115results about How to "Reduce false match rate" patented technology

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 discloses a scattered workpiece recognition and positioning method based on point cloud processing, and the method is used for solving a problem of posture estimation of scattered workpeics in a random box grabbing process. The method comprises two parts: offline template library building and online feature registration. A template point cloud data set and a scene point cloud are obtained through a 3D point cloud obtaining system. The feature information, extracted in an offline state, of a template point cloud can be used for the preprocessing, segmentation and registration of the scene point cloud, thereby improving the operation speed of an algorithm. The point cloud registration is divided into two stages: initial registration and precise registration. A feature descriptor which integrates the geometrical characteristics and statistical characteristics is proposed at the stage of initial registration, thereby achieving the uniqueness description of the features of a key point. Points which are the most similar to the feature description of feature points are searched from a template library as corresponding points, thereby obtaining a corresponding point set, andachieving the calculation of an initial conversion matrix. At the stage of precise registration, the geometrical constraints are added for achieving the selection of the corresponding points, therebyreducing the number of iteration times of the precise registration, and reducing the probability that the algorithm falls into the local optimum.

The invention discloses an image stitching method based on overlapping region scale-invariant feather transform (SIFT) feature points and belongs to the technical field of image processing. Aiming at the problems that the algorithm computation is large and subsequent matching error and computing redundancy are easily caused due to the non-overlapping region features because of extraction of the features of the whole image in the conventional image stitching algorithm based on features, the invention provides an image stitching method based on the overlapping region SIFT feature points. According to the method, only the feature points in the image overlapping region are extracted, the number of the feature points is reduced, and the algorithm computation is greatly reduced; and moreover, the feature points are represented by employing an improved SIFT feature vector extraction method, the computation during feature point matching is further reduced, and the mismatching rate is reduced. The invention also discloses an image stitching method with optical imaging difference, wherein the image stitching method comprises the following steps of: converting two images with optical imaging difference to be stitched to a cylindrical coordinate space by utilizing projection transformation, and stitching the images by using the image stitching method based on the overlapping region SIFT feature points.

The embodiment of the invention discloses a method for processing matched pairs of characteristic points of images. The method comprises the following steps of: extracting the characteristic points of the images to be matched and acquiring invariables of the characteristic points; extracting significant areas of the images to be matched and acquiring the matched pairs of the significant areas among the images to be matched; and according to the invariables of the characteristic points, acquiring the matched pairs of the characteristic points among the matched pairs of the significant areas as the matched pairs of the characteristic points among the images to be matched. The embodiment of the invention also provides an image retrieval method and image retrieval equipment. In the embodiment of the invention, a space constraint relationship among the characteristic points, namely the significant areas where the characteristic points are positioned, is fully utilized, so the error matching rate of the matched pairs of the characteristic points can be reduced, and the acquisition accuracy of the matched pairs of the characteristic points is improved.

A mobile robot positioning method based on stereoscopic vision specifically comprises the following steps: shooting an image sequence through a binocular camera mounted on a robot; extracting scale invariant features in the stereo image sequence, matching and tracking the features through a circle matching technique; solving rotating translation matrixes among adjacent frames through a unit quaternion method; performing iterative optimization for multiple times through a minimum re-projection error method to obtain an optimal motion parameter. According to the invention, for the condition that the GPS and other traditional positioning methods are in failure in a certain environment, the positioning method based on stereoscopic vision is provided; for the problem that wrong matching easily appears during feature matching, the circle matching technique is provided to match and track the features, so that the accuracy of feature matching is improved; since the unit quaternion method is adopted to solve the rotating translation matrixes, computational accuracy is improved; the minimum re-projection error method is provided finally to obtain the relatively accurate motion parameter by performing iterative optimization for multiple times.

The invention discloses a semi-automatic image cutting method based on nuclear transfer, which mainly solves the problems that the existing interactive image cutting method can not keep data consistency, the cutting result is not stable, and edges are not clear. The method is realized through the following steps: inputting an image to be cut, obtaining a superpixel set of the input image through a mean shift method, calculating a superpixel color column diagram feature set, building a similarity matrix W through a Bhattacharyya coefficient formula, performing interactive operation on the precut image by a user to obtain a seed superpixel set, building a must-link constrain set M and a cannot-link constrain set C, transferring constrain information M and C to the whole nuclear space through a nuclear transfer method, so as to obtain a nuclear matrix R, clustering the nuclear matrix R through a k mean clustering method, so as to obtain a clustering mark vector, and outputting and displaying a cutting result. The method has the advantages of good data consistency, stable cutting result, simplicity of operation and clear edges, and can be applied to image retrieval, the technology of converting 2D to 3D, target detection and tracking, and medical image analysis.

