1114results about How to "Improve extraction accuracy" patented technology

A method of automatically narrowing data search space and improving accuracy of data extraction using known constraints in a layout of extracted data elements for classified documented is provided. The method includes: analyzing each document to classify it within a document category, each category having a corresponding set of expected layouts; analyzing each electronic document to automatically extract images and text features; automatically constructing a data structure including a layout of the extracted features and layout relationships amongst the extracted features, wherein each of the extracted features in the layout maintains a reference to neighboring features and wherein closely related features are merged to form a combined feature; automatically narrowing data search space by detecting and removing parts of the layout that are not associated with any data elements using the data structure; and automatically detecting data using the extracted feature layout and the layout relationships amongst the extracted features.

The present disclosure relates to diagnostic apparatus and method for home appliance which make it easy to diagnose a breakdown of a home appliance, using a signal including product information and outputted from the home appliance, in which a signal is converted in accordance with signal distortion generated in the process of transmitting the output signal or in a terminal, so that the product information of the home appliance is easily extracted, accuracy of a diagnosis result is improved, an after-sales service for a breakdown of the home appliance is easily provided, and reliability of the product is improved by the improvement of accuracy of the diagnosis result.

A document image processing device and method for extracting a title region and a mark attached by the user from a document image to use them as document tag information. A region with a region average character size larger then a predetermine extraction judging value is extracted as a title region by title region extracting means. As a result, title regions can be extracted from one document image. A mark that the user makes on an input image is extracted by mark extracting means, and characteristic value of the mark is found by calculating means. Document tag information to be imparted to the input image is selected from reference tag information according to the characteristic value and the attribute value of the reference tag information imparting means. Thus, document tag information is automatically imparted to a document image.

The invention discloses an object-oriented remote sensing image coastline extraction method which belongs to the field of full automatic remote sensing image coastline extraction. The steps are as follows: carrying out filtering treatment on a remote sensing image; selecting the segmentation algorithm to segment the remote sensing image; utilizing sample points for classifying the remote sensing image after segmentation; utilizing the seed growth method for extracting a seawater region; and finally utilizing the corresponding judgment criterion to extract a coastline. By utilizing the method disclosed by the invention, the accuracy of the existing coastline extraction algorithm can be improved, the correct rate of seawater extraction after segmentation and classification treatment throughthe object-oriented method is improved and the accuracy of the coastline extraction is ensured. As the coastline changes frequently, the method can provide convenience for maintenance and updating ofcoastline information of all levels of national fundamental geographic information databases.

The invention provides a text relation extraction method and system based on a hierarchical knowledge graph attention model, and the method comprises the steps: firstly selecting a training text set and a knowledge graph according to a to-be-processed text, and carrying out the preprocessing of the training text set and the knowledge graph; Then constructing a hierarchical knowledge map attentionmodel, and training the model by using the preprocessed training text; And finally, marking the head entities and the tail entities of all sentences in the to-be-processed text, and inputting the marked to-be-processed text into the trained hierarchical knowledge map attention model to obtain a relation result. The method has the advantages that the hierarchical knowledge graph attention model isprovided according to the technical scheme, the knowledge graph is used for representing the weights of sentences and words distributed to the model, the accuracy rate and the recall rate of relationprediction of the model are improved, and then the extraction accuracy of text relations is improved.

The invention discloses a method for extracting circular facula sub-pel center and a device thereof. In the proposal of the invention, firstly, the initial position of the circular facula center is determined by utilizing a center of mass method; then, the pel position of the circular facula center is determined in a window by taking the initial position as a center according to the characteristic value of Hessian matrix; finally, image gradation function which is continuously distributed in a neighborhood of the pel position of the circular facula center is determined by using second order taylor expansion, and the position of extreme point of the image gradation function is the sub-pel position of the circular facula center. The searching range of the circular facula center is narrowed to a smaller area fast by the proposal of the invention, thus leading the follow-up operation such as gauss convolution, the calculation of S and the searching of the extremum to be carried out in the smaller area, and the calculated amount is small and the extracting speed is rapid. Therefore, the sub-pel position of the circular facula center is determined by the taylor expansion and the extracting precision of the circular facula center is ensured.

