532results about How to "Realize automatic extraction" patented technology

The invention discloses a deep convolutional neural network end-to-end based image retrieval method by layered deep searching. The method is characterized by mainly including a convolutional neural network used for deep learning, training and recognition, a rapid visual segmentation algorithm for searching image objects, a quick comparison method used for rough searching with a Hash method and Hamming distance fast images, and a precise comparison method for first k ranked images based on images from a candidate pool P. By the method, automation and intelligence level in searching images with images can be effectively heightened, search results can be acquired accurately, and demand of image retrieval in the big data era is satisfied with less storage space and high retrieval speed.

The invention belongs to the technical field of image processing, in order to realize automatic extraction and segmentation of retinal blood vessels, improve the anti-interference ability to factors such as blood vessel shadow and tissue deformation, and make the average accuracy rate of blood vessel segmentation result higher. The invention relates to a retinal blood vessel image segmentation method based on a multi-scale feature convolutional neural network. Firstly, retinal images are pre-processed appropriately, including adaptive histogram equalization and gamma brightness adjustment. Atthe same time, aiming at the problem of less retinal image data, data amplification is carried out, the experiment image is clipped and divided into blocks, Secondly, through construction of a multi-scale retinal vascular segmentation network, the spatial pyramidal cavity pooling is introduced into the convolutional neural network of the encoder-decoder structure, and the parameters of the model are optimized independently through many iterations to realize the automatic segmentation process of the pixel-level retinal blood vessels and obtain the retinal blood vessel segmentation map. The invention is mainly applied to the design and manufacture of medical devices.

The invention discloses an age estimation method and device. The method comprises the steps of inputting a pre-processed to-be-tested human face image into a trained deep convolutional neural network,and acquiring an age estimation result of the to-be-tested human face image. The age estimation method and device have the beneficial effect of acquiring highly accurate age estimation result.

A method for picking up and analysing information automatically includes using XML daily record analysing module to enter automatic picking up procedure, compairing packed up daily record with safety rule to judge whether system operation is abnormal or not, recording biased byte value of each analysis, detecting variation of each target device or daily record file at fixed time to realize automatic analysis to increased part of daily record according to recorded state information of dialy record.

The invention discloses an automatically-extracting power line method in random laser point cloud data. According to the automatically-extracting power line method, a ground point cloud is segmented and filtered through a point cloud of an elevation; on direction projection of the segmented point cloud, a straight line segment is detected and further back-projected to three-dimensional space so as to obtain a three-dimensional line segment; while the three-dimensional line segment serves as a clustering nuclear, a distance clustering method is used to obtain an accurate point cloud of the power line and generates a power line vector node based on the accurate point cloud of the power line; and a power line vector is outputted, so that automatic extraction of the power line is realized. The automatically-extracting power line method in the random laser point cloud data is efficient in extracting, and can automatically extract the power line.

The invention provides a road zebra crossing automatic extraction method based on vehicle-mounted laser scanning point cloud, and relates to public traffic road zebra crossings. According to the method, global positioning system data for recording vehicle positions and tracks in real time is used for extracting a plurality of cross sections from the vehicle-mounted laser scanning point cloud data, and the road and non-road classification is realized through detecting the elevation mutation of road shoulders of the roads in the scanning line data; then, the three-dimension road data is converted into an intensity characteristic image with space distribution characteristics, the laser scanning point normal distribution characteristics are utilized for dynamically cutting the road zebra crossings, the GPS (global positioning system) track data is used again for calculating the linear morphology closed operation direction and size, and the extraction of the road zebra crossings is realized. Through the cross section subdivision on the vehicle-mounted moving scanning data, and the three-dimension road surface data detection is converted into the detection of the elevation mutation of the road shoulders of the roads in the two-dimension profile for realizing the road and non-road classification. Compared with a method of directly processing mass three-dimension data, the method has the advantages that the calculation quantity is small, and the efficiency is high.

The invention discloses an automatic extraction method of a conversation text topic, comprising preprocessing conversation text data and detecting question-answer pairs in the preprocessed conversation text; segmenting topics of the conversation text, clustering language chunk groups after topic segmentation and extracting a topic sentence from the clustered language chunk groups. The conversation text topic is more accurate by adopting the extraction method, enabling a user to search or retrieve conversation records of interest from the extracted topic sentence, thereby improving the user experience.