The invention discloses a binocular three-dimensional reconstruction method comprising the following steps of: 1) acquiring an image of an object to be reconstructed by two image acquisition apparatuses with the same model in order to obtain a left image and a right image; 2) calibrating the image acquisition apparatuses by a chessboard method, computing the internal and external parameter and a lens distortion coefficient, processing the left image and the right image to remove distortion in the images; 3) extracting features from the two processed images in the step 2) to obtain the feature points of the two images; 4) matching the feature points by using the feature points of the two images in the step 3) to obtain feature point pairs; 5) performing epipolar geometric constraint detection on the obtained feature point pairs to remove mismatched feature point pairs; and 6) computing three-dimensional coordinates of corresponding points, in a world coordinate system, of the obtained feature points by using the reserved feature point pairs. The binocular three-dimensional reconstruction method may reduce abnormal points and acquires an accurate three-dimensional reconstruction model.

The embodiment of the invention discloses a team battle matching method and server, and relates to the field of the Internet and games. The mismatching rate of teams can be reduced and the matching efficiency of team battles can be improved. The method comprises the following steps: firstly, configuring teams with the same team people numbers in all teams which participate in battle application to the same queues; then selecting a random team from the queue in which the number of team people configured is the maximum; judging whether the team people number of the selected team is equal to a preset number; if the team people number of the selected team is equal to the preset number, configuring the team to carry out battle.

A binocular three-dimensional reconstruction method includes the following steps: step A, image collection: using two color cameras with the same model and arranging the two color cameras horizontally in a parallel mode, and enabling an target object to be arranged within the collecting range of the two cameras, wherein a image collected by a left camera is an original left image and a image collected by a right camera is an original right image; step B, camera calibration, and image rectification and pretreatment: adopting a chessboard method to demarcate and obtain inner and outer diameters of the two cameras, and then rectifying the original left image and the original right image, and next, pretreating the images; step C, hierarchical matching: conducting hierarchical feature extraction and matching to the images according to strong and week of feature lines of the images; step D, three-dimensional reconstruction: after completing the delamination which reaches required precision, according to a parallax principle, enabling obtained matching information of the two images to be calculated into three-dimensional information of the target object, adopting a method of plane interpolation, and then completing the three-dimensional reconstruction. The method has the advantages of being low in algorithm complexity and capable of achieving the three-dimensional reconstruction well and the like.

The invention discloses a stereo matching algorithm for computer vision, which is implemented based on a fully connected neural network and Edge-aware Disparity Propagation (EDP). The stereo matchingalgorithm comprises the steps of: firstly, calculating the initial matching cost of an input stereo image pair through a fully connected neural network, and using the WTA algorithm to find a corresponding disparity value, and generating an initial disparity map; performing consistency detection, and using the EDP algorithm and geodetic distance filtering to reconstruct the matching cost of a disparity inconsistent region, obtaining a new disparity value of the inconsistent region, and filling the new disparity value into a hole disparity map to obtain a complete disparity map; and finally, generating a final disparity map through sub-pixel enhancement optimization. Experimental values show that the algorithm can effectively reduce the mismatch rate of stereo matching and improve the accuracy of the disparity map, has lower mismatch rate especially in a non-occlusion region.

The invention relates to a laser-based amphibious three-dimensional visual detection device and a detection method, and belongs to the technical field of three-dimensional visual detection devices. The invention is an amphibious device capable of measuring a three-dimensional object and obtaining the position and attitude of the object by matching dot-line lasers to provide active features and combining land binocular stereo vision measurement with underwater multi-feature structured light measurement. According to the method, 25 pieces of line laser and 1 piece of dot laser are successfully taken as features, the technical problem of binocular stereo vision passive measurement can be solved, and meanwhile, the dot laser is taken as a marker to match the remaining light bars, the misidentification of the light bars caused by the underwater complex environment can be avoided, and the matching accuracy can be improved. Each picture can obtain 25 times more data than traditional structured light measurement, so that the underwater clear object feature data can be obtained, and the three-dimensional position and attitude under various complex conditions such as complexity, high occlusion, point cloud sparseness and the like can be solved with high accuracy.

