407results about How to "Improve match rate" patented technology

Steps after a user finds an attractive product until he or she buys it are successively performed so that he or she can buy it stressfreely. The user photographs an attractive product with a portable camera telephone terminal 1 and transmits image data 7 to a server 3 through Internet 2. The server 3 has a DB 4 in which images such as various types of products, persons, and logo marks are registered and a DB 5 in which related information of image data registered in the DB 4 is registered. The server 3 performs an image recognizing process for the image data 7, retrieves data that match the image data 7 with a high match rate from the image data registered in the DB 4, and extracts corresponding related information from the DB 5. The related information is provided to the user through the Internet 2. By only transmitting the image data 7, the user can obtain related information of an attractive product. When the user accesses a site 6 for the product corresponding to the related information, he or she can obtain detailed information and buy the product.

Ink strokes of cursive writing are segmented to make the cursive writing more like print writing, particularly with respect to the number of strokes of a character. A stroke-segmentation module first finds the local extrema points on a stroke of input ink. Then the local extrema points are stepped through, two (or three) at a time. The stroke-segmentation module may compare the three (or four) ink segments that are adjacent to the two (or three) local extrema points to a set of predefined stroke-segmentation patterns to find a closest matching pattern. Strokes are then segmented based on a stroke-segmentation rule that corresponds to the closest matching pattern. Additional stroke segmentation may be performed based on the change of curvature of the segmented ink strokes. Then, a character-recognition module performs character recognition processing by comparing the segmented ink strokes to prototype samples at least some of which have been similarly segmented.

Ink strokes of cursive writing are segmented to make the cursive writing more like print writing, particularly with respect to the number of strokes of a character. A stroke-segmentation module first finds the local extrema points on a stroke of input ink. Then the local extrema points are stepped through, two (or three) at a time. The stroke-segmentation module may compare the three (or four) ink segments that are adjacent to the two (or three) local extrema points to a set of predefined stroke-segmentation patterns to find a closest matching pattern. Strokes are then segmented based on a stroke-segmentation rule that corresponds to the closest matching pattern. Additional stroke segmentation may be performed based on the change of curvature of the segmented ink strokes. Then, a character-recognition module performs character recognition processing by comparing the segmented ink strokes to prototype samples at least some of which have been similarly segmented.

The invention discloses a method for quickly matching Chinese addresses in a multi-level manner on the basis of address feature words, and belongs to the field of data spatial research in the geographic information science. The method particularly includes two links of creating standard Chinese address matching dictionaries and matching the addresses. Chinese words of standard Chinese addresses are segmented on the word segmentation basis of the address feature words, the standard Chinese address matching dictionaries are created by the aid of double-array trie trees and Hash operation, database retrieval modes are replaced by bidirectional scanning and Hash operation, and geographic spatial coordinates of the Chinese addresses to be matched are acquired, so that the Chinese addresses can be matched. The method has the advantages that integral address matching procedures can be completed in memories of computers, and the address matching speed can be increased owing to bidirectional scanning and a simultaneous word segmentation and matching mode; the problem that partial Chinese addresses cannot be matched due to missing of address elements can be solved according to Chinese address classifying, layering and combining rules, and the address matching accuracy can be improved.

The invention discloses a method for sea-land vector map data integration and fusion, which includes the steps of data pre-treatment, data format conversion, data integration display, same-name entity matching, same-name entity combination and the like. A digital audio tape (DAT) data format put forward by the method enables map control information, index data and position coordinate data to be packaged on one file and stored with layers as the unit. The storage format is convenient for management of layers, and when data of a corresponding layer is read, the data file only needs to be opened once, so that data reading speed of data is greatly improved, and simultaneously the size of the map data file is reduced to a certain degree. In addition, the same-name surface entity matching method based on simplification of polygon is capable of replacing all point entities forming a surface entity with four points or eight points obtained through simplification to participate in a matching process, thereby being capable of reducing complexity, improving matching rate, and simultaneously being capable of storing simplified results to be in the mode of files to be convenient for secondary use.