The invention relates to a three-dimensional human face reconstruction method based on stereoscopic vision of two eyes. Steps for implementing the method are as follows: (1) using two cameras to respectively take a picture of a human face from two different angles; (2) respectively determining an internal parameter matrix and an external parameter matrix of each camera; (3) carrying out extreme line correction and picture conversion according to the determined data; (4) utilizing a Harris corner detection operator to extract characteristics of the human face, and carrying out initial matching by utilizing a local template window method and extreme line restriction; (5) starting from an initial matching assemblage, and obtaining a sparse matching assemblage by utilizing a seed point growing algorithm; (6) taking the sparse matching assemblage as a guiding point and carrying out a dynamic planning algorithm so as to finish dense matching; and (7) calculating three-dimensional coordinates of actual physical points on the human face according to the determined data and the matching relationship, thereby reconstructing a three-dimensional point cloud of the human face.

The invention relates to an environment noise identification classification method based on a convolutional neural network. The method comprises the following steps of: S1, extracting natural environment noise, and editing the natural environment noise into noise segments with duration of 300ms to 30s and a converted frequency of 44.1kHz; S2, carrying out short time Fourier transformation on the noise segments, and converting a one-dimensional time-domain signal into a two-dimensional time-domain signal to obtain a sonagraph; S3, extracting a MFCC (Mel Frequency Cepstrum Coefficient) of the signal; S4, forming a training set with 80% of all the noise segments and forming a testing set with the residual 20% of all the noise segments; S5, carrying out noise classification by a convolutionalneural network model; and S6, training a classification model by the training set, and verifying accuracy of the model by the testing set so as to complete environment noise identification classification based on the convolutional neural network. According to the invention, the sound segments are input, sound feature information is extracted, an output is a classification result, and automatic extraction on the sound feature information can be implemented.

The invention relates to a deep convolutional neural network based SAR image sea ice sorting method. The method includes steps of S01, segmenting a prior sea ice SAR image; S02, performing data pre-treatment; S03, performing model training and establishing a model; S04, performing treatment on a to-be-sorted sea ice SAR image; S05, combining sorting results. The method is advantaged in that the deep convolutional neural network model constructed in the method can realize characteristic automatic extraction based on images and excessive manual intervention is not needed; the method is an end-to-end SAR image sea ice sorting method and can reach sea ice monitoring service level and meet requirements on real time performance for offshore workers; image characteristics can be extracted automatically depending on a large amount of mark samples by the model and reliance on specialist knowledge is not needed; convergence is accelerated by utilizing a random gradient descending method and model training condition can be judged according to a loss function and accuracy; problems of gradient vanish or gradient diffusion in network parameter optimization counter-propagation are solved by utilizing an normative approach.

The invention provides a method for registering a remote sensing image based on a local contour covariance matrix, which combines a corner feature, a local split image and a local edge contour together as a local feature to implement an extraction of control points, has the local contour covariance matrix introduced into a registering processing, uses the local contour covariance matrix to implement an automatic extraction of a rotating and scaling factor, and combines a technology of self adaptive selection of a local window to reduce an effect of the twiddle and scaling factor on the registering processing so as to improve the robustness of the registering processing, thereby implementing an accurate registering between remote sensing images in a mode of large rotation and scaling without human participation.

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 embodiment of the invention discloses an image labeling method, system, and device and a computer-readable storage medium. The method comprises: a training sample set and a to-be-labeled image into a convolutional neural network, wherein the training sample set employs a back propagation algorithm to minimize a cross entropy loss function in the convolutional neural network to adjust the weight of the convolutional neural network for training and the weight of the trained convolutional neural network is loaded again to extract a sample network feature set of the training sample set and a test network feature set of the to-be-labeled image; according to the sample network feature set, the test network feature set, and a label set, the probability of belong to each kind of label in the to-be-labeled image is calculated and thus a label probability set is generated; and according to the label probability set, labeling is carried out on the to-be-labeled image. On the basis of combination of the deep learning and label propagation algorithms, the high-level semantic features of images are extracted automatically, so that the efficiency and accuracy of image labeling are improved.

