239results about How to "Find out exactly" patented technology

The invention discloses a power equipment infrared image fault positioning, identification and prediction method. The power equipment infrared image fault positioning, identification and prediction method comprises the following steps: 1) collecting power equipment infrared thermal image data; 2) classifying the infrared images to form a data set; 3) constructing a convolutional neural network model; 4) separating out faulty power equipment; 5) monitoring faulty power equipment in real time, and longitudinally collecting temperature data; 6) positioning a fault part, segmenting the infrared image of the power equipment, and extracting a fault area; 7) diagnosing a fault area, and judging a fault level; 8) predicting an equipment state trend; 9) uniformly outputting and displaying the information; 10) storing the fault level; 11) making four types of infrared image data sets; 12) building a target detection model and training; 13) directly detecting an infrared image of power equipmentto be detected through a target to obtain a fault position and a fault level; 14) repeating the step (5); 15) repeating the step (8); and 16) repeating the step (9) facilitating positioning of the fault position, fault level judgment and prediction of the fault equipment and giving a maintenance suggestion.

A detecting and error-correcting method capable of rapidly and accurately obtaining an element center and a deflection angle is disclosed. Interested edge area can be rapidly and accurately found out through an automatic edge area extraction method based on combination of binary morphology and image reduction, and subsequent edge extraction image scope is minimized as possible. Pixel level edge is roughly positioned at first, and accordingly a sub-pixel edge is rapidly extracted by one-dimensional curve fitting method. The method overcomes the defect of complex algorithm of a conventional sub-pixel edge extraction algorithm based on two-dimensional images, and the time of extracting accurate sub-pixel edge can be shortened. Weighted least square rectangular edge fitting algorithm based on linear hazen paradigm can detect straight lines on any positions of an image, and effectively minimizes influence of outlier on fitting precision due to uneven edges. The central positions of a plurality of elements and deflect angles can be rapidly and accurately detected at once, and the efficiency of visual detection is improved.

The invention discloses a multi-feature cyclic convolution significance target detection method based on an attention mechanism. The method comprises the following steps: ; the method comprises the following steps of: 1, analyzing common characteristics of a salient target in a natural image, including spatial distribution and contrast characteristics, using an improved U-Net full convolutional neural network, performing pixel-by-pixel prediction by adopting an encoder-decoder structure, and performing multi-level and multi-scale characteristic fusion between an encoder and a decoder by adopting a cross-layer connection mode; secondly, a large number of clutters can be introduced to interfere with the generation of a final prediction graph by carrying out concentage fusion on coding end features and decoding end features, so that an attention module is introduced to calibrate full-pixel weights from two angles between channels and between pixels, the task-related pixel weights are enhanced, and the background and noise influence is weakened; and 3, a multi-feature cyclic convolution module is used as a post-processing means, the spatial resolution capability is enhanced through iteration, the edge of an image region is further refined and segmented, and a finer significant target mask is obtained.

The invention discloses a multi-channel network-based video human face detection and identification method. The method comprises the following steps of S1, performing video preprocessing: adding time information to each frame image; S2, detecting a target human face and calculating a pose coefficient; S3, correcting a human face pose: for m human faces obtained in the step S2, performing pose adjustment; S4, extracting human face features based on a deep neural network; and S5, comparing the human face features: for an input human face, obtaining eigenvectors by utilizing the step S4, matching a matching degree of an eigenvector of the input human face and a vector in a feature library by utilizing a cosine distance, and adding a class to alternative classes, and if the cosine distances between a feature of the to-be-identified human face and central features of all classes are all smaller than a set threshold phi, regarding that a database does not store information of a person, and ending the identification, wherein the cosine distance between the class and the to-be-identified human face is greater than the set threshold phi. The multi-channel network-based video human face detection and identification method with relatively high accuracy is provided.

The invention relates to a method and a device for identifying association relation between network users. The method comprises the following steps of determining a user pair according to the identification attributive characters of users; judging whether the user behavior occurs between the users of the user pair for a data object with the specific attributive character; and when the user behavior occurs between the users of the user pair for the data object with the specific attributive character, utilizing parameters of the user behavior to compute a weight of the association between the users of the user pair in order to identify whether the users of the user pair have the association. According to the technical scheme of the invention, the association between the users can be identified from all users quickly and accurately.

