67results about How to "Quick and accurate extraction" patented technology

The invention discloses a remote-sensing image coastline extracting method based on an information vector machine. The remote-sensing image coastline extracting method includes the steps of firstly obtaining a remote-sensing image containing coastline information, calculating the normalization difference water body index according to wave band information of the remote-sensing image, and obtaining a NDWI image; then selecting sample points on the NDWI image, extracting the color characteristics, the textural characteristics and the types to construct a training sample, and training an IVM model; then automatically classifying all pixels of the image through the trained IVM model, and partitioning seawater from land of the remote-sensing image; finally extracting a coastline with the graying and binarization image processing technology. The remote-sensing image coastline extracting method has the advantages that the coastline extracting accuracy under noise pollution is remarkably improved, the coastline is rapidly and accurately extracted, and an efficient technological means is provided for measuring, identifying and analyzing the coastline.

The invention discloses a method for extracting video captions quickly and accurately, and the procedure thereof is as follows: extracting video key frame brightness images; generating a texture, projecting and positioning caption strips; confirming the initiation frame and the termination frame of the caption strips; and extracting characters. Texture diagrams of the video key frame brightness images are extracted by wavelet transform; the positions of the caption strips of the key frame images are positioned by utilizing the texture diagram projection method; the same caption of two frames are judged by the caption point match method; and the initiation frame and the termination frame of the caption strip can be quickly confirmed in the video, therefore, all the continuous frames of the caption strips can be confirmed only by invoking one localization; and the characters in the caption strips are extracted by the texture noise elimination method.

The invention discloses a forest burned area automatic extraction method based on a time sequence remote sensing image. The method comprises the following three steps: 1, selecting cloudless remote sensing images before and after occurrence of a forest fire, calculating variable quantity of each band and a vegetation index difference image of two time phase remote sensing images, and carrying out weighted fusion on the variable quantity of each band and the vegetation index difference image to obtain a difference image; 2, carrying out linear fitting on near infrared bands before and after the occurrence of the fire to obtain a fitting error image, then carrying out binarization classification on the fitting error image, and setting a classification result as an initial curve of a C-V (Chan-Vese) model; 3, inputting the obtained difference image and the initial curve as the C-V model, and carrying out segmentation on the difference image obtained in the step 1 by using the C-V model to realize the automatic extraction of burned area information. By adopting the extraction method disclosed by the invention, a burned area can be fast and automatically extracted by using the remote sensing images, and the method is strong in operability, clear in extracted burned area contour, high in detection precision, and strong in practicability, and further can be used for extracting forested region burned areas in a variety of land satellite image data.

The invention discloses a method of rapidly positioning an area of interest of a palm print. The method comprises five processes of marking valley points among fingers of a training sample; training a convolution nerve network used for detecting valley points among fingers of a low-resolution palm print; extracting a key point among fingers of the low-resolution palm print and acquiring a subgraph which takes the key point as a center in a high-resolution palm print; aiming at the subgraph, rapidly extracting an accurate valley point among fingers and positioning the area of interest of the palm print according to the valley point. During a process of training the convolution nerve network, the training sample is amplified and then is input into a network for several times in batches, and a learning rate is gradually reduced along with increase of the batches so as to increase a training speed during the training. During an extraction process, the key point among the fingers of the low-resolution palm print is firstly positioned, and then the subgraph taking the key point as the center is acquired in the high-resolution palm print through mapping, the convolution nerve network or a tangent method is used in the subgraph to carry out local search so as to rapidly extract the valley points among the fingers, and finally, the points are used to accurately position the area of interest of the palm print.

The invention relates to a Canny operator-based registration image edge extraction method. The method comprises the following steps of: A, smoothing a registration image which is registered on site; B, judging the brightness of the edge points, and determining the gradient intensity and the gradient direction of the registration image; C, performing non-maximum suppression on each local gradient region in the global gradient map in the step B, and determining edges in the registration map; Step D, further extracting real edge points in the registration image by utilizing a threshold method; And E, carrying out connectivity analysis on the edge points subjected to threshold processing to form a final edge image. According to the method, the Canny operator is used for carrying out edge extraction on the target on the registration image, smoothing processing is carried out on the image, noise is reduced, and error detection and omission in the image edge extraction process can be prevented while the detection precision is improved.