The invention discloses a data exchange platform. The data exchange platform comprises a data source adapter, a kettle platform and ETL tool, a task scheduling engine based on Quartz and a target-terminal adapter. Interfaces accessed into all business systems can be provided, automatic data extracting, data verification, data dissemination and the like are achieved through uniform interface specifications, and meanwhile data synchronism, historical data migration and the like are supported. For the following, data integration and uniform controlling of all external data sources are achieved through the data exchange platform, suitable decoupling between all the external data sources and all the business systems is carried out, and product life-cycle management of data is achieved; for thepreceding, processing and grouping of integrated data are achieved, flexible and rapid data services are provided, and therefore a powerful support is provided for achieving business collaboration anddecision supporting in the data aspect.

The invention discloses a robot welding path autonomous planning method based on 3D point cloud data. The robot welding path autonomous planning method comprises the following steps of acquiring the original three-dimensional cloud data of a workpiece weld profile, and preprocessing the original three-dimensional cloud data; according to the workpiece weld characteristics, constructing ruler CAD models with the same characteristics, and converting characteristic surface information of ruler CAD models as three-dimensional point cloud data; performing local registration operation on the three-dimensional point cloud data of the ruler CAD models and the three-dimensional point cloud data generated after the workpiece weld profile is preprocessed; according to the local registration operation, performing continuous characteristic searching, thereby obtaining complete workpiece weld information; extracting weld position and posture information marked in the complete workpiece weld information; and processing the extracted weld position and posture information, and planning the robot welding path. The self-adaptability of robot welding can be improved, and the workload of scene teachingand off-line programming of operating staff can be remarkably reduced.

The invention discloses an intelligent heart sound diagnostic system and method based on in-depth learning and relates to the fields of bio-signal processing, pattern recognition, big data and in-depth learning. The method comprises the following steps: 1) acquiring heart sound audio data by a user through heart sound acquisition equipment or intelligent wearable equipment; 2) transmitting the data to a cloud server through a network, and storing and archiving the heart sound audio data; 3) segmenting the heart sound data on the cloud server by adopting a heart sound segmentation algorithm based on a logistic regression-hidden semi-Markov model, and performing automatic characteristic extraction and classification on the segmented heart sound data by using a one-dimensional convolutional neural network; 4) feeding diagnostic results to the user through a network and storing the results on a cloud so as to be provided for related institutions and designated hospitals as clinical historyreference of the user; and 5) expanding the heart sound data of the user confirmed by a professional doctor serving as training data into a heart sound database of a cloud server, so that the diagnostic capability of the heart sound diagnostic system is continuously improved.

The invention discloses a newborn pain degree assessment method and system based on facial expression recognition, and the method comprises the steps: building a newborn pain facial expression image data set which comprises a preprocessed newborn facial expression image and a corresponding expression class label; constructing a DCNN (deep convolutional neural network) for estimating the pain degree of a newborn, employing a disclosed large-size data set with a label for the pre-training of a network, obtaining an initial weight parameter value, carrying out the fine tuning of the network through the facial expression image data set, and obtaining a trained network model; inputting a to-be-tested newborn facial image into the trained network for facial expression classification and recognition, and obtaining a pain degree assessment result. The method can make the most of the features extracted through the DCNN, can obtain a better pain degree assessment result through a small-size newborn pain degree facial expression image data set, and is a new method for the development of a system for automatic assessment of the pain degree of the newborn based on the facial expression recognition.

The invention relates to the technical field of image recognition, in particular to a method for identifying lodging regions of wheat in multiple growth periods based on transfer learning, and the method comprises the following steps: A, shooting an RGB image and / or a multispectral image of a to-be-recognized wheat field; b, splicing and cutting the images to obtain a complete image of the wheat field to be identified; and C, importing the complete image of the wheat field to be identified into the trained DeepLabv 3 + model to identify the lodging area. The method is based on a DeepLabv 3 + network model. two methods are constructed by adopting a transfer learning mode to realize extraction of lodging regions of wheat in multiple growth periods; based on unmanned aerial vehicle images anda transfer learning method, lodging wheat characteristics in multiple periods can be effectively obtained, high-precision wheat area automatic extraction is achieved, it is possible to accurately detect a wheat lodging area, and powerful data support is provided for researching wheat lodging influence factors. The method is little affected by the environment and convenient to implement, and afterthe DeepLabv 3 + model is trained, the lodging area can be automatically recognized only by shooting the image of the wheat field to be recognized and importing the image into the model.

