1842results about How to "Improve matching accuracy" patented technology

A World Wide Web-based melody retrieval system takes a sung melody as a query and retrieves the song's title or other information from a music database over a WWW network which comprises a method of obtaining search clues with the maximum quantity of information from pitch and span (dynamic threshold determination) and a method of effectively reducing the number of answer candidates (coarse-to-fine matching), thus increasing the matching accuracy, and it is characterized in that a user can retrieve music or media with music by singing.

The invention discloses a method for acquiring a range image by stereo matching of multi-aperture photographing based on color segmentation. The method comprises the following steps of: (1) carrying out image correction on all input images; (2) carrying out the color segmentation on a reference image, and extracting an area with consistent color in the reference image; (3) respectively carrying out local window matching on the multiple input images to obtain multiple parallax images; (4) removing mismatched points which are generated during the matching by applying a bilateral matching strategy; (5) synthesizing the multiple parallax images into a parallax image, and filling parallax information of the mismatched points; (6) carrying out post-treatment optimization on the parallax image to obtain a dense parallax image; and (7) converting the parallax image into a range image according to the relationship between parallax and depth. By acquiring range information from multiple viewpoint images and utilizing image information provided by the multiple viewpoint images, the invention can not only solve the problem of mismatching brought by periodic repetitive textural features and shelter and the like in the images, but also can improve matching precision and obtain an accurate range image.

An authentication system incorporates pluralilty of fused biometric modalities which are combined with quality indicia of the biometric samples. Fusion processing maps individual modality data along with one or more quality measurements to a fused score for purposes of making an authentication decision.

The invention discloses a system for visually detecting workpiece appearance defects based on image processing, comprising an industrial personal computer, a coaxial light source, a CCD (charge coupled device) industrial camera, an image acquisition card and a rejecting mechanism; both the coaxial light source and the rejecting mechanism are connected with the industrial personal computer; the CCD industrial camera is connected with the industrial personal computer through the image acquisition card; wherein the coaxial light source is used for providing a diffuse reflection light source for a workpiece to be detected; the CCD industrial camera is used for taking an image of the workpiece at a detection station; the rejecting mechanism is used for rejecting from a production line, workpieces detected to be defective; the industrial personal computer has a defect detection module based on image processing; the system is high in detection efficiency and easy to implement.

Disclosed herein an information processing device that compares an input image with a model image, the device including: a storage; an object feature point extractor; and a feature comparator.

The invention discloses a three-dimensional reconstruction method based on coding structured light, comprising the following steps: 1) projecting structured light to an object to be measured, and capturing an image modulated by the object to be measured by a camera; 2) matching an optical template, comprising: (2.1) positioning the optical strip boundary, scanning along each row of the image, determining a pixel point with strong gray variation as a candidate marginal point, and searching a local domain; and (2.2) matching the optical strip: adopting a color cluster method to build a color matching proper vector, comparing image color with a projected color, and defining Euclidean distance between the color proper vector and the cluster center to distribute the colors of red, green, blue and white of the candidate optical strip; and 3) using a calibrated system parameter for three-dimensional reconstruction of the object to be measured, determining the relation between a space point coordinate and the image coordinate point thereof by the calibrated conversion matrix parameter; and restoring three-dimensional spatial coordinate from the image coordinate of a feature point. The invention can simplify calculation process and has high matching precision and high reconstruction precision.

The invention discloses a workpiece apparent defect detection method based on machine vision. Firstly, a visual system is used for guiding a robot, the pose of a target workpiece is precisely positioned according to a template matching algorithm based on a gray value, and then workpiece apparent defect detection is carried out and comprises the steps that firstly, a workpiece image is acquired and pretreated through median filtering; secondly, the target workpiece is divided through a global threshold value, and workpiece pose correction is carried out; thirdly, burr interference of the edge of the workpiece is removed through mathematical morphology open operation; fourthly, notches, material sticking, cracking, indentation, needle eyes, scratches and foaming apparent defects are detected. The method solves the problems that the artificial detection speed is slow, the efficiency is low and precision is poor; the problems that according to current vision detection, defect types are singular, the imaging quality is poor, and the false drop rate is high are solved, and the automation degree of precise workpiece production and product quality are improved.