The invention discloses an image processing method in an underwater vision SLAM (Simultaneous Localization And Mapping) system. The image processing method comprises the following steps: establishing an underwater imaging model; processing influence of underwater environment factors on a camera imaging; carrying out image feature extraction and matching. According to the method, extraction on feature points of an underwater environment image in the later period is reinforced; according to a data correlation method which is newly disclosed on the basis of the improved SLAM system, the feature points are extracted in a matched manner, so that the feature points can be more rapidly and accurately extracted; real-time performance of an SIFT algorithm is improved. The method which uses a relative position factor of a binocular camera and a way point position factor as auxiliary conditions for carrying out correlation can effectively solve the problems of mismatching and poor matching efficiency in the data correlation.

The invention discloses a vehicle detection method for an intelligent automobile under severe weather conditions, and solves the problem that the intelligent automobile is difficult to detect in severe environments such as night, rainy days, foggy days and snowy days. The method includes: performing vehicle target detection and image capture through a millimeter-wave radar and a thermal imaging camera respectively; projecting a vehicle target detected by the millimeter-wave radar to an infrared thermal image captured by the thermal imaging camera through coordinate transformation to obtain approximate position distribution of the vehicle detection target in the infrared thermal image; extracting and segmenting a vehicle region of interest; performing hypothesis verification on the region of interest of the vehicle by adopting a DMP target detection algorithm, and performing regression prediction on a detection frame of the vehicle; fusing the detection targets of the millimeter-wave radar and the thermal imaging camera by calculating a cost matrix, and finally tracking the fused vehicle detection target by adopting Kalman filtering.

The invention discloses a keywords extraction method which is used for extracting keywords from a paragraph of words. The method comprises the following steps: paragraph segmentation, namely, segmenting the paragraph of characters into words; stop words elimination, namely, eliminating stop words from the words obtained through segmentation in the paragraph segmentation step; word characteristic reduction, namely, carrying out reduction on word characteristics of the words with the stop words extracted, wherein a stem analytical algorithm is adopted for reducing the words into the stem prototypes; and keywords determination, namely, determining the occurring frequency of each item prototype in the paragraph of words based on the stem prototypes, and determining the keywords based on the occurring frequency. With the adoption of the keywords extraction method, the keywords can be extracted from one paragraph of words rapidly and accurately, so that the keywords extracted by a computer are approximate to keywords obtained through understanding of a human brain to the furthest.

The invention provides a block chain system and an application method thereof. The block chain system comprises one or more routing nodes and one or more storage nodes; each routing node is used for determining one or more target storage nodes in one or more storage nodes when receiving data and sending the data to one or more target storage nodes; a storage condition of one or more target storagenodes for storing the data is monitored, and when it is monitored that one or more target storage nodes complete data storage, data storage task information is broadcasted; each storage node is usedfor storing the data when receiving the data sent by the routing nodes. Therefore, according to the scheme, the efficiency of storing the data can be improved.

The invention discloses a parallel EPLL-based pulsating high frequency signal injection method signal extraction system and strategy comprising the following steps: carrying out Clark and Park transformation for three-phase current, thus obtaining d and q shaft currents; improving an EPLL strategy so as to extract the amplitude of the q shaft current containing a rotor position signal and the q shaft high frequency signal amplitude, thus obtaining a pre-estimate rotating speed and angle, and q shaft second harmonic signal positive negative value correction angles; adjusting PI of DC signals so as to obtain d and q shaft real voltages; carrying out Park inverse transformation for the d and q shaft real voltages so as to obtain [alpha] and [beta] shaft voltages; modulating an inverter so as to control a motor. The system and strategy can realize non-position operation of the motor at a low speed, can obtain the pre-estimate rotor position signal, can improve the filter delay problems on the basis of the pulsating high frequency signal injection method, can simplify the signal extraction process, and thus providing response speed and signal extraction precision superiority.

