed Tomography) image. The method comprises the following steps: reading a 3D CT image data sequence and dividing the sequence by using a clustering method; describing each marked supervoxel from supervoxel level, pixel level and supervoxel neighborhood relation level respectively to obtain a characteristic vector for describing each supervoxel; constructing a supervoxel training set, training a classifier by using the characteristic vector corresponding to the supervoxel, then classifying the supervoxel by using the trained classifier and marking the category; if a certain supervoxel is marked as the bone tissue according to a supervoxel mark value, removing the supervoxel from the original CT data, and otherwise, remaining. The supervoxel obtained by dividing the 3D CT data is used as a basic operation unit, bone tissue templates on different parts are not required to be learnt and trained, and meanwhile the problem of inaccurate bone tissue extraction caused by mismatching of the templates is also avoided.

The present invention discloses an interface defining method, and a data transmission method and apparatus, so as to solve the problem in the prior art that a defined interface is relatively poor in accuracy and scalability. The method comprises: an interface tool defining an interface by using a preset interface description language; generating an interface file according to the defined interface; and providing the interface file to a server, so that the server converts data in an original format into data in a target format according to a correspondence, between each field in the original format and each field in the target format, comprised in the interface file. According to the method and apparatus in the present application, the interface is defined by using the preset interface description language, and the interface description language can well describe the fields, so that each field in the target format can be accurately described, thereby effectively improving accuracy of interface defining. In addition to describing an HTTP based interface, the interface description language can also describe an interface based on other protocols, thereby effectively improving scalability of interface defining.

The embodiment of the invention provides a tracking target determination method and device. The method comprises the steps that a current frame image of a to-be-determined tracking target is obtained,and alternative targets in the current frame image are determined; all the alternative targets are sent to a preset deep learning network, wherein the deep learning network is obtained by training sample images in advance and is used for extracting appearance features containing position information of pixels; first appearance features, extracted through the deep learning network, of all the alternative targets are obtained; second appearance features of first tracking targets corresponding to a previous frame image are obtained; according to the first appearance features and the second appearance features, each alternative target is matched with each first tracking target; and according to the alternative targets successful in matching, the tracking target in the current frame image is determined. Through the scheme in the embodiment, the accuracy of the determined tracking target can be improved.

A channel logo segmentation method based on a fully convolutional channel logo segmentation network includes the steps of constructing a multi-type fine-grained channel logo data set, the channel logoimage set containing a total number of 8400 images covering 42 categories; according to a channel logo region extraction method of pixel-by-pixel labeling, establishing a binary label image set corresponding to the channel logo image set and converting the binary label image set into an L-type single-channel gray image set; establishing a fully convolutional channel logo segmentation network of an end-to-end encoder-decoder structure, training the fully convolution label segmentation network on the channel logo data set, inputting a test image of any size in the channel logo image set to thetrained fully convolutional channel logo segmentation network, and generating a pixel-level segment result of the same size as the input image. The method of the invention helps the deep convolutionalneural network to realize powerful performance and solves the problem that the tiny target segmentation results are not precise enough. The network model can delineate a tiny target, thereby improving the spatial precision of the output, and is suitable for being applied to the tiny target segmentation method.

The invention discloses a robust human face image principal component feature extraction method and identification apparatus. The method comprises: by considering low-rank and sparse characteristics of training sample data of a human face image at the same time, directly performing low-rank and L1-norm minimization on a principal component feature embedded through projection, performing encoding to obtain robust projection P with good descriptiveness, directly extracting a low-rank and sparse principal component union feature of the human face image, and finishing image error correction processing; and by utilizing the embedded principal component feature of a training sample of a robust projection model, obtaining a linear multi-class classifier W* for classifying human face test images through an additional classification error minimization problem. When test samples are processed, a union feature of the test samples is extracted by utilizing a linear matrix P and then the test samples are classified by utilizing the classifier W*; and by introducing a thought of low-rank recovery and sparse description, the principal component feature, with better descriptiveness, of the human face image can be obtained by encoding, the noise can be eliminated, and the effect of human face identification is effectively improved.