The invention discloses a binocular stereoscopic vision matching method combining depth characteristics. The binocular stereoscopic vision matching method comprises: obtaining a depth characteristic pattern from left and right images through a convolutional neural network; calculating a truncation similarity measurement degree of pixel depth characteristics by taking the depth characteristics as the standard, and constructing a truncation matching cost function combining color, gradients and depth characteristics to obtain a matched cost volume; processing the matched cost volume by adopting a fixed window, a variable window and a self-adaptive weight polymerization or guide filtering method to obtain a cost volume polymerized by a matching cost; selecting an optimal parallax error of the cost volume by adopting WTA (Wireless Telephony Application) to obtain an initial parallax error pattern; then finding a shielding region by adopting a double-peak test, left-right consistency detection, sequence consistency detection or shielding constraint algorithm, and giving a shielding point to a parallax error value of a same-row point closest to the shielding point to obtain a parallax error pattern; and filtering the parallax error pattern by adopting a mean value or bilateral filter to obtain a final parallax error pattern. By adopting the binocular stereoscopic vision matching method combining the depth characteristics, the incorrect matching rate of three-dimensional matching can be effectively reduced, the images are smooth and image edges including edges of small objects are effectively kept.

The invention relates to an indoor positioning method of an intelligent mobile phone. The indoor positioning method is technically characterized by comprising the following steps of 1, in a walking process of a pedestrian, performing pedestrian dead reckoning, such as judging a moving state of the pedestrian, calculating the number of generated steps, calculating a walking distance and reckoning the direction change of the pedestrian through an acceleration sensor and a gyroscope which are arranged in the intelligent mobile phone, thus obtaining reckoned position information; and 2, constructing an indoor environment magnetic field map, performing magnetic field matching by a particle filtering algorithm according to information which is acquired by the indoor environment magnetic field map and a magnetic field sensor in real time, and correcting the reckoned position information so as to obtain a final position of the pedestrian. According to the indoor positioning method, the robustness of indoor positioning is improved from two aspects of a method for resisting various handheld mobile phones and a method for processing different step lengths; the matching precision is improved by improving the stability of a magnetic field matching judging and determining weight and reducing a system accumulation error, so that long-time accumulated errors are avoided; the system complexity is reduced by reducing the amount of external equipment.

The invention discloses a method and system for map matching of transportation vehicle GPS (Global Position System) data. In the invention, a map matching process of the transportation vehicle GPS data is completed by structuring high-efficiency map matching grid indexes and using a simplified and accurate matching calculation, so that the operation efficiency of matching of the transportation vehicle GPS data is greatly improved, and the processing efficiency is thousands times higher than that of the traditional map matching method; and the matching positioning of a vehicle in a complex road network is assisted and judged by tracking the historical matching road state of the vehicle and using the topological connectivity of the road network, so that the accuracy of the map matching of the transportation vehicle GPS data can be greatly improved.

The invention relates to a target detection system and method based on self-adaption combined wave filtering and multilevel detection. The system includes a target detection module combing a moving target detection unit with an obvious target detection unit, a target identification module based on a convolutional neural network and a target tracking module based on combination decision and multi-channel image features; the target detection module, the target identification module and the target tracking module tightly cooperate with one another to constitute the stable and reliable target detection system together. System output includes detected candidate target position information, target classification information and position information of selected targets obtained by target tracking. The target detection system is achieved on a high-performance multi-core DSP chip and used for carrying out targeted optimization on the multi-core DSP chip, real-time target detection and target tracking are achieved, and a function of quickly identifying a target is achieved. The target detection system and method have the advantages of being high in practicability and feasibility and convenient to integrate into various solution schemes with target detection demands, and can achieve intelligent target detection, identification and tracking.