The invention belongs to the technical field of additive manufacturing on-line measurement, and discloses an on-line detection method of the contour accuracy of a molding layer in the laser selectivemelting process. The on-line detection method includes the following steps that S1, a model of a to-be-machined part is sliced, and an auxiliary image is generated; S2, powder spreading is conduced, and then powder is melted and molded in a laser selective mode; S3, an image of a basic molding area is collected, and a contour of a segmented image is extracted; S4, the image contour is subjected tothree-dimensional reconstruction to obtain an image actual contour; S5, the image actual contour Cr is compared with a contour of a corresponding slicing layer to analyze the accuracy, if the requirements are met, the step S6 is executed, and otherwise, machining is completed; and S6, whether machining of the to-be-machined part is completed or not is detected. According to the on-line detectionmethod of the contour accuracy of the molding layer in the laser selective melting process, the range of image contour extraction can be effectively and accurately delimited, and then the number of iteration times in the calculation process is greatly decreased; and thus the extraction accuracy of the image contour reaches the sub-pixel level, and accurate and rapid contour detection is realized.

The invention relates to an extraction method and device of digital information. The method comprises the following steps: obtaining text information selected by a user; determining whether the text information contains the digital information or not; when the text information contains the digital information, extracting the digital information in the text information; and when a monitoring result shows that an appointed entrance is activated, filling the digital information into the appointed entrance. When the digital information in the text information is extracted, the user does not need to accurately select the digital information in the text information, the user can firstly select the text information which contains the digital information. Therefore, under the situation that the click identification rate of the touch screen of a mobile terminal is low, the digital information in the text information can be quickly and accurately extracted and is input to the appointed entrance.

The invention belongs to the field of electric power data processing, and relates to a power grid key node identification method, device and system. The method comprises the following steps: constructing a power grid topological structure with power effectiveness according to effectiveness rules of power node injection power and line transmission power in a power grid; in the power grid topological structure, calculating an average electrical distance and an electrical betweenness center of power nodes, respectively allocating weight vectors, and constructing a criticality evaluation model ofthe nodes; and calculating a power node nuclear value according to the criticality evaluation model, carrying out recursive network decomposition according to an interval to which the power node nuclear value belongs, and taking the power node contained in the last layer of sub-network as an identified key node set. According to a series of rules, rule characteristics of a large-scale power systemare mined, a standard power grid topological structure is constructed, network characteristics and electrical characteristics of the power system are comprehensively considered, a power grid key nodeidentification model based on multiple factors is established, and efficiency and accuracy of key node identification in a complex power grid system are improved.

The invention relates to the technical field of text extraction. The invention provides a website navigation bar information extraction method and device, electronic equipment and a storage medium, and the method comprises the steps: downloading source codes of a main page and any subpage of a to-be-extracted enterprise website domain name, obtaining a first HTML code, analyzing the first HTML code into a first node DOM tree, and obtaining a second HTML code, and analyzing the second HTML code into a second node DOM tree; removing outer links of the first node DOM tree and the second node DOMtree to obtain a third node DOM tree and a fourth node DOM tree; extracting navigation bar information by using an NAV tag method, an A tag density method, a maximum public area method and a keyword link block method, then performing duplicate removal and filtering, calculating a node score of each node, and outputting the navigation bar information of the enterprise to be extracted. According tothe method, the navigation bar information is extracted through the NAV tag method, the A tag density method, the maximum public area method and the keyword link block method, so that the accuracy andthe efficiency of extracting the navigation bar information in the page are improved.

The invention provides a content extracting method and device based on semantic analysis and rules. And the traditional content extracting method based on a template is combined on the basis of the content extracting method and device based on semantic analysis and rules. By use of the method and device provided by the invention, the advantages that the template extracting speed is fast, the data is accurate are realized, and the semantic analysis and rules are strong in parsing adaptability are realized; through the combination of two ways, the content data can be quickly and accurately extracted.

The invention discloses an intelligent full-text retrieval method and system based on semantic comprehension, and the method comprises the steps: segmenting a received search statement into short texts, and carrying out the word segmentation operation of the short texts, so as to obtain a word segmentation library corresponding to the short texts; constructing a semantic information vector and a dependency relationship vector of the short text, wherein the semantic information vector comprises a head word and a word-meaning co-occurrence word of the short text; and based on the semantic information vector and the dependency relationship vector of the short text, performing similarity calculation on the short text information and related information in the intelligent index database to obtain a search result set. According to the method and system, the original data is divided into the multiple short texts to form the search text vector, and the similarity between the search text and the index database text is calculated by calling the semantic understanding interface of the artificial intelligence platform, so that the accuracy of full-text retrieval can be improved.