The invention provides an image pushing method, an image pushing device and a mobile terminal. The method comprises the following steps: acquiring a user image at least comprising a user face; identifying face information of a character in the user image, wherein the face information of the character comprises facial feature information and position information of the character face in the image; comparing the character face information in the user image with a preset database; and pushing an image with the highest matching degree with the character face information of the user image in the preset database according to the comparison results between the character face information in the user image and the preset database. According to the technical scheme, the problem in the prior art that the matching result is inaccurate by matching a single comparison factor with a photo similar to the user through the facial features is solved.

The invention relates to a circuit for detecting the dead pixel of an LED display screen and a method thereof. The circuit comprises an LED display unit, a control circuit and a drive circuit, wherein the drive circuit comprises a line drive circuit and a column drive circuit; the LED display unit consists of M*N LED(s), wherein M is more than or equal to 1, N is more than or equal to 1, and M and N are integers; the drive circuit is internally provided with a sampling resistor and an A/D converter; and the two ends of the sampling resistor are connected with the input end of the A/D converter. The working principle of the circuit comprises the steps of detecting the voltage of the circuit by the A/D converter through the sampling resistor which is arranged in the drive circuit of the LED display screen to compute the electric current which pass through each LED, and comparing the computed electric current with the normal work electric current of the LED to judge that the LED is the dead pixel when some LED electric current value is deviated from the normal work electric current. Compared with the prior art, the invention is simple and reliable in circuit, can fast and exactly find out the dead pixel, is suitable for not only the production process of the LED display screen but also the remote on-line detection of the installed and used LED display screen, and greatly improves the detection efficiency.

The invention provides a distributed implementation method for analyzing user WiFi activity track rules. Residential areas of MAC can be obtained according to the activity rule track of the MAC, the time of the activity rule track is analyzed to predict which areas a target object may occur within which time period, and an MAC set with similar activity track rules as the MAC can also be inquired. The mined tracks can be semantized to find some commuter modes of users, for example commuting routes; and mined interested areas can be semantized to working areas and residence areas of the users.

The invention discloses a landmark building identification and detection method based on deep learning. The method comprises the following steps of inputting a to-be-identified landmark building imageinto a DenseNet network to obtain a feature block diagram containing the target building feature information, and then sending the feature block diagram into a region suggestion network to predict abinary category of the feature block diagram and the coordinates of a target building in an original image; completely mapping a prediction candidate box to the feature block diagram by using a RoI Agign method; finally, carrying out classification and frame regression on the more accurate feature block diagrams to obtain the prediction probabilities and the coordinate positions of different landmark buildings, removing the redundant candidate frames through a non-maximum suppression method, fusing the diagrams with the wider coverage regions, and finally realizing the identification and detection of the landmark buildings. According to the method, the prediction of the landmark building candidate frames is more accurate, the prediction range is larger, and the method has the better identification capability on the landmark building images in the complex environment.

The invention discloses a positioning and partition method for a human tissue cell two-photon microscopic image, and belongs to the technical field of image processing. The positioning and partition method for the human tissue cell two-photon microscopic image mainly solves the problem that too many errors occur when the human tissue cell two-photon microscopic images are partitioned by means of the prior art. The positioning and partition method comprises the steps of preprocessing, namely converting an image to be partitioned into a grey-scale image, clustering the preprocessed images and obtaining an edge graph of the image, conducting centralized positioning with the edge graph, positioning the centers of cell nucleuses accurately, obtaining a point set of the cell nucleus centers, and finally obtaining the edges of the cell nucleuses accurately combining an active contour model. Compared with the prior art, the positioning and partition method for the human tissue cell two-photon microscopic image has the advantages of being accurate in edge extraction, high in positioning efficiency, short in evolution time and the like. Further, the disturbance of granular noises and uniform brightness distribution in the image is avoided, and the positioning and partition method can be used for extracting the edges of the cell nucleuses in the cell two-photon microscopic image.

The invention provides a method and a device for identifying multiple touch points. The method comprises the following steps of: acquiring a frame image by two cameras respectively, and acquiring shape, position and transverse size information of touch object images in the two images; judging whether the types of the touch objects are the same according to the shape information; if the types are different, identifying pen touch and finger touch, and calculating touch point coordinates of a pen and a finger respectively; if the types are the same, calculating the angle between each touch object and the connecting line of the two cameras and crossing point coordinates of all connecting lines according to the position information; calculating the approximate distance between the touch objects and the cameras and the approximate position coordinates according to the transverse size information; and verifying the crossing point coordinates and the approximate position coordinates to find out a real touch point. By using the method and the device, the touch objects of different types can be accurately identified, and the problems of cost increment and design difficulty increment caused by increasing the auxiliary camera are effectively avoided when multiple touch points of the same type are positioned.