The invention provides a deep text matching method and device based on word migration learning, and the method comprises the steps: firstly, carrying out the fusion of a BERT model and carrying out the pre-training of the BERT model during the training of a deep matching model; secondly, utilizing a pre-trained BERT model to respectively represent sentences in the input sentence pairs by using initial word vectors, and then performing similarity weighting on the sentences in the sentence pairs represented by the initial word vectors to obtain weighted sentence vectors; and finally, according to the loss value corresponding to the similarity value of the statement vector, adjusting model parameters, and carrying out text matching on the input statement by utilizing a depth matching model obtained through parameter adjustment. The parameters of the pre-trained BERT model are no longer randomly initialized parameters, and part-of-speech prediction is added into the pre-trained BERT model,so that the word vector semantic information is enriched. Therefore, semantics, represented by word vectors, of sentences in the sentence pairs are more accurate through the trained BERT model, and the matching accuracy of the trained model is promoted to be improved.

The invention discloses a stereoscopic vision-based real low-texture image reconstruction method, which is implemented by the following steps: (1) shooting images by two cameras at two proper angles at the same time, respectively, wherein one image is used as a reference image and the other image is used as a registered image; (2) calibrating the inside and outside parameter matrixes of the two cameras respectively; (3) performing epipolar line correction, image transformation and Gauss filtration according to the calibrated data; (4) calculating a self-adaptive polygonal prop window of each point in the two calibrated images, and calculating the matching of pixel points to obtain a parallax space diagram; (5) completing dense matching by executing a tree dynamic programming algorithm pixel by pixel in the whole diagram; (6) extracting error matching points according to a left and right consistency principle, and performing parallax correction to obtain a final parallax diagram; and (7) calculating the three-dimensional coordinates of actual object points according to the calibrated data and a matching relationship to construct the three-dimensional point cloud of an object.

The invention relates to a three-dimensional point cloud-based patrol robot vision system and a control method. According to the system, point cloud data of a patrol environment is collected by adopting an RGBD video camera, and a three-dimensional map of the patrol environment is created based on a point cloud fusion technology; obstacle avoidance and optimal path planning are performed based onan artificial potential field method; based on a convolutional neural network identification algorithm, three-dimensional features of objects are fused, a target object in the patrol environment is identified, and according to a mapping relationship between the target object and the camera, three-dimensional coordinates of the target object are accurately located; based on a wireless network system, real-time data obtained by a patrol robot is transmitted to a control terminal quickly in real time; and an operator can monitor or play back a patrol condition through the control terminal in realtime, and can control the robot to execute a patrol task through the terminal. According to the control method of the system, a working environment of the patrol robot is not influenced by change ofambient light, and the patrol task can be finished in a dark light condition.

The invention provides a finger vein identification method and system based on a convolutional neural network. The method comprises that a finger vein image which is registered originally is collected; the originally registered finger vein image is processed, and a finger vein ROI (Region of Interest) image is extracted from the finger vein image; a finger vein depth feature in a finger vein ROI is extracted via the convolutional neural network, and input to a support vector machine for classification, and further a finger vein enhanced image is obtained; binarization is carried out on the enhanced image to obtain a finger vein feature image; and a model is used to extract features of the pre-identified finger vein image, and the features are matched to realize identity authentication. According to the invention, the finger vein features can be extracted effectively, and the identification precision of the finger vein identification system is improved obviously.

The invention relates to an image detection method for detecting foreign matter at the bottom of a vehicle. The method sequentially includes the following steps of image input, image registration, binary image acquisition and foreign matter detection. The method can be used for fast recognizing whether the bottom of the vehicle going in or out of a barrier gate carries foreign matter or not, a vehicle chassis image provided by a vehicle chassis imaging system is subjected to signal analysis and processing, a real-time reliable foreign matter detection method is provided for detection of the chassis of the vehicle, and therefore the foreign matter at the bottom of the chassis of the vehicle can be fast, accurately and efficiently detected, the detecting accuracy is improved, interference of manual operation and environment changes is eradicated, and the requirement for real-time detection of the bottom of the vehicle in the security and prevention market in China is met. The time needed for recognition is no longer than five seconds, and the recognition efficiency is high.