The application provides a voice-assisted recognition method and device, a storage medium and electronic equipment. The electronic equipment can acquire mouth motion information of a face image in a shooting range while acquiring voice information containing multiple voices, then, a voice-assisted training model at a pre-training position is utilized to analyze and calculate the mouth motion information and multiple voices, target voices output by users of the electronic equipment can be quickly and accurately extracted from the voices, thereby guaranteeing that the output text information isthe content said by the users in a noisy scene, and the efficiency and the accuracy of voice recognition in the noisy scene are greatly improved.

The invention discloses a human body slice image automatic segmentation method based on shared matting and ROI gradual change. The method comprises the following steps: S1, roughly doodling an interested area; S2, selecting an internal region A of the region of interest; S3, covering the region of interest and reading a boundary region of the region of interest; S4, defining a foreground region of the internal region A as a white part, defining a middle part as a gray part, defining a background region as a black part, and obtaining a tripartite graph of the region of interest; S5, obtaining a black-and-white mask of the region of interest by adopting a shared matting algorithm according to the tripartite graph information, and obtaining a foreground skeleton of the black-and-white mask by applying a refinement algorithm; S6, taking the foreground skeleton as a foreground graffiti of the next picture, and obtaining a tripartite graph of the next picture by adopting a flooding algorithm; and S7, repeating the steps S5 to S6 until all the pictures are processed to obtain black and white masks of the regions of interest of all the pictures.

The invention discloses a gesture error correction method, system and device in an augmented reality environment. The method comprises the following steps: firstly, acquiring a first gesture depth mapand gesture depth information of a hand; judging whether the current gesture is recognizable or not by utilizing a gesture recognition model; when the gesture depth map cannot be identified, determining the minimum feature distance of the current second gesture depth map according to the Hausdorff distance, determining a third gesture depth map through the Hausdorff distance, and finally, inputting the third gesture depth map into the gesture identification model, so that gesture error correction is realized. The system comprises an information acquisition module, a preprocessing module, a first input module, a judgment module, a minimum feature distance determination module, a third gesture depth map determination module and a second input module. The device comprises a processor and a memory connected with the processor, and the processor can execute the gesture error correction method. Real-time performance and accuracy of gesture error correction can be greatly improved.

The invention discloses a binocular inertial simultaneous localization and mapping method based on point-line feature fusion. Length suppression, near-line merging and broken-line splicing methods are added after online feature extraction. Point, line and IMU data information are effectively fused in the sliding window based on optimization, the problem that a point feature visual SLAM algorithm in a weak texture environment is low in precision and unstable is solved, higher-precision pose estimation is achieved, length suppression, near-line merging and broken-line splicing strategies are utilized, while rapid extraction is guaranteed, the problem that the line segment extraction quality of a traditional algorithm is not high is solved, the mismatching rate of system line features is reduced, and the method can better adapt to indoor weak texture and texture-free scenes.

The invention provides a building area extraction method and system based on an SAR image. The building area extraction method based on the SAR image includes: acquiring a first image on the basis ofthe SAR image; generating a local space index image and a variogram image on the basis of the first image; respectively screening sampling points on the basis of the local space index image, the variogram image and the first image to obtain corresponding sets; and processing sampling points in each set according to preset growth criteria to respectively construct to-be-selected building area images, superimposing all the to-be-selected building area images to obtain a final building area image. Through the method or the system provided by the invention, a building area can be quickly and accurately extracted from the Sentinel-1 SAR image, automatic extraction of construction land of a large region is realized, and a foundation is laid for global building area mapping.

The invention provides a method and apparatus for determining a time domain parameter of the partial discharge current of a high voltage IGBT. The method includes the following steps: pre-screening partial discharge currents of a high voltage IGBT to obtain partial discharge pulses; then counting the partial discharge pulses that are pre-screened, and extracting the current magnitudes of the partial discharge pulses; and eventually based on the current magnitudes of the partial discharge pulses, obtaining a time domain magnitude of the partial discharge pulses. According to the invention, thetime domain parameter of the partial discharge pulses obtained by the method herein can sensitively detect and identify the defect of insulation in the process of designing and manufacturing the highvoltage IGBT. And in case the large amount of original data obtained from tests do not undergo denoise pre-processing, the method herein can accurately and rapidly extract the time domain parameter ofthe partial discharge pulses of the partial discharge currents, and further acquires a partial discharge mode in presence of the defect of insulation of the high voltage IGBT.