The invention provides an NPC three-level inverter open-circuit fault diagnosis method based on an instantaneous frequency, and relates to the field of power electric appliance fault diagnosis. The method comprises the following steps: acquiring a three-phase output current on a grid side of an inverter, estimating the instantaneous frequency of the three-phase output current by using a weighted sliding Hilbert transform method, constructing an instantaneous frequency residual value in combination with a theoretical instantaneous frequency, and solving a mean value of the instantaneous frequency residual value; and obtaining a normalized current period mean value according to the three-phase output current of the grid side of the inverter, and comparing the mean value of the instantaneousfrequency residual value and the current period mean value with a set fault threshold value to judge a fault type and a kind. According to the technical scheme of the invention, a ''flying wing effect'' between data segments can be well overcome, and open-circuit fault diagnosis of a single power switching tube based on the three-phase output current can be realized.

The invention discloses an equipment testing method and device applied to a thermal power plant. The method comprises the steps of controlling an unmanned aerial vehicle to fly according to a preset testing route; collecting terahertz images or infrared images of inner and outer walls of thermal power plant equipment in the flight process of the unmanned aerial vehicle, wherein the thermal power plant equipment comprises a boiler, a chimney, a desulfurizing absorption tower and an air cooling island; and analyzing the terahertz images or the infrared images and determining the position of the defect of the equipment and the heat dissipation condition. The defects of the inner and outer walls of the thermal power plant equipment are detected by using the unmanned aerial vehicle instead of manpower, so that the human cost can be reduced, the time can be saved, and the testing efficiency is improved.

The invention belongs to the technical field of digital signal processing and discloses a voice underdetermined blind identification method and device based on frequency spectrum correction and data density clustering to improve algorithm precision, improve algorithm efficiency remarkably and improve the robustness of the algorithm on noise. According to the technical scheme, the method comprises the steps of 1, conducting hanning windowed L point 50% overlapped short time Fourier transform (STFT) to obtain an observation frequency spectrum Xm(t,k); 2, conducting frequency spectrum correction on the STFT observation frequency spectrum frame by frame; 3, for a specific time frame t0, conducting mode purification on all harmonic wave parameters; 4, executing the step 2 and the step 3 frame by frame, and collecting SAS modes of the all the time frames to form a monosource domain omega={zi, i=1,...,P}, wherein P is the number of modes of the monosource domain; 5, conducting data density clustering on the SAS modes in the monosource domain. The method and device are mainly applied to digital signal processing.

This center of gravity position detection device calculates the center of gravity position of a suspended load suspended by a rope, and is provided with a measured value acquisition unit which, for at least two different states, acquires measured values from data indicating the state in which the suspended load is suspended, and a center of gravity position calculation unit which calculates the center of gravity position of the suspended load on the basis of a state model of the suspended load which includes at least the center of gravity position of the suspended load as an unknown variable and the measured values acquired by the measured value acquisition unit.

The invention discloses a large scale three dimension (3D) wireless sensor network node location method based on convex partition. A partition method involved in the large scale 3D wireless sensor network node location method based on the convex partition includes: confirming boundary nodes of a 3D wireless sensor network area and obtaining a boundary node set; confirming a concave node set and cluster head concave nodes in the boundary nodes; dividing the 3D wireless sensor network area into a plurality of subsidiary area sets appropriate in concave and convex rate and area number according to the cluster head concave nodes. The large scale 3D wireless sensor network node location method based on the convex partition includes: using an improved location method and a DV-Hop distance measurement method on subsidiary areas after partition to confirm relative distance information of the nodes and position information of anchor nodes; unifying relative position information of the subsidiary areas so as to obtain absolute position information of the nodes in a whole large scale area. The large scale 3D wireless sensor network node location method based on the convex partition is suitable for node location in the large scale 3D wireless sensor network area, and is high in location accuracy and high in speed.

The embodiment of the invention discloses an intelligent question question-answering method and a related device, which are used for improving the accuracy of an intelligent question and answer system. The method comprises the steps that a target user problem is acquired; a target generation problem corresponding to the target user problem is generated according to a problem optimization model, the problem optimization model is obtained based on generative adversarial network training, and the problem optimization model comprises a generator and a discriminator; whether the generation qualityof the target generation problem is higher than a first preset threshold value or not is judged according to the discriminator, wherein the generation quality is used for indicating the probability that the target generation problem is a standard problem; and a target answer is determined according to the target generation question if so.