The invention discloses a POS-aided method for matching aerial images, which comprises the following steps: firstly, utilizing an exterior orientation element obtained by the POS to construct a homonymous nucleofilament constraint equation and predict the initial parallax of an image; then, establishing an image pyramid according to the initial parallax and an approximate one-dimensional image correlation which carries out nucleofilament constraint layer by layer on the image of the pyramid; and finally adopting the matching of least-square images to confirm homonymous image points and pick mismatched points, thereby obtaining the homonymous image points of the images to be matched. The invention which adopts the POS-aided image matching method to automatically measure image points has the advantages that not only the application potential of the POS can be fully developed; but also the matching rate and the matching efficiency of automatically rotating points can be improved; and the problems that the rotating points of the images are so difficult to be matched that the rotating points are required to be measured manually and interactively are solved, for example, the rotational angles of certain images are too large, the image texture is not obvious, and the topographic relief is bigger.

There is provided a method of processing a digital image including: (a) obtaining a plurality of images; (b) converting the plurality of images into histograms; (c) setting one of the plurality of images as a reference image and another of the plurality of images as a comparison target image; (d) adjusting a distribution of the histogram of the reference image to match a distribution of the histogram of the comparison target image to produce an adjusted reference image; (e) comparing a difference between the adjusted reference image and the comparison target image to produce a masking image; (f) applying the masking image to the comparison target image to produce an adjusted comparison target image; and (g) combining the reference image and the adjusted comparison target image to produce a high dynamic range (HDR) image. Accordingly, even if there is a complex motion on a subject, a clear image without an image overlap or a ghost effect may be obtained when producing the HDR image.

The invention discloses a method for re-identifying vehicles in sequence images of a mentoring video. The method comprises the following steps of performing image feature extraction: for all shot video data, firstly detecting out vehicle images appearing under a camera, dividing the vehicle images of the monitoring video into a plurality of equilong vehicle image sequences according to the vehicles and the camera, calculating grayscale histograms of two channels Cr and Cb in a YCrCb space, and finally obtaining vehicle contour images by using a Gabor filter and original image convolution; performing a set-to-set-based metric learning method: training a step of learning a metric function of relative distances between vehicle target image sequences, namely, performing metric learning based on a thought of maximization of a probability that inner-class distances are shorter than between-class distances; and performing between-set metric learning-based vehicle re-identification. According to the method, the complexity of an algorithm is lowered while feature dimensions are reduced; and after between-set distance measure is added based on a re-identification effect, the identification rate of the algorithm is increased and the identification effect is improved.

The invention relates to the field of image processing and image identification, and particularly relates to a human motion local characteristic representation method and an application in behavior identification thereof. The human motion local characteristic representation method comprises the following steps that firstly kinetic energy of human bone joints, the position coordinates of the human bone joints, the direction change vectors of the human bone joints and posture potential energy of the human bone joints are extracted from the perspective of energy according to human behavior biological and kinematic characteristics so that a local characteristic combination matrix is constructed; and then dimension reduction is performed on the local characteristic matrix by utilizing K-means clustering and bag of word (BOW) characteristics are extracted, characteristic vectors used for behavior identification are formed through combination of human joint angle characteristics, and the characteristic vectors are applied to the field of human behavior identification so that the method has great effect through experimental verification.

The embodiment of the invention provides a leader sequence transmission method, device and system, and relates to the field of communications. According to the leader sequence transmission method, device and system, the intensity of leader sequence signals received by a base station can be enhanced, therefore, the matching probability of the base station is increased, and the efficiency at which a user has access to the base station is increased. The method includes the steps that system information broadcasted by the base station or a physical broadcast channel (PBCH) is obtained, wherein the system information or the PBCH comprises a preset physical random access channel (PRACH), the preset PRACH comprises preset m wireless frames, and m is a positive integer; it is determined that the repetition times of leader sequences used for random access at preset resource positions of the preset PRACH are n, wherein n is larger than or equal to 1 and smaller than or equal to 10 m; the leader sequences are sent at the n preset resource positions of the RRACH to the base station.