A weak texture detection-based cross-scale cost aggregation stereo matching method of the present invention belongs to the computer vision field, in particular relates to a stereo matching method of the weak texture images, and comprises the following steps of inputting two color images which are separately a left image and a right image, utilizing the gradient information of the left image to carry out the weak texture detection and segmentation on a picture; calculating the matching cost according to the gradient information and the color information of the left image and the right image; taking the above weak texture detection and segmentation results as the references to carry out the Gaussian filter-based internal scale and cross-scale cost aggregation; adopting a winner taking strategy to calculate the parallax; adopting the left and right consistency detection and an adaptive weight-based method to refine the parallax, and outputting a parallax image. The method of the present invention achieves the technical purposes of improving the weak texture area matching correct rate under the premise of guaranteeing the texture area matching correct rate, and obtaining the better parallax image.

The invention relates to a panorama stitching method for removing a moving object in a moving video, comprising the following steps: (1) acquiring an appropriate number of video screenshots according to the length of a video, wherein the screenshot selection interval is shorter is the video is shorter, and the screenshot selection interval is longer if the video is longer; (2) projecting an original image to a sphere in order to prevent a final panorama from distortion and deformation; (3) carrying out SURF feature point detection on each frame of image to obtain a feature point description sub vector as the basis of matching; (4) carrying out feature point matching; (5) calculating a perspective (homographic) transformation matrix of every two frames of images based on matching feature points, and calculating forward and reverse homographic matrix transformation of two adjacent frames; (6) obtaining the size of the panorama based on the homographic transformation matrix of adjacent frames; and (7) inversely mapping the panorama to a sequence frame, building a candidate point list for each pixel point of the panorama, and removing a moving object, thus, completing generation of the panorama. The resolution of panorama generation is high, background overlay is accurate, and the 'ghost' phenomenon of the moving object is avoided.

The invention provides a representation method for knowledge markup languages of a Chinese question answering system and the Chinese question answering system. The representation method for the knowledge markup languages of the Chinese question answering system comprises a basic representation method for knowledges, a semantic representation method for the knowledges, a representation and instantiation method for knowledge templates, a quotation and association method for the knowledges, a representation and disambiguation for named entities in the knowledges, and a representation method for expression, mathematic and hypertext symbols as well as time in the knowledges. The invention furthermore provides the Chinese question answering system which uses the representation method. According to the representation method provided by the invention, the further research and application of the Chinese question answering system in the scientific research circles and the business circles are pushed.

The invention discloses an AUV multi-strategy navigation method based on submarine topography matching. The method comprises the steps: constructing a topography adaptive region selection model basedon a machine learning algorithm, wherein the topography adaptive region selection model includes a suitable topography matching region, a topography matching transition region, and an unsuitable topography matching region; during specific navigation, analyzing a type of a navigation area through collected real-time data, and carrying out navigation based on a multi-beam sonar underwater topographyelevation image matching auxiliary method for the suitable topography matching region; for the topography matching transition region, starting a side-scan sonar, and carrying out underwater side-scansonar image matching by using a multi-beam depth finder and the side-scan sonar data to assist ship position calculation for navigation; and for the unsuitable topography matching region, starting the side-scan sonar, and carrying out navigation based on simultaneous localization and a map construction algorithm. The navigation method has better significance and value in practical application.

The present invention relates to an image synthesis restoration method based on local structure features, comprising the following steps: inputting an image to be restored; initializing confidence coefficients of all pixel points in the image to be restored; selecting a template block of each pixel point by respectively taking each pixel point on a damaged boundary of the image to be restored as the center; calculating the priority of the template block where each pixel point on the damaged boundary of the image to be restored locates, and performing descending sort; searching an area which has the highest similarity with the template block having the greatest priority from an undamaged area in the image to be restored to serve as an optimal matching block; copying pixel values in the optimal matching block to the template block; and updating the confidence coefficients and returning to execute corresponding steps until the image to be restored is completely restored. The image synthesis restoration method based on local structure features provided in the present invention can improve image restoration accuracy for a damaged image with a complicated texture structure. The present invention also relates to an image synthesis restoration system based on local structure features.