The invention discloses an intelligent rotary machine fault depth network feature identification method. A vibration sensor is arranged at a to-be-detected rotating mechanical part of a train rollingbearing; collecting an original vibration sequence when the rolling bearing works; decomposing and reconstructing the original vibration sequence through a singular spectrum analysis method; extracting a root-mean-square value of the reconstructed vibration sequence; standard deviation, skewness index and peak value; a fault position is judged by using a rotary machine fault position diagnosis model obtained by training of a support vector machine; and then, carrying out ensemble empirical mode decomposition on the reconstructed vibration sequence, calculating the permutation entropy values ofa group of decomposed intrinsic mode components, taking the permutation combination of the permutation entropy values as a detection characteristic, and judging the fault type by using a rotary machine fault type diagnosis model obtained by training of a support vector machine. The fault position and the fault type of the rotary machine can be detected more timely, and the fault diagnosis accuracy and reliability are improved.

The invention discloses a preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification. The method includes the steps: manually selecting images shot by a vehicular camera, calculating layered HOG symmetric characteristics of an image library in a normalized manner, and performing training by the aid of an ELM (extreme learning machine) to obtain a classifier; performing shadow processing for the video images acquired by the vehicular camera to obtain hypothetical car sub-graphs; judging the obtained hypothetical sub-graphs by the classifier to obtain detection results. The method can accurately detect preceding cars in real time and has strong robustness for illumination intensity change.

The invention discloses a vehicle type fine identification method and system. The method comprises the following steps: carrying out graying and standardizing processing on an obtained original vehicle image to obtain a standardized image; calculating gradient and direction of each pixel point of the standardized image; carrying out direction gradient histogram feature extraction and local linear constraint coding on the standardized image according to the calculated gradient and direction to obtain encoding vector of the standardized image; processing the standardized image obtained after local linear constraint coding through a weight space pyramid according to the obtained encoding vector to obtain final expression vectors of the vehicle image, wherein the final expression vectors of the vehicle image comprise position information and semantic information of the vehicle image; and inputting the final expression vectors of the vehicle image to a pre-trained linear support vector machine classifier for vehicle type identification. The vehicle type fine identification method and system have the advantages of high accuracy, low complexity, high robustness and rich detail features, and can be widely applied to the field of picture processing.

The invention discloses a classification model construction method used for macula degeneration region segmentation. The method includes the following steps: selecting multiple fundus images, conducting graying processing on the fundus images to obtain multiple gray scale images, and sampling foregrounds and backgrounds of the gray scale images to obtain samples; adopting a generalized low-rank approximate method to obtain a transformation matrix, conducting dimension reduction on the samples on the basis of the transformation matrix, and obtaining a low-rank approximate matrix of the samples; adding label information into the low-rank approximate matrix of the samples to perform a supervision function, and constructing manifold regularization items; establishing a target function through the generalized low-rank approximate method and the manifold regularization items, solving the target function through an iterative optimization method, and obtaining an optimal transformation matrix and an optimal low-rank approximate matrix of the samples; and constructing a classification model on the basis of the optimal low-rank approximate matrix and the label information. The classification model can extract low dimensional and also highly distinguishable feature descriptors, and can improve the segmentation precision.

The integrated optimization method is for controlling production planning, dispatching and emulation on mixing batch assembly line (i.e. assembling multiple products on a assembly line at same time). The method includes procedures: smoothing customers' requirements on products; based on smoothed requirements on products, mixing integrated planning model of mixing batch assembly line is solved by using a branch delimit algorithm; obtaining best production planning, i.e. categories and quantities of products as well as dispatch or assembling sequence; observing best dispatching procedure for mixing batch assembly line by animated emulating the dispatch; determining whether customers' requirements on products are satisfied; if yes, formal plans and dispatch is ready to be excuted; based on formal plans and dispatch, arranging and controlling parts to be assembled on assembly line; based on moving speed determined by using animated emulation, controlling operation of assembly line.