The invention provides a method for extracting content of a web page, which comprises the following steps: 1) performing a paragraph merger operation on all characterization nodes in a DOM tree constructed by the web page; 2) optionally selecting a node from a node set bearing the longest paragraph, backtracking from the selected node to a root node of the DOM tree, calculating the difference of punctuation amounts of contexts, which are born between a father node of each node and a subtree thereof and between the node and a subtree thereof, and forming an increment sequence in turn; 3) searching the increment of a first 0 value or a first increment smaller than a former adjacent increment and smaller than or equal to a latter adjacent increment in the increment sequence; 4) if the increment is found, determining that the context born by the corresponding node of the increment and the subtree thereof is the content of the web page; 5), if the increment is not found, determining that the context born by the root node of the increment and the subtree thereof is the content of the web page. The method for extracting the content of the web page has the advantages of high extraction precision, high implement speed, low maintenance cost and excellent adaptability.

The invention discloses a method for extracting a center line of a laser stripe and belongs to the technical field of machine vision. The method comprises the following steps: S1, carrying out de-noising processing on an optical stripe image to obtain a noiseless laser stripe image; S2, carrying out region extraction on the noiseless laser stripe image obtained in the step S1 so as to obtain a small-area rectangular region comprising all the laser stripes; S3, removing pixel isolated points in the rectangular region obtained in the step S2; S4, processing the laser stripe which is obtained in the step S3 and of which the edge is not provided with burrs by adopting a Gaussian convolution method; S5, processing the smooth fuzzy laser stripe obtained in the step S4 twice by a gray weighted centroid method and extracting a secondary center line of the smooth fuzzy laser stripe; S6, carrying out nonuniform B spline fitting on the secondary center line obtained in the step S5 for three times so as to obtain the optimized center line of the laser stripe, i.e. implementing extraction of the center line of the laser stripe. According to the method disclosed by the invention, accuracy of extracting the center line is greatly improved; moreover, the method has a wide application range.

The invention discloses a high-resolution remote sensing image building automatic extraction method and system by using shadow. The method comprises the steps of: firstly performing sample automatic extraction of shadow, vegetation, bare land, and building on an image, combining an SVM classification principle and realizing automatic classification, and extracting a building initial result; and then performing post-processing optimization on the initial result, the post-processing optimization including morphology pre-processing, missed detection supplement and building verification or the like, and outputting the building final result. The system includes a sample area automatic extraction module, an SVM classifier training and outputting module, and a post-processing optimization module. The building extraction automation degree is effectively improved and the precision can be ensured by designing the sample automatic extraction method and effective post-processing optimization measures.

The method includes following steps: (1) preprocessing coarse web pages, and normalizing all web pages to more normative format; (2) primary positioning contents of web pages including information to be extracted in files of web page; (3) according to specified rules to extract metadata from contents picked-up by primary positioning operation. First, the invention distinguishes core area from hash in large area. Then, aiming at core area, the invention carries out regular extraction so as to raise accuracy of extraction greatly. The invention also can extract metadata in web page according to multiple rules. Based on given priorities, the multiple rules determine matching sequence, and carries out refined process according to method of extraction in two stages.

The invention discloses a rice yield remote sensing estimation method based on MODIS data, which comprises: 1) acquiring the MODIS09 remote sensing data in the growth process of the rice in the region to be monitored, and calculating an vegetation index EVI and a water index LSWI; 2) extracting a rice pixel based on the relation between the water index LSW1 and the vegetation index EVI; 3) acquiring the vegetation index EVI corresponding to the rice in different stages; 4) calculating the average of the vegetation indexes EVIs of different set administrative regions in different stages to acquire the average EVI of the set administrative regions; 5) establishing a statistical model for the average EVI of the set administrative regions and the rice yield; and 6) estimating the rice yields of the set administrative regions using the vegetation index EVI in the growth period of rice in the year to be estimated, and calculating the rice yield of the region to be monitored. The temporal resolution, the spatial resolution and the estimation precision of the invention are relatively high, and the invention is applicable to monitoring rice yield in a large scale.

The invention discloses a generation method of single document summaries. The method includes the steps of clustering paragraphs of a document to be summarized, and defining each class as a semantic block; calculating similarity of each sentence pair in the semantic blocks to score one sentence with the other sentence, and defining the sentence with highest score as a sentence expressing partial core content in each semantic block; and connecting the sentences to generate a summary according to emergency sequence of core sentences. Word similarity and named entity recognition are introduced to single document summaries, so that extracting precision of summaries is higher. Clustering speed is increased by means of single pass. Single document summaries can be extracted accurately. The generation method is high in accuracy of extracting news and announcement documents.