The invention discloses a method for panoramic inspection on a substation. The method comprises the following steps: selecting image acquisition points according to specific conditions of the substation; acquiring image information by an inspector holding inspection equipment; performing fault judgement on a panoramic image by the inspection equipment according to preset fault criteria during the process of inspection performed by the inspector holding the inspection equipment; and finally, performing key survey and maintenance aiming at a fault point. The inspection equipment comprises an inspection device, a location device and a signal processor. The method can be used for acquiring panoramic information of the substation once via image splicing technology, finding the fault point in the substation quickly and accurately, reducing omission in inspection, and improving the work efficiency of the inspector to a great extent.

The invention discloses a five-shaft cradle type numerical control machine tool non-deformation cutting three-dimensional geometrical modeling method, for solving the technical problem of poor modeling accuracy by use of a conventional method. The technical scheme is that first of all, a movement amount of each movement shaft of a five-shaft machine tool in a processing process at any instantaneous time by use of a linear interpolation method; then, performing modeling by use of a two-dimensional contour in which a group of parallel planes are intersected with a cutter and a workpiece, and through combination with a kinematic chain of the machine tool, reducing a space three-dimensional problem to a two-dimensional plane problem; establishing a mathematical expression of an instantaneous cutting edge of the cutter on a selection layer of the workpiece in a work coordinate system; solving an envelope boundary curve and an internal envelope area of the instantaneous cutting edge on each layer of a part at any time; solving a non-deformation cutting shape on the selection layer at the moment by subtracting an envelope area at the moment by use of an envelope area at last moment; and obtaining a three-dimensional non-deformation cutting geometry of the period by stacking all non-deformation cutting shapes generated on all layers of the workpiece. The modeling precision is high.

The invention discloses a pedestrian recognition method based on Gaussian and depth features. The method includes 1, collecting a training set and a test set; 2, quickly extracting pedestrian features; 3, extracting semantic features of pedestrian feature; 4, training and extracting feature with that model; 5, testing that performance of the JDAG model; step 6, reordering the final result queue. Beneficial effect: Combining the semantic features extracted by convolution neural network with the unique features (color, texture) of pedestrians, the distinguishability of features is improved. Next, the Re-ranking method rearranges the previous similarity scores so that a picture is given and the pedestrians in the picture set with the same ID as the pedestrians in the picture can be found moreaccurately.

The invention discloses a method for detecting network performance problems and positioning failure nodes. The method includes the steps: establishing an IP (internet protocol) metropolitan area network topology structure comprising hierarchical relation among network nodes and corresponding relation of each network node and an IP address pool of a terminal that the network node belongs to; capturing IP data packets of all terminal access service platforms through a service platform outlet of an acquisition device on an IP metropolitan area network core layer; tracking TCP (transmission control protocol) data flow of each terminal, making statistics on TCP packet sum A and TCP duplicate packet number B of each terminal, computing TCP duplicate packet rate Nr=B/A, and then recording corresponding relation of an IP address of each terminal and the Nr; computing a current value Pc and a historical reference value Pb of a ratio of failure terminals that each network node belongs to; according to the metropolitan area network topology structure, traversing from top to bottom to detect whether or not the current failure problem terminal ratio Pc of each network node is out of limit, and if yes, prompting that the network node or the downstream node thereof has performance failure; finishing if network topology traversing is finished. By the method, the failure nodes with the ratio of the failure terminals out of limit can be found rapidly and accurately.

The invention relates to a sensitive information analysis system and a sensitive information analysis method. The sensitive information analysis system comprises a core processing module, a core database, a user interface and a sensitive person interface. The sensitive information analysis method comprises the following steps that: a system engineer installs a transparent gateway, a search engine and the sensitive information analysis system on a server; a user enters the sensitive information analysis system, adds the sensitive information into the system and sets automatic operation; when the time is up, the system automatically executes the operation, calculates the sensitivity of each data block according to the sensitivity of the sensitive information, displays the sensitive blocks with higher sensitivity to the user to judge whether the sensitive blocks are true sensitive blocks or not, further analyzes the true sensitive blocks, extracts words which may be new sensitive words, extracts data sources which may be new sensitive sources and then displays the words and the data sources to the user; the user judges whether the words and the data sources are new sensitive words and new sensitive sources; and all the steps are repeated. The sensitive information analysis system and the sensitive information analysis method have the advantage of finding out the sensitive information and the sensitive persons quickly and accurately.