The invention relates to a map matching method of floating car data. An optimal matching path is guaranteed and an alternative path set of each global positioning system (GPS) point is minimized at the same time by means of multistandard dynamic planning technology, and therefore the matching efficiency is improved. In addition, aiming at the characteristics of map matching, a traditional shortest path analytical method is improved, and the matching speed is further improved by reducing a mark initialization process and a searching method from multiple points to a single point. Compared with other current methods, the map matching method of the floating car data is better in matching accuracy and matching performance, can achieve real-time map matching processing of low sample frequency floating car data in a city traffic network, and has business popularization and application prospects.

The invention relates to a method and device for matching users in a game application. The method comprises the following steps that the historical primary attribute values of the users are acquired according to the historical game data of the users; the evaluating coefficients of current game running environment parameters of the users are acquired; processing is carried out according to the historical primary attribute values of the users and the evaluating coefficients of the current game running environment parameters, and the current primary attribute values of the users are obtained; the users with the current priamry attribute values meeting set conditions are matched as a group. According to the method and device for matching the users, the current game running environment parameters of the users are estimated, processing is carried out according to the evaluating coefficients of the environment parameters and the historical primary attribute values of the users, the current primary attribute values of the users are obtained, the users meeting the set conditions are matched, obtained matching is more accurate and reasonable, matching accuracy and matching efficiency are improved, the situation that a large number of users not meeting the set conditions are matched is avoided, and network resources are saved.

The invention discloses an image segmentation and adaptive weighting-based stereo matching method. The method comprises the steps of parallax initialization and parallax optimization. The parallax initialization comprises the steps of adopting corrected left and right images as a reference image and a target image respectively, and segmenting the image; based on the constructed combination segmentation information and an adaptive-weighting cost function, calculating matching costs E (p, pd, d) between a current to-be-matched pixel pint p in the left image and all candidate matching points pd in the right image, and selecting a candidate matching point of a minimum matching cost as an optimal matching point for the pixel pint p; repeating the above steps till all pixel points in the left image are traversed in the raster scan order. In this way, an initial parallax image is obtained. The parallax optimization comprises the steps of fitting the parallax plane of the obtained initial parallax image, suppressing the abnormity of the initial parallax image, and recovering the edge of the initial parallax image. The method is advantaged in that unreliable points in the initial parallax image obtained based on the calculation of the optimal matching point are re-corrected, while abnormal small areas are merged to adjacent normal areas. Meanwhile, edge pixels are recovered, and the parallax error is eliminated. The matching accuracy is improved.

The invention relates to an indoor positioning method based on fuzzy theory and multi-sensor fusion. The indoor positioning method is technically characterized by comprising the steps of according to multiple sensors arranged in an intelligent mobile phone, acquiring data of an acceleration sensor, a gyroscope and a barometer of the intelligent mobile phone firstly, and then the fuzzy theory is applied to the acquired data, speculating the motion state, step number and step length of a pedestrian so as to obtain the calculated position information; meanwhile, acquiring the user height from the information of the barometer; and establishing an indoor environment magnetic field map, acquiring information in real time by utilizing the indoor environment magnetic field map and magnetic field sensors, matching magnetic fields through the fuzzy theory and correcting the calculated position information to obtain the final position of the pedestrian. According to the method, the indoor positioning robustness is promoted through two aspects of multiple intelligent mobile phone carrying methods and processing of different step lengths. Through the matching of the fuzzy theory of the magnetic field, the accumulated errors can be reduced, the accuracy is improved, and the system complexity can be lowered.