A salient feature enhancement sampling method based on octree-like indexes is proposed. Firstly, based on the octree-like spatial segmentation, the adaptive point cloud feature enhancement is appliedto the model with serious local missing. Then, the subspace-based nearest neighbor sampling algorithm is used to downsample the enhanced point cloud. Finally, the skeleton is extracted from the enhanced point cloud according to the obtained sample points. The experimental results show that the octree-like space can realize the efficient management of the point cloud and improve the accuracy and timeliness of the skeleton. The adaptive point cloud enhanced sampling strategy makes the extracted skeleton more repeatable and descriptive. The optimized algorithm in the invention is applicable to point cloud models in various domains, and has good adaptability and robustness.

The invention discloses a multi-stage vehicle-mounted laser point cloud road change monitoring method, which comprises the following steps of: firstly, analyzing the characteristics of a vehicle-mounted laser road point cloud, preprocessing road scene point cloud data into information and irrelevant information related to point cloud registration, further separating out a road surface point cloudby taking kerbs special for a road scene as boundary assistance, establishing an octree index based on a kerb boundary and ground object point cloud data, and extracting feature points at equal intervals by adopting a normal vector extraction feature region and an LSP (Local Surface Patches) feature detection method; and finally, using a 4PCS (4-Points Congruent Sets) registration algorithm basedon the feature points for completing coarse registration, and using an ICP (Iterative Closest Point) algorithm of a bidirectional KD tree index for completing fine registration. Change monitoring is performed on a road by taking the precision requirement of change monitoring as a constraint condition, so that the registration precision is higher and the registration speed is higher.

The invention discloses a multi-state multi-task system reliability parameter analyzing method. The method comprises the steps that AS trees are utilized to describe logic structure relation of each stage of a multi-stage task system, and a multi-state multi-stage task system simulation algorithm is given based on a discrete event simulation thought. The simulation algorithm can calculate multiple parameters such as multi-state PMS system reliability, required input parameters are few, programming realization is simple, the simulation algorithm can well describe complicated behaviors of the multi-stage task system, multiple reliability parameters for knowing a system reliability parameter can be obtained, and the simulation algorithm has very strong universality and practicability.

The invention discloses a method for achieving a distributed system behavior simulated analysis tool based on a behavior tree. The method comprises the steps that a graphic symbol of a pel corresponding to a typical behavior tree factor is established, a simulation function on a behavior tree model expressed by an XML document is designed and achieved, a traversal algorithm of the behavior tree model based on XML is designed and achieved, an input interface through which a user defines whether a behavior is reachably verified or not within time cost is given, a variable for recording traversal time cost is designed, an external java programmatic interface defined by the behavior in the model is achieved, and therefore a simulation tool can call a function defined by the user. By means of the method, the graphic simulated analysis tool based on the behavior tree can be achieved, the tool achieved through the method can achieve system modeling, a method is provided for interactive logic reasonability, timeliness and reliability analysis of a distributed system, and the advantages of being simple in model, high in description capacity, accurate and comprehensive are achieved.

The present invention discloses a DCE and LSS-based infrared image and visible image registration method. The method comprises the steps of S1, through the DCE, respectively describing the target contour of a to-be-registered infrared image and the target contour of a visible image, extracting each DCE feature point in the vertex of a corresponding target contour, and obtaining DCE feature point sets respectively corresponding to the to-be-registered infrared image and the visible image; S2, based on the LSS method, respectively describing each DCE feature point in each DCE feature point set; S3, according to the description result of step S2, calculating the matching degree among each DCE feature point in the to-be-registered infrared image and the visible image so as to obtain the corresponding relationship between each DCE feature point; S4, according to the corresponding relationship obtained in the step S3, conducting the image transform to obtain a registered image and outputting the registered image. According to the technical scheme of the invention, the advantages of low registration implementation complexity, good registration accuracy, high efficiency and the like are realized.