The invention discloses a system for extracting a picture abstract based on page partitioning, which comprises a page preprocessing module, a page sorting module, a page partitioning module, a subject block identifying module and an information extracting module. The invention also discloses a method for extracting the picture abstract based on the page partitioning, which specifically comprises the following steps that: 1, a page is crawled on the Internet; 2, the page preprocessing module preprocesses the page; 3, the page sorting module sorts the preprocessed page; 4, the page partitioning module partitions a semantic chunk; 5, the page partitioning module sends the subject type page to the subject block identifying module, and the subject block identifying module identifies the subject block and then sends the subject block to the information extracting module; and the page partitioning module sends the non-subject type page to the information extracting module; and 6, the information extracting module downloads the picture and is correlated with the page. The system and the method have the advantages of high extracting speed, high accuracy and good effect.

The invention discloses a method for extracting a target object from an interactive video, comprising the following steps of: obtaining the frame difference image of three adjacent frames of images by a frame difference method as a rough estimate of the target object; then combining with the coordinate position of the rectangular region of the front frame of image to determine the rectangular region containing the target object in the present frame of image; after that, structuring the energy function of each frame of image in an original video, and realizing the accurate extraction of the target object; adequately utilizing the coordinate position of the rectangular region of the front frame of image during the process of determining the rectangular region containing the target object inthe frame of image, thus, the target object region of the present frame can be accurately determined even if aiming at the conditions of target object suddenly turning to static or background shakingand the like; and finally establishing a fraction model for each frame of image after the target object region is determined, wherein the target object can be accurately extracted by a graph cut method, thereby greatly improving the extraction accuracy.

The invention discloses a webpage information extracting method and system. The webpage information extracting method comprises the following steps of: analyzing a webpage to be extracted as a document object model (DOM) tree, and obtaining a template corresponding to the webpage to be extracted; traversing the DOM tree according to webpage division fineness defined by the template, and dividing the corresponding webpage into content blocks; and outputting content and type information of the content blocks according to an output rule defined by the template. With the adoption of the webpage information extracting method and system provided by the invention, the precision of extracting webpage information can be improved.

The invention provides a nonlinear active-disturbance-rejection control method for a vertically rising and falling repeatedly-used carrier. The method comprises the following steps of step 1,attitudecontrol model establishment; step 2,TD tracking differentiator design; step 3,nonlinear expansion state observer design; step 4,nonlinear active-disturbance-rejection controller design; and step 5,nonlinear feedback control law design. The nonlinear active-disturbance-rejection control method for the vertically rising and falling repeatedly-used carrier is provided for the defects of longer attitude response time,weak anti-noise ability,lower attitude control precision and the like of an existing active-disturbance-rejection controller,combines a nonlinear TD tracking differentiator with a finite time convergence characteristic,an expansion state observer with fixed time convergence and a nonlinear feedback control law to form a novel nonlinear active-disturbance-rejection controller,thereby improving the complex external interference rejection ability of a system,and also improving the attitude control precision and response speed.

The invention provides a rapid extraction method for dense point cloud oriented toward city three-dimensional modeling. The rapid extraction method comprises the steps of: 1) preparing data; 2) establishing a coordinate grid; 3) converting an image into a gray scale image; 4) judging whether a difference value between a gray value of each target pixel point and a gray value of a neighborhood pixel point is greater than a threshold value A, if so, regarding the target pixel point as a candidate angular point, and if not then rounding; 5) calculating a corner response function (CRF) value of the candidate angular point, and judging whether the CRF value of the candidate angular point is greater than a threshold B, is so regarding the candidate angular point as a feature point, and if not then rounding; 5) matching homonymy points; 7) generating seed points; 8) conducting regional diffusion; 9) conducting gross error elimination; 10) and generating dense point cloud to form a tin for establishing a three-dimensional model. Compared with the prior art, the rapid extraction method provided by the invention has the advantages of small error, high precision, high operation efficiency, wide application extension and the like.

The invention provides a method for extracting a keyword from a text. The method comprises the following steps: dividing the text into words to obtain the words corresponding to the text; labeling word characteristics of the words; adding statistical information to the words; indexing the words; inputting label information, the statistical information and indexing information of the words into a deep learning extraction model so as to obtain importance weights of the words; according to the importance weights, selecting at least one corresponding word as the keyword of the text. In correspondence to the method, the invention further provides a device for extracting the keyword from the text.