The invention discloses a police service cloud image recognition vehicle management and control system based on geographic space-time constraint. The system comprises an image data collection and decomposition module, an image retrieval module, a data storage module and a geographic information service module. The system is characterized in that the image data collection and decomposition module is responsible for collecting road images, extracting vehicle characteristics in the images and transmitting vehicle characteristic information in the images to the data storage module, the data storage module is used for storing the road images and image characteristic element data concerning the vehicle characteristic information in the images, the image retrieval module is used for responding to image retrieval requests of requesters, acquiring retrieval results meeting the image retrieval requests in an image retrieval mode mixing the image characteristic element data with geographic space-time constraint according to the image retrieval requests of the requesters and feeding the retrieval results back to the requesters, and the geographic information service module is used for responding to vehicle positioning requests of the requesters, acquiring vehicle space position information through a map server served by a geographic information platform and feeding the vehicle space position information back to the requesters.

The embodiment of the present invention discloses a picture recommending method, apparatus and system. According to the embodiment of the present invention, tags with same, close or related semanteme are clustered to obtain clustered tags; then when picture recommendation needs to be carried out, picture information of a current picture is acquired; according to the picture information, a corresponding clustered tag set is acquired and according to the clustered tag set, a picture is selected to generate a candidate pool; and then a similarity between the current picture and each picture in the candidate pool is calculated and on the basis of the similarity, picture recommendation is carried out. According to the scheme, a matching ratio can be greatly improved, a statistical effect is strengthened, and a picture recommending effect is improved integrally.

The invention relates to a double-channel grinding method of a double-row ball bearing. The method is characterized in that a centerless clamp is adjusted, wherein an axial positioning support of the centerless clamp is ground by using an equipment self-grinding mode, and the end face runout of the axial positioning support is controlled within 2 micrometers; the angle alpha of an equipment workpiece frame is adjusted according to the requirements of the sizes of a ground loop, comprising the diameters of channels De / di, the positions of the channels a / b, the channel spacing a1 / b1 and the curvature radiuses of the channels Re / Ri, and the angle adjustment range is between 0 degrees and 25 degrees; a control cabinet of a CNC digital control system is used for controlling the shape of a single-point trimming grinding wheel of an arc trimmer; the positioning surface of the ground loop is fixed on the centerless clamp, and the control cabinet of the CNC digital control system is adopted to control the feed travels of an equipment workpiece X axis and a grinding wheel grinding frame Z axis meanwhile; after locking, a rotating type electric main shaft and a rotating equipment workpiece main shaft drive the ground loop so as to ensure that the grinding wheel enters a first row of channels to be ground; and after the grinding of the first row of channels is finished, the grinding wheel is fed into the second row of channels through the CNC digital control system and locked to grind the second row of channels.

A social media identity / construct is created for a person by locating social media profiles of the person at least by one or both of i) recognizing image(s) in one or more social media postings related to the person and ii) correlating geolocation data embedded in one or more social media postings related to the person or the person's social media connections, analyzing, monitor social media outlets and channels to create a social media profile for the person.

The invention relates to an image copying and pasting detection method based on circular description operator CSIFT (Colored scale invariant feature transform). The image copying and pasting detection method comprises the following steps: 1) inputting a colorful image, and calculating the color invariant of the colorful image, and taking the gradient direction of the color invariant as a direction characteristic; 2) detecting a scale space extremum line, and positioning key pints; 3) determining the main directions of the key points by utilizing a circular description operator to generate the 24-dimensional operators of all key points; 4) taking Euclidean distance between the operators of every two key points as measurement for judging key point similarity, pairing all key points in the image, determining two suspicious areas in the image, and finding all paired key points in the two suspicious areas; and 5) if the amount of the paired key points is greater than a set threshold value, judging that the image is tampered by copying and pasting, and otherwise, judging that the image is not tampered by copying and pasting. The method is favorable for accurately and efficiently detecting whether the same image is tampered by copying and pasting.