The invention discloses an impermeable surface monitoring method and device based on Landsat time series images. The method comprises step 1, aiming at the Landsat time series images after registration, establishing time series LST, time series BCI and time series NDVI; step 2 interpolating the missing data in each time series, eliminating the noise and year of each time series In step 3, the stable time series representing the impervious surface is selected from the smooth time series as the class label time series samples, and the difference between the non-class label time series samples and the class label time series samples is measured. Similarity distance, obtain the similarity distance map corresponding to each time series; step 4 uses the C4.5 algorithm to classify the similarity distance map at different time intervals, and monitors its temporal and spatial variation rules by using the temporal and spatial distribution characteristics of impervious surfaces. The invention improves the monitoring accuracy of the impermeable surface.

The invention provides a weld bead contour determining method and device, a processor and a system. The determining method comprises the steps: obtaining an initial weld bead image; performing first edge detection on the initial weld bead image to obtain a middle weld bead image; determining a region of interest according to the middle weld bead image; and performing predetermined processing on the region of interest to obtain a weld bead contour, the predetermined processing at least comprising second edge detection. According to the determination method, the first edge detection, the region of interest determination and the second edge detection are sequentially carried out on the obtained initial welding bead image, so that a relatively accurate welding bead contour can be obtained in a relatively short time; the problem that in the prior art, due to the fact that base metal is polished to form light reflection in the welding process, the edge contour of the welding bead is not clear and not obvious is solved, it is guaranteed that the welding bead contour can be rapidly and accurately extracted from the welding bead image, and then it is guaranteed that the effect of follow-up automatic welding according to the welding bead contour is good.

The invention provides a standard text labeling and standard atlas construction method and device. The method comprises the following steps: extracting a standard element text and a non-standard element text from a to-be-labeled standard text based on a structured template; extracting a standard element triple from the standard element text based on a preset keyword and a preset rule; reading and understanding the non-standard element text, and extracting a non-standard element triple; and respectively carrying out entity fusion on the standard element triples and the non-standard element triples, and carrying out text labeling on the triples after entity fusion. The standard element triples are extracted on the basis of the preset keyword and the preset rule, and the non-standard element triples are extracted on the basis of reading comprehension, so that the triples used for constructing the standard atlas can be rapidly and accurately extracted, and the problem that the efficiency of constructing the standard atlas is low due to the fact that texts are manually labeled in a traditional method is solved.

The invention discloses a three-dimensional detection method based on structured light and multiple light field cameras, and the method comprises the steps: building a three-dimensional detection system comprising a structured light source and a plurality of light field cameras, and placing a reference object of a to-be-detected target object in a working range of the three-dimensional detection system, carrying out structured light-based three-dimensional reconstruction on the reference object to obtain a reference object three-dimensional model, then placing a to-be-detected target object in a working range of the three-dimensional detection system, and carrying out structured light-based three-dimensional reconstruction on the to-be-detected target object to obtain a to-be-detected target object three-dimensional model; and performing three-dimensional detection on the reference object three-dimensional model and the to-be-detected target object three-dimensional model, and outputting a three-dimensional position of a key point of the to-be-detected target object. The advantages of the light field camera and the structured light in the field of close-range three-dimensional reconstruction are fully utilized, and three-dimensional detection of the target object in the working range can be accurately and efficiently completed.

The invention relates to a method and device for extracting official document elements based on OCR character recognition, belongs to the technical field of intelligent file processing, and solves theproblems that an existing method wastes manpower and time and is low in efficiency. The method comprises the following steps: scanning a paper official document file containing official document element information to obtain an electronic official document file; dynamically generating an algorithm selection box based on the electronic official document file, and obtaining a corresponding algorithm based on the algorithm selection box; acquiring official document element information in the electronic official document file based on an algorithm; and storing the obtained official document element information to a corresponding field position of an official document element form in a business information processing system. The invention is simple and feasible, is easy to implement, achievesthe quick extraction of the official document elements, saves the manpower and cost, and improves the efficiency.