A method and a system for controlling service data access are disclosed, the method first establish a service database for storing various necessary services of enterprise and a resource database for storing various resources of enterprise, wherein the each sort of necessary services in the service database comprises relevant service positions, in the meantime, each service position comprises permitted service operations; each sort of resources in the resource database comprises the relevant elements, then an associating module associates all the service operations in the service database with each sort of resources and elements in the resource database; when staff need to operate the resources and elements of the enterprise, an inquiring module inquires the service database according to the service position of the staff to determine the permitted service operations for the service position, and an association determining module determines the resources and elements associated with the inquired service operations, in the end, an authorizing module authorizes the resources and elements to the relevant staff, achieving authorization management for service data access.

The invention discloses a monocular camera and millimeter wave radar external parameter calibration result accuracy evaluation method. A plurality of metal balls are arranged in a scene, after data information of the metal balls in all coordinate systems is converted, statistics and comparison calculation are completed, and four indexes describing the accuracy of an external parameter result are obtained, and two indexes are in a camera image plane and two indexes are in a radar point cloud space.

A channel logo segmentation method for a pixel-level channel logo recognition network based on cross-layer feature extraction includes the steps of modifying a fully connected layer of an existing classification network to a convolutional layer, combining features of input images output by low and high layers, extracting cross-layer features that integrate local and global features of the images,and constructing three pixel-level channel log recognition networks of different cross-layer architectures; using existing channel logo data sets as training and test data, including a channel logo image set and a binary label image set corresponding to the channel logo image set, and extracting three cross-layer features of the channel logo image set respectively by using the three pixel-level channel logo recognition networks; training the pixel-level channel logo recognition networks; and testing the channel logo image set by using the three pixel-level channel logo recognition networks trained on the channel logo data set of different types respectively, and finally producing pixel-level segmentation results of the same size as the input images. The channel logo segmentation method ofthe invention has a stronger description capability and discrimination capability.

The invention discloses a distributed principal component analysis neural network modeling method for a chemical exothermic reaction. According to the method, input and output data of an object is collected, a principal component analysis method is utilized for dividing a distributed parameter system into an autoregressive linear model and a non-linear model by reducing dimensionality, and the autoregressive linear model is identified through a least square method. For the non-linear model, the least square method is utilized for establishing an RBF neural network model of the object, and then parameters of the RBF neural network model are optimized through a genetic algorithm. The established models have high precision, and dynamic properties of the process object can be well described.

The invention discloses a picture description method of guiding an attention mode on the basis of an attribute probability vector. The method comprises the following steps: inputting an image to obtain a feature graph through a fully convolutional neural network, and then obtaining the attribute probability vector through a multi-instance learning algorithm layer; selecting certain threshold values for the obtained attribute probability vector to initialize hidden states of c0 and h0 of a long short-term memory (LSTM) unit; guiding the attention mode through the attribute probability vector, and combining a state of ht-1 of a description statement LSTM of a previous moment to generate an encoding vector, which currently needs to be attended, at a region which is on the feature graph and attended by the attention mode of a current moment; outputting an output state of ht of the current moment by the description statement LSTM according to the current encoding vector; and becoming the state of the previous moment by the output state of the current moment, and repeating the previous operations until generation of a description language is completed. Compared with other methods, the picture description method of guiding the attention mode on the basis of the attribute probability vector of the invention obviously improves effects, is better in comprehensive performance of evaluation indexes, and can basically be competent for general picture description needs.

The invention discloses a method for establishing an equivalent model of external characteristics of a wind power plant of a DFIG, A DFIG overall equivalent model response is shown in the specification, wherein P<total_res>, Q<total_res>, P<s_res>, Q<s_res> and P<grid_res represent a total active response of the DFIG, a stator reactive response of the DFIG and a grid side active response of the respectively, Pc and Qc represent a controlling equivalent model active response and a controlling equivalent model reactive response respectively, and a grid side reactive response is ignored here, and subscripts '-res' represent that responses are equivalent model responses, so that responses are distinguished from simulation model responses. The method for establishing the equivalent model of external characteristics of the wind power plant of the DFIG is high in fitting precision and fast in simulation speed.