The invention relates to a method and a system for identifying a form in a layout file and belongs to the technical field of mode identification in the field of computer information processing. The conventional mode identification technology cannot effectively identify and automatically extract the form of a layout. In the method and the system, firstly, independent characters of the layout are combined and organized into content blocks by utilizing automatic combination technology; and secondly, form identification and content combination are performed according to spatial positions, character information and typesetting information of the content blocks. Through the method and the system, the form can be rapidly identified at high efficiency, and the form content is accurately organized through the analysis of the position and the typesetting information of the content of the paper layout.

The invention discloses an improved image splicing key frame fast extraction method, and relates to the field of computer vision. The method eliminates the influence of image distortion through image preprocessing, detects feature points by using the SIFT, realizes the automatic extraction of key frames by taking a matching ratio of the feature points of adjacent frames as the similarity measure of images, splices the key frame images, greatly reduces the splicing time of the sequence images, screens initial matching pairs by using an improved RANSAC algorithm, calculates an accurate transformation matrix between the images to achieve image registration, and finally achieves seamless image splicing by using a weighted smoothing fusion method. The method has strong robustness, and the spliced images are natural and clear.

The invention discloses a video region-of-interest extraction method based on visual perception characteristics and encoding information, and relates to the field of video encoding. The video region-of-interest extraction method comprises the following steps of (1) extracting luminance information of a current encoding macro-block from a primary video stream, (2) identifying a space domain visual characteristic saliency region through an inter-frame prediction mode type of the current encoding macro-block, (3) using a mean motion vector, in the horizontal direction, of a previous encoding macro-block and a mean motion vector, in the perpendicular direction, of the previous encoding macro-block as dual dynamic thresholds, identifying a time domain visual characteristic saliency region according to the result of comparison among a motion vector, in the horizontal direction, of the current encoding macro-block, a motion vector, in the perpendicular direction, of the current encoding macro-block and the dual dynamic thresholds, and (4) defining a video interest priority through combination of the identification result of the space domain visual characteristic saliency region and the identification result of the time domain visual characteristic saliency region, and achieving automatic extraction of a region of interest of a video. According to the video region-of-interest extraction method, the important encoding basis can be provided for the video encoding technology based on the ROI.

The invention relates to a panorama three-dimensional laser sensor data calibration method and apparatus. The apparatus includes a three-dimensional laser sensor and a calibration device. The method includes the following steps: firstly, dividing scan data of a two-dimensional laser sensor into two parts on average on the basis of the positive portion and the negative portion of a yL axis, then forming a laser scan plane, and forming two pieces of space three-dimensional point cloud data by the positive and negative laser data of the yL axis when a rotary holder drives the two-dimensional laser sensor to rotate 360 degrees; secondly, extracting laser data characteristic points needed by calibration from the two pieces of space three-dimensional point cloud data; and thirdly, optimizing distances among the laser points in the two pieces of space three-dimensional point cloud data, and then obtaining a calibration parameter. The problems of uncontrollable human errors and long time caused by hand-operated click matching are solved, factors that are likely to affect the data precision are fully considered, the calibration precision of the sensor is improved, and the universality of the calibration method is improved.

The invention relates to a system and method to filter the wasted mails on internet. At present there is a lot of wasted mails, human use keywords to filter it, build black name list, and set regulation to prevent, but the result is unobvious, hard to maintain, high requirement for professional skill. The invention consists of hardware, software which is installed on hardware and supported network, the software includes Linux operating system, analyzing engine for wasted mails, extended black name list, internet library and management interface. The periphery supporting network includes complain device for wasted mails, analyzing system for wasted mails, which can work as transparent gateway before LAN or mail sever, take over MX interface to work before mail sever. The method includes the hardware connection, telnet management, management of wasted mails and anti-waste setting, and the operation is convenient, easy to maintain, high identifying rate, high stability, low misjudgment rate, and plug and play.

The invention provides an automatic extraction method for power transmission lines based on airborne laser point cloud data. Along with the development and application of laser radar technologies, airborne laser radar is also gradually introduced into the inspection for power transmission lines, the laser radar generates a lot of point cloud data after scanning a power transmission line corridor,and extraction for the power transmission lines and power transmission towers from the point cloud data is a basis for subsequent data analysis. However, the traditional manual classification method is great in workload and very time-consuming. The automatic extraction method provided by the invention for power transmission lines based on airborne laser point cloud data eliminates ground points byadopting adaptive TIN, filters irrelevant point data by adopting an eight neighborhood method and various filters based on the height and the density and retains points of the power transmission lines and towers. Finally, the power transmission lines are fit by a general linear equation, and the power transmission towers are extracted by using a shaped-based filter.