The invention claims inspection method and device of an Internet deep packet. The method comprises the following steps of: sequentially filtering the contents of the Internet packets section by section through a Bloom filter, delivering suspected content parts to a matching unit for matching when suspected attacking packets are discovered; respectively carrying out parallel matching on suspected parts in the suspected attacking packets with a plurality of characteristic words in a TCAM of the matching unit, and acquiring the judgment result that whether the suspected parts are attacking packets, if the suspected parts are not attacking packets, filtering the rest contents section by section through the Bloom filter, and otherwise, discharging the packets and generating a warning. The invention improves the efficiency of characteristic word matching and greatly reduces the energy consumption via filtering the non-characteristic word contents of the non-attacking packets and the attacking packets through the Bloom filter, and only sending the suspected parts of the suspected attacking packets to the matching unit for carrying out matching inspection.

The invention discloses an auxiliary design method of virtual terminals on the basis of a general template and key character matching. The auxiliary design method comprises the following steps of: firstly manually associating the virtual terminals as the general template, then conforming the position of the virtual terminals in a file to be associated with the positions of the virtual terminals with the same function in the general template by key character matching, so that the automatic association of most virtual terminals with different types in different manufacturers is realized; and for fewer virtual terminals without automatic association, the manual association can be carried out after the search range is shrunk. The auxiliary design method disclosed by the invention has the advantages that design workload can be effectively reduced and the design efficiency and the design quality are improved.

The embodiment of the invention provides a voice recognition method and device. The voice recognition method comprises the steps that candidate word strings corresponding to voice signals are acquired; a field language model is scheduled, and the corresponding probability enhancement coefficient of the candidate word strings in the field language model is determined; a basic language model is scheduled, and the basic matching probability of the candidate word strings in the basic language model is determined; the probability enhancement coefficient is combined with the basic matching probability so that the total matching probability of the candidate word strings is acquired; and the probability of a word sequence formed by each candidate word string is calculated according to the total matching probability of each candidate word string, and language texts corresponding to the voice signals are generated according to the word sequence of the highest probability. According to the voice recognition method and device, the general speaking habits in the basic language model are comprehensively considered, and the specific wording habits of the field language model are also considered so that the higher recognition rate of the voice signals can be provided.

The present application relates to the field of computer technology, in particular to a pharmaceutical data processing method, apparatus, computer device and storage medium based on machine learning.The method comprises the following steps of: acquiring a plurality of drug data, wherein the drug data comprises an instruction manual, the instruction manual comprises a plurality of fields and corresponding text contents; extracting the text content corresponding to the preset fields in the instruction manual; obtaining a classifier, inputting the text content into the classifier, outputting a disease class corresponding to the drug data through the classifier, and adding the disease class to the drug data; obtaining a disease classification coding table, wherein the disease classification coding table comprises a disease category and corresponding coding data; the disease category of the drug data is matched with the disease classification coding table, and the corresponding coding datais added to the drug data. The method can effectively improve the matching rate of the drug data and the corresponding disease categories.

The invention relates to a battery grouping method. The method comprises the following steps: (1) discharging batteries to be grouped according to a preset multiple ratio so as to obtain a discharge curve of voltage changing; (2) selecting the voltage platform of each battery to be grouped, and calculating the average value of the voltage platform; (3) calculating the differences or quotients between each voltage platform value and the average value, and taking the calculated differences or quotients as the comparative values; (4) sequencing the comparative values from large to small or otherwise, individually carrying out subtraction calculation or division calculation on two neighbored comparative values, taking the calculation results as the deviation values, and when the absolute value of a deviation value is larger than a first preset differential voltage factor or less than a second present differential voltage factor, dividing the two batteries corresponding to the two neighbored comparative values into different groups.