The invention discloses a high-resolution SAR image individual building extraction method, and the method comprises the steps: firstly carrying out domain ontology modeling through combining the imaging features of a high-resolution SAR image individual building and the characteristics of a complex morphological structure, and constructing an individual building body semantic model in a high-resolution SAR image; secondly obtaining image regions with the good homogeneity and clear boundary through employing SAR image segmentation based on an object, wherein the image regions are the basic processing units extracted by building base units; combining with the related rule of the building base units in the body model, wherein the extracted image object characteristics comprise regional features, shape features, geometric features, texture features, and topological features; forming the object expression of body semantic description according to the body semantic rule and the object features, guiding the image objects to be automatically recognized as the building base units, and achieving the large-scale individual building recognition which takes the semantic knowledge as the center. The method can extract a large-scale individual building in the high-resolution SAR image accurately and quickly.

The invention discloses a regional water body rapid dynamic extraction method combining optics and radar. The method comprises the following steps: obtaining optical remote sensing image data of a monitoring region; obtaining radar remote sensing image data of the monitoring area; preprocessing the optical remote sensing image data and the radar remote sensing image data respectively; obtaining aspectral index based on the preprocessed optical remote sensing image data; extracting an optical remote sensing area water body classification map based on the spectral index; based on the optical remote sensing area water body classification map, extracting a radar remote sensing area water body classification map in the preprocessed radar remote sensing image data; and obtaining a monitoring area water body map based on the optical remote sensing area water body classification map and the radar remote sensing area water body classification map. According to the invention, dynamic extractionand monitoring of water bodies in a regional range can be realized.

The invention discloses a method for extracting a tower target in an unmanned aerial vehicle routing inspection power transmission line image, and belongs to the technical field of pattern recognition and intelligent computation technologies. The method for extracting the tower target comprises the steps of converting an image acquired by aerial photographing from an RGB color space into an HSI (hue / saturation / luminance) space, carrying out binarization processing on an I component image of the HIS space after being subjected to median filtering so as to acquire a binary image of an electric power tower profile; carrying out angular point detection on the binary image of the electric power tower profile by adopting an improved Susan angular point detection method, and marking positions of all angular points in the binary image of the electric power tower profile; and detecting the mass center of an angular point sequence based on a moment invariant mass center detection algorithm, and marking the position of the electric power tower in an original image by taking the mass center as a center. Through applying the technical scheme disclosed by the invention, the electric power tower can be extracted accurately in real time, and a guarantee is provided for fault diagnosis of electric power components such as insulators, hardware fittings and the like on the tower.

The invention provides a keyword extraction method. The method comprises the steps of performing word segmentation operation on all analysis statements of an extracted text to obtain word units of theanalysis statements; determining dependency association degrees between candidate words in the analysis statements and other candidate words in the analysis statements; determining word attraction between the candidate words and other candidate words in the extracted text; calculating related weights between the candidate words in the analysis statements and other candidate words in the analysisstatements; creating weighted undirected graphs of the analysis statements; based on a text sorting algorithm, calculating vertex scores of the vertexes of the weighted undirected graphs; and according to the vertex scores, sorting the candidate words corresponding to the vertexes, and extracting keywords in the candidate words. The invention furthermore provides a keyword extraction apparatus. The related weight between the two candidate words serves as a weight edge in the text sorting algorithm, so that the calculation accuracy of the text sorting algorithm is improved and the keyword extraction accuracy is improved.

The embodiment of the invention provides a method and a device for extracting a water body of a satellite remote sensing image. The method for extracting the water body of the satellite remote sensing image comprises the following steps: obtaining satellite remote sensing image data and utilizing a preset threshold in a preset decision-making tree to classify the water body information in the satellite remote sensing image data, thereby obtaining a water body extracting result after being classified. The device for extracting the water body of the satellite remote sensing image comprises an acquiring unit and a decision-making tree unit, wherein the acquiring unit is used for acquiring the satellite remote sensing image data, and the decision-making tree unit is used for utilizing the preset threshold in the preset decision-making tree to classify the water body information in the satellite remote sensing image data, thereby obtaining the water body extracting result after being classified. The method and device provided by the embodiment of the invention have the technical effect that the water body extraction precision of the satellite remote sensing image is increased.

The invention discloses a method for improving the location precision of micro seismic sources for mining. Through wave velocity zoning, a wave velocity model is more in line with the velocity structure characteristics of the complex rock masses of mines in a micro seismic monitoring area (such as a mine roadway, a goaf, an ore body or a surrounding rock), the extraction precision of the actual arrival time of seismic wave is higher, and thus, the precision of seismic source inversion and location is improved. In the process of seismic source inversion and location, an initial iteration seismic source parameter depth threshold is set through the layout and monitoring range of the space envelope of wave detectors, and a weighting factor is introduced during inversion according to the waveform quality. Therefore, the location precision of seismic sources is improved.