The invention discloses an electric car charging station harmonic wave detection method in a microgrid in the electric engineering microgrid technical field. First, harmonic current in the microgrid are pretreated and is broken down into narrow band frequency signal families S1. Next, continuation treatment is carried out for the narrow band frequency signal families S1 through adoption of a self-adaption step-be-step continuation method to obtain frequency signal families S. Further, empirical mode decomposition is carried out for the frequency signal families S to obtain natural mode components. Hibert conversion is carried out for the natural mode components, and finally instantaneous frequency signals and instantaneous amplitude signals of the components are obtained. The electric car charging station harmonic wave detection method not only is suitable for analyzing nonlinear and nonstationary signals in power quality of the microgrid, but also can better reflect frequency characteristics and amplitude characteristics of the signals by analyzing linear and stationary signals in the power quality of the microgrid than other time-frequency analysis methods, and has more specific physical significance.

The invention is suitable for the field of mobile terminals and provides a method and a system for managing an address book, and a mobile terminal. The method comprises the following steps: acquiring communication information; respectively extracting candidate name information and candidate attribute information in the communication information according to a preset recognition system; establishing a corresponding relation of the candidate attribute information and the candidate name information; adopting the candidate name information as contact person name information, adopting the candidate attribute information as the contact person attribute information and storing the contact person name information and the contact person attribute information corresponding to the contact person name information into the address book. The invention realizes the automatic extraction of the contact person name information when a mobile terminal user manages the address book, improves the management efficiency of the address book, and brings convenience to the user.

The invention discloses an atmospheric correction method for multispectral data of a remote sensing satellite. The method comprises the following steps of: (1) reading data, and converting the data into apparent radiance data and apparent reflectivity data; (2) calculating decision factors, i.e. a ratio vegetation index, a soil adjustable vegetation index and a normalization water body index in adecision tree method according to the apparent reflectivity data; (3) acquiring apparent reflectivity data in a dark target area according to the apparent reflectivity data, and calculating air path reflectivity; (4) determining the aerosol optical depth of each multispectral waveband according to the relation among the air path reflectivity, a Rayleigh reflectivity ratio and the aerosol optical depth, and determining the total optical depth of a position of 550 nanometers according to the aerosol optical depth and the Rayleigh scatting optical depth of each multispectral waveband; and (5) acquiring an inversion parameter from a lookup table according to the total optical depth of the position of 550 nanometers, and solving land surface reflectivity through a radiative transfer equation to finish atmospheric correction.

The invention discloses a pedestrian detection method based on double-layer spatio-temporal context information. The method includes the steps: firstly, performing dimension estimation for an original image to obtain an interested area; then, extracting the double-layer spatio-temporal context information by extracting characteristic-layer spatio-temporal context, instance-layer spatio-temporal context and instance-layer timing sequence context so as to extract apparent characteristics to construct a basic pedestrian detector; and finally, combining the double-layer spatio-temporal context information with the apparent characteristics by the aid of a spatio-temporal context model. On the basis of the basic pedestrian detector based on the apparent characteristics, the double-layer spatio-temporal context information relevant to pedestrian detection is extracted automatically, and the apparent characteristics and the double-layer spatio-temporal context information are combined by the aid of the spatio-temporal context model. The double-layer spatio-temporal context information is introduced, influences of a complicated background and local blocking on detection performance are effectively overcome, and pedestrian detection recall rate is increased while pedestrian detection precision is improved.

The invention relates to the field of identification of vehicles, in particular to a taxi dispatching method and a taxi dispatching system on the basis of a video monitoring system. Vehicle information can be automatically extracted by using videos provided by monitoring probes at intersections and the like and applying an intelligent video analysis technology on the basis of the existing video monitoring system which almost covers all traffic road conditions and parking lots of an urban area, position information of taxis of a whole city is marked on a map by networking of the monitored videos to gather information of a plurality of cameras, and regions with probable demands on the taxis are discovered on the basis. The taxi dispatching method and the taxi dispatching system aim to overcome the shortcoming that information of various taxi dispatching companies on the basis of a GPS (global positioning system) is not shared, comprehensiveness of the information is guaranteed by the internet of vehicles on the basis of video monitoring, and a feasible solution is provided for intelligent dispatching, traffic guidance and the like by the internet of vehicles.