The invention relates to a pedestrian re-identification method based on multi-region feature extraction and fusion, including the following steps: using a residual network to extract global features,and adding a pedestrian identity classification module for global feature extraction and optimization in the training stage; constructing a multi-region feature extraction sub-network for local feature extraction, and carrying out weighted fusion of local features; setting a loss function including the loss of the classification module and the loss of a feature fusion module; training the networkto obtain a model, and extracting the feature vectors of a query set and a test set; and in the measuring stage, re-measuring the feature distance by using a cross-nearest neighbor method. The methodis reasonable in design. Global and local feature are effectively combined. The distance measurement method is optimized. A good pedestrian re-identification result is obtained. The overall matching accuracy of the system is greatly improved.

The invention discloses a rolling bearing fault feature extraction method based on the signal sparse representation theory, and the method comprises the following steps: constructing an over-complete dictionary representing local damages of a rolling bearing through employing a multi-stage inherent frequency unit impulse response function; recognizing the multi-stage inherent frequency and damping ratio of the rolling bearing and a sensor system from a vibration response signal through a related filtering method, and obtaining an optimized dictionary; solving a sparse coefficient through employing a matching tracking algorithm, and improving the solving speed and precision through reasonable segmentation; reconstructing an impact response signal of each segment, and obtaining the sparse representation of a fault feature signal; carrying out time domain index statistic characteristic analysis of time intervals of adjacent impact response components in a sparse signal, and diagnosing the type of a fault through combining a mean value and a mean square deviation value. The method has the advantages of an analytical method and an adaptive method, improves the precision of waveform features, and can iron out the defects that a conventional method based on Fourier transform is not suitable for rotating speed fluctuation.

The invention provides a content filtering method and a device. The method comprises steps of respectively extracting key words from input rule conditions; dividing the rule conditions into one or more groups in accordance with extracted key words, and precompiling group matching data sets for the extracted key words; respectively precompiling accurate matching data sets for groups of rule conditions corresponding to the extracted key words; obtaining content to be filtered; conducting key word matching for the content to be filtered by the aid of the group matching data sets; conducting accurate matching of rule conditions for the content to be filtered by the aid of the accurate matching data sets of the groups of rule conditions which correspond to the matched key words; and implementing corresponding filtering strategies in accordance with matching results of accurate matching. Group prefiltering is conducted for rule conditions, so that the quantity of each of groups of rule conditions is little, and the occupied memory is reduced. The matching accuracy is high based on accurate matching of rule conditions after group prefiltering.

The invention discloses a splicing method for measuring image data with large forgings assisted by lasers, and belongs to the field of image processing and computer vision detection. The method comprises the steps that two different laser devices are used for constructing feature points on a metal surface, and the image data can be spliced under complex working conditions; two sets of adjacent binocular cameras carry out image data rough matching through light spots projected by point laser devices, meanwhile, a conversion matrix is obtained, linear laser devices are used for projecting grids, the rough matching conversion matrix is adopted as an initial value, image data fine matching is carried out, and splicing is achieved after redundant data are removed. The three light spots of different sizes are adopted as feature points in image data rough matching, error matching is effectively avoided, and matching precision is improved. The conversion matrix obtained through rough matching is adopted as the initial value in fine matching, and the computation quantity of correcting the conversion matrix is reduced. In the splicing process, a measurement system is taken as a multi-binocular measurement system, the method has no precise requirement for the position of a target, and implementation is easy.

The invention relates to an online community interaction method and device and an online community platform. The method comprises the following steps: digging and classifying a current question raised by a questioner on the online community platform to obtain classification information; according to the classification information, obtaining historical questions corresponding to classification; calculating a similarity between the historical questions and the current question, calculating to obtain the similarity between answers corresponding a preset quantity of historical question which exhibit the higher similarity with the question and the current question, and taking the answer corresponding to the historical question which exhibits the highest similarity as a smart recommendation answer; and pushing the smart recommendation answer to the questioner. The questioner can quickly obtain the answer of the question, and the question matching accuracy of an ask-answer community and user activeness are improved. In addition, the user can obtain the answer of the question through on-line help seeking so as to further improve question answering efficiency, interaction among users can be enhanced, and the user activeness of ask-answer community is further improved.