The invention discloses a banknote image inclination correction and region extraction method, which is characterized by comprising the steps of carrying out fuzzy enhancement on an acquired banknote image, and enhancing a contrast ratio between a banknote and a banknote background color or pattern; segmenting the banknote image, wherein the banknote image is segmented into a banknote background region and a banknote region; scanning to acquire edge pixels of an image of the banknote region; acquiring image edges of the banknote region by adopting least square method straight line fitting if the banknote edges are uneven or zigzag-shaped, calculating fitting errors of the image edges of the banknote region, and calculating an inclination angle of the image of the banknote region according to the image edge, whose fitting error is minimum, of the banknote region; carrying out inclination correction on the image of the banknote region according to the inclination angle of the image of the banknote region, and rotating the inclined image of the banknote region into a horizontal image; and extraction the banknote region. According to the invention, fuzzy enhancement is carried out on the banknote image, the banknote edges are acquired by least square method straight line fitting, the accuracy in inclination correction for the banknote image is improved, and the extraction effect of the banknote region is good.

The invention discloses a high-precision binocular camera calibration method, which is completed by using a pair of camera equipment and a calibration plate of a circular characteristic image, and comprises the following steps of: firstly, extracting control point image coordinates in the calibration plate to carry out camera primary calibration; Constructing a front view image of the calibrationboard by using the obtained camera parameters, solving control point image coordinates in the front view image by using a digital image correlation method, correcting the image coordinates in the calibration board, and performing camera calibration again; And finally, establishing a binocular camera overall optimization model according to the fixed position relationship between the cameras, and optimizing camera parameters. According to the calibration method provided by the invention, the coordinates of the control points in the calibration plate are accurately obtained, the optimization model of the calibration parameters of the stereo camera is simplified, and a high-precision calibration result can be obtained.

The present invention discloses a method for detecting and locating insulator defects in power transmission line image via a drone, and belongs to the technical field of image extraction and recognition. The detection and location method for the insulator defects comprises the steps of: S1. converting an aerial image from an RGB color space into an HSI (Hue-Saturation-Lightness) space, and carrying out binarization on extracted H component image and S component image separately to obtain a binary image of a preliminary contour of an insulator string; S2. using an ant colony algorithm based on a particle swarm optimization parameter to extract a contour of an insulator monomer; and S3. using a least square method to perform ellipse fitting on the contour of the insulator monomer, and locating the insulator defects by detecting the space between the contours of insulator monomers. By using the technical scheme provided by the invention, the detection precision of the insulator defects can be significantly improved, background interferences are reduced, and the operation speed is relatively high.

The invention discloses a cross-language multi-source vertical domain knowledge graph construction method, and relates to the technical field of knowledge engineering. According to the technical scheme, the method comprises the steps that vertical domain translation completes parallel corpus construction through content and link analysis according to input cross-language texts, domain dictionaries, domain term libraries, domain materials and data, and automatic translation of foreign language texts is achieved based on a trained translation model on the basis of preprocessing; domain knowledgepre-annotation training realizes active learning annotation based on text word segmentation and text clustering, to-be-annotated corpus screening based on an analysis topic is completed, and a confirmed service annotation data set is generated; an optimal algorithm is selected, and semantic feature extraction and entity relationship extraction are completed based on deep learning in combination with vertical domain translation data and an actual scene; and domain knowledge fusion and disambiguation are performed on knowledge from different sources through network equivalent entity combination, so that the cross-language multi-source vertical domain knowledge graph can be obtained.

An evaluation pair extraction unit analyzes a document which is present on a network to extract an evaluation pair composed of an object and an evaluation expression, and registers it in an evaluation pair database. An evaluation pair analysis unit executes analysis such as category estimation, specific evaluation expression extraction, and comparative analysis by utilizing the evaluation pair of the database. The evaluation pair is data in which the object and the evaluation expression are combined; wherein the object is a unique object including a commercial product name, maker name, or the like which can be evaluated, and the evaluation expression of the evaluation pair is a phrase representing evaluation including like, hard to use, or the like. An evaluation pair extraction unit gives an evaluation pair extracted from a model document to a machine leaner to cause it to learn a rule which can serve as an evaluation pair, and, after the learning, correctly combines an object with an evaluation expression from an unknown document to extract an evaluation pair.