The invention discloses a method for quickly segmenting an effective region of a palmprint on line based on direction detection, and mainly overcomes the deficiencies of high time complexity and poor real-time property of an existing online palmprint segmentation technology. The method comprises the following steps: 1, inputting a palmprint image and calculating the palm direction by gradient of a profile; 2, correcting the palmprint image and selecting alternative points by the gradient direction of the corrected profile; 3, performing hierarchical clustering on the alternative points until three types of alternative points are clustered, and confirming the position of a valley point by a position relationship among the centers of the three types of alternative points; 4, finishing palmprint segmentation by taking the valley point as a reference point. According to the method, the speed of segmenting the effective region of the palmprint on line is increased by a relation of the gradient direction of the profile point of the palm and the palm direction; the method can be used for identity verification in mobile equipment.

The invention discloses a WSN wireless communication module fault diagnosis method based on a fuzzy neural network. A fuzzy neural network current model is established by using emission consumption parameters corresponding to a DHT11 temperature and humidity sensor under different temperatures and voltages for the fault diagnosis of a wireless communication module. For data subjected to normalization processing, firstly an initial structure and parameters of the fuzzy neural network are adaptively determined by using subtraction clustering, then parameter optimization and adjustment are carried out on the model by using a hybrid learning method combining the particle swarm optimization algorithm with the least square method, and finally fault diagnosis is carried out on a test sample by using a trained diagnosis model. According to the WSN wireless communication module fault diagnosis method disclosed by the invention, the advantages of fuzzy reasoning and the neural network are integrated, an improved learning algorithm is adopted, the fuzzy neural network current model of the wireless communication module is established for the relation among the current, the voltage and the faults of a WSN, and the model is short in training time, high in convergence speed and high in fault diagnosis efficiency.

The invention discloses a method and a device for positioning a high-resistance grounding fault section of a power distribution network and a storage medium. The method comprises the following steps of acquiring zero-mode current recording data of fault fixed-section terminals; selecting zero-mode current initial wave heads according to the zero-mode current recording data, and aligning travelingwave heads of the zero-mode current recording data of the terminals; when the amplitude values of the zero-mode current initial wave heads are not smaller than a rated current value, according to thezero-mode current initial wave heads, calculating a product integral of sampling points of each terminal and the next terminal after breakdown, and determining a first grounding fault section according to a calculation result, thereby obtaining an initial wave head judgment result; performing cross wavelet transformation on the zero-mode current recording data acquired by each terminal and the next terminal, and determining a second grounding fault section according to a transformation result, thereby obtaining a cross wavelet phase judgment result; and when the initial wave head judgment result and the cross wavelet phase judgment result are consistent, obtaining a fault fixed-section result. The fault section can be accurately judged; and the method and the device have relatively high adaptability.

The invention relates to a hand language searching method which comprises the following steps of: S1: carrying out video decoding on the hand language, extracting texture characteristics, color characteristics and outline characteristics of the bottom layer of an image, generating a hand language characteristic library according to hand language motion characteristics, acquiring hand language characteristics comprising coordinate characteristics of two-hand motion, speed characteristics of two-hand motion and shape characteristics of two hands by adopting a method for tracking two hands by the hand language to form time sequence characteristics, and storing in character strings of the hand language characteristics; S2: expressing the video input by a user by the coordinate characteristics, the speed characteristics and the shape characteristics, and acquiring the character strings of the video input by the user; and S3: carrying out distance measurement on the character strings of the video input by the user and the character strings of the hand language characteristics by adopting a character string editing distance algorithm, and acquiring the similarity of the video input by the user and the video in the hand language characteristic library. The invention solves the problem of fast and robust two-hand positioning in the hand language, the problem of two-hand characteristic expression and the problem of fast and effective hand language similarity measurement.

The invention relates to a communication fault detecting device and method in an air conditioning system which has an outdoor unit connected with a plurality of indoor units, aiming at making use of the wired/wireless controller to cut off the communication of the indoor units even when the communication lines connecting the outdoor unit and the indoor units are not separated, so that the indoor unit with communication fault can be more quickly and easily found. To achieve the aim, the communication fault detecting method provided by the invention in the air conditioning system which has an outdoor unit connected with a plurality of indoor units includes: whether a communication fault happens to the air conditioning system is judged; when the communication fault occurs, selectively separate the indoor units through the communication lines and then find out the indoor unit with communication fault.