The invention relates to a geocoding-free rapid image splicing method of a low-altitude unmanned plane. A key frame image is extracted from a video to carry out geocoding-free rapid image splicing to generate a panoramic image. The method comprises the following steps: extracting a key frame image from a video stream to obtain all to-be-spliced images; carrying out feature point extraction including establishing a multi-dimensional space unit, constructing a pyramid image, determining a key point position, and removing an unstable point on all to-be-spliced images and carrying out screening on the key point by using a SUSAN algorithm to obtain a final feature point; carrying out feature matching according to respective feature points of a corresponding reference image and a to-be-matched image; and carrying out smoothening on the to-be-spliced images by using a gain compensation method, carrying out weighted average fusion on spliced overlapped areas to eliminate seams and sawteeth, and carrying out splicing to generate a panoramic image. According to the technical scheme, the method can be applied to fields of emergency surveying and mapping, disaster prevention and mitigation, land monitoring, emergency monitoring, pipeline layout and inspection of the oil sector, and circuit layout and power line inspection of the power sector and the like.

The present application relates to a text perception based emotion feedback method and apparatus, a computer device and a storage medium. The method comprises: calculating obtained to-be-fed back textcontent by using a text content perceptor to obtain the first weight distribution ratio and the second weight distribution ratio; inputting the to-be-fed back text content into a trained neural network to obtain an emotion classification result; calculating emotion words in the to-be-fed back text content according to a preset rule in a preset dictionary to obtain a corresponding emotion index result; according to the first weight distribution ratio and the second weight distribution ratio, weighting the emotion classification result and the emotion index result to obtain an emotion analysisresult; obtaining an emotional opinion result corresponding to the to-be-fed back text content; and according to the emotion analysis result and the emotional opinion result, selecting a target emotional feedback comment reply corresponding to the to-be-fed back text content from a preset comment reply association library. By virtue of the technical scheme of the present application, the matchingrate between the to-be-fed back text content and the target emotional feedback reply can be improved.

The invention discloses a fabric property picture collection and recognition method based on deep learning. The method comprises the steps that multiple fabric property pictures are acquired, macro information and micro information of the fabric property pictures are collected, and a training set is generated; the training set is trained through a deep learning model; deep features which are trained through the deep learning model and contain global information and local information at the same time are extracted, and linear discriminant analysis is performed on the deep features to complete training of the deep learning model; and fabric recognition is performed through a cosine distance according to the trained deep learning model. Through the method, multiple fabric property recognition problems including weaving process problems, background color process problems, surface process problems, printing process problems, spinning process problems and the like are solved, meanwhile, the trained model contains the local information and the global information at the same time, and the accurate recognition rate and the matching rate of a local pattern and a global pattern of a fabric are increased. The invention furthermore discloses a fabric property picture collection and recognition system based on deep learning.

The invention discloses a matching method of a power battery pack. The power battery pack is obtained by sequentially connecting cells in parallel and in series. The matching method of the cells connected in parallel comprises: acquiring the capacity of each cell; performing cycle charge and discharge according to a charge-discharge rate required by the purpose, and recording charge and discharge curves; calculating a voltage-dependent electric quantity change rate dQ / dV curve of the discharge curve, wherein the dQ / dV curve has two peaks; searching for voltage values corresponding to the peaks according to the dQ / dV curve, finding electric quantity values of the corresponding cells in the discharge curve according to the voltage values, and calculating a difference Dn between the electric quantity values; acquiring the self-discharge rate of each cell; acquiring an internal resistance value of each cell; grading Dn according to the values into a plurality of large groups, and classifying the cells into the corresponding large groups according to the tested Dn values, wherein the cells in each large group are classified into a plurality of medium groups based on the value of the self-discharge rate, and the cells in each medium group are classified into a plurality of small groups according to the internal resistance values of the cells; selecting a plurality of cells in the same group for matching.

The invention discloses a real-time image splicing method for police unmanned aerial vehicle investigation and evidence obtaining. The method includes the three steps of improved ORB algorithm operation, image registration and image fusion. According to the method, a multi-scale space is constructed; a significance analysis model is adopted to obtain an optimal angle point detection threshold value, and feature points are extracted; an ORB descriptor is adopted to describe the feature points; and a Hamming distance and RANSAC-combined method is adopted to achieve fast matching. As indicated by results, an improved ORB algorithm maintains the superiority of speed, and the matching rate of the improved ORB algorithm is improved as for images on which scale, angle of view, rotation and illumination change occur.

The invention discloses a deep learning-based small-area fingerprint comparison method. The method comprises the following steps of 1) finding position and direction information of a feature point in a small-area fingerprint; 2) according to the obtained position and direction information of the detail feature point in the step 1), performing rotation normalization on an image by taking the feature point as a center and the direction of the feature point as an X axis, and capturing a small block B of a set size; 3) performing convolutional network training: adopting a deep residual error network for a network model of a convolutional neural network, and training a training sample by using a Caffe framework; 4) performing semantic feature extraction; 5) performing fingerprint registration: enabling a user to perform registration in combination with a corresponding instruction in a registration process, wherein a registration template is composed of a feature point union set of a registration fingerprint image; and 6) performing fingerprint comparison, wherein a comparison score is determined by a mean value of a plurality of maximum values of similarity in to-be-matched images and the registration template. The deep learning-based small-area fingerprint comparison method provided by the invention is effectively suitable for small-area fingerprint comparison and is high in reliability.

The invention provides a embedded link short message point-to-point identification method and system. The invention comprises: embedding a plug-in unit of handset software into handset, and converting the conventional point-to-point short message into identification short message; or, a end-to-end direct sending of point-to-point identification service technology user. A short message center gateway forwards short messages, a searching server analyzes, matches short messages, adds identification, and returns the text identification of point-to-point short message. The steps is: receiving a SMS information form a friend, analyzing the message, matching the message content according to the message identification keyword object base, implanting corresponding link and identification respectively, writing the message with implanted point-to-point identification into inbox and storing, and displaying to the user. The invention can increase inquiry result accuracy rate, improve matching speed and searching efficiency, and can represent the new technology development tendency.

A magnetic ink character reading apparatus, magnetic ink character reading method and program, and a POS terminal apparatus reduce recognition errors and thereby improve the read rate. The magnetic ink character reading apparatus reads a text string of magnetic ink characters using both a magnetic reading mechanism and optical reading mechanism to obtain magnetic ink character recognition (MICR) results and optical character recognition (OCR) results, which are compared. The OCR process is repeated if the results differ. The positions of the read character blocks are compared to find character blocks that are offset perpendicularly to the base line of the magnetic ink characters, and the OCR process is repeated. The area to which the OCR process is applied again is near the position of the offset character block corrected in the direction perpendicular to the line of magnetic ink characters to be in line with the character blocks for which the MICR result and OCR result were the same.

The invention relates to an integration method for depth feature and traditional feature based on AdaRank. The main technical characteristics comprise: dividing image data, respectively establishing aimed at different parts and training a depth convolution and a neural network, used to obtain depth features; extracting traditional features from pedestrian re-identification data, including LOMO features, ELF6 features, and Hog3D features; selecting the following three metric learning methods, KISSME, kLFDA, and LMNN; all the features and the three metric learning methods being combined and spanned to a Cartesian product, to obtain a series of weak sorters; using an AdaRank algorithm, performing ensemble learning on the weak sorters, to finally obtain a strong sorter. The method is reasonable in design, and combines depth learning, multi-feature, metric learning, and ensemble learning, and learns in an integrated manner through establishing the weak sorters, so integrated performance of a system is much better that using a single feature and a single metric learning, system integrated matching ratio is greatly improved, and good performance is obtained.