The invention discloses a novel self-adaptive weight stereo matching method. The novel self-adaptive weight stereo matching method comprises: obtaining a reference image and a target image; obtaininga representation method for each pixel supporting weight value; improving the traditional self-adaptive algorithm, selecting a gray scale absolute difference in a CIELab color space to replace a grayscale absolute difference in a RGB space; using the similarity measure method of Census nonparametric transformation to use a relationship between a central pixel point and pixels in the neighborhoodto replace a gray value of the central pixel point; and finally performing parallax post-processing of left and right consistency detection, sub-pixel enhancement, and median filtering. The inventioneffectively improves the matching precision, enhances the robustness of the stereo matching, and improves the block effect generated by the traditional algorithm in a low texture region.

The invention discloses a nonlinear similarity index based fault classification and diagnosing method, which aims to solve the three key problems facing the fault classification and diagnosing in practical industrial process: 1) limited number of available training samples in reference to fault type; 2) available fault training data in the initial stage of operation state conversion in which the non-linear degree of the data is relatively strong; and 3) overlapping occurrence of sampled data for different faults in the spatial distribution. According to the invention, firstly, the characteristic variables are selected for each referred fault type; and the characteristic variables which can distinguish the fault from normal data are selected. Then, the characteristic variables are just used to match the similarity between the online fault data window and the reference fault data window. The similarity calculation method is based on the non-linear similarity index of window data, and the fault type can be determined online according to the minimum nonlinear similarity value. Compared with the traditional classification and diagnosing method, the method of the invention significantly reduces the dimension of the variables by the selection of the characteristic variables, which not only reduces the limitation of the training data greatly, but also eliminates the negative impact of non-characteristic variables on the calculation of the nonlinear similarity index. In addition, this method implements the fault similarity matching through the non-linear similarity matching of the window data in the spatial distribution, and can avoid the misclassification of overlapping data maximally.

The invention provides a personnel indoor positioning method based on an adaptive SIFI (scale invariant feature transform) algorithm. The method is characterized by: firstly, carrying out feature extraction on an global motion image sequence; then, taking a correct matching point as a sample set to estimate a camera motion model so as to obtain an actual offset of a camera. In the method of the invention, a Lagrange parabola interpolation is introduced. Last three times of model matching results are used to predict an overlapping region of a reference frame image and a current frame image. On the overlapping region, feature points are extracted and feature matching is performed so that a lot of information redundancies in a video image sequence can be eliminated, a processing speed of the each frame image can be accelerated, the validity of the feature points to be matched can be increased and mismatching can be reduced. Therefore, the algorithm is accurate and real-time and can be used for the indoor personnel positioning system.

The invention belongs to the field of digital image processing technology and discloses a fingerprint image splicing method based on detailed points and a distance image. According to the method, a detailed point pair with the highest matching score in two fingerprint images is obtained according to a fingerprint image matching algorithm based on the detailed points to serve as a reference point pair of rough matching, and rotational translation parameters corresponding to rough matching are solved; and whether rotational translation parameters of fine matching are optimal rotational translation parameters is judged according to a ridge matching error of an overlapping region between a detailed image of the first image and the distance image of the second image nearby the rotational translation parameters of rough matching. Through the method, the probability of erroneous matching is lowered, and splicing errors generated due to erroneous matching of the detailed point pairs are reduced; the problem that the number of the matching detailed point pairs is small due to the fact that the overlapping region is small is avoided; and moreover, the ridge matching error of the overlapping region between the fingerprint images is used to solve the optimal rotational translation parameters, so that fingerprint splicing errors are further reduced.

The invention discloses a method and system for retrieving multimedia information based on a bit vector. The method comprises the steps that the characteristic data of the current multimedia information are extracted to obtain the n-dimensional high-dimensional characteristic vector of the current multi-media information; the n-dimensional high-dimensional characteristic vector is converted through a projection matrix to obtain an m-dimensional middle vector; all the elements of an m-dimensional threshold vector are compared with the corresponding elements of the middle vector respectively, and binaryzation is carried out on the middle vector according to a comparison result to obtain the m-dimensional bit vector of the current multimedia information; m is smaller than n; according to the obtained bit vector, a bit vector similar to the bit vector is searched for in a multimedia characteristic database, and the multimedia information corresponding to the found bit vector is used for being output as a retrieved result. The method ensures original vector identification capacity, and after the high-dimensional characteristic vector of the multimedia information is mapped into a low-dimensional bit vector, the retrieval efficiency based on the bit vector is higher, and the retrieval loss based on the bit vector is smaller.

The invention relates to a space rope tying robot camera automatic motion detection and compensation method. The space rope tying robot camera automatic motion detection and compensation method is characterized in that regions contained moving objects in a front frame image and a rear frame image can be calculated; feature point detection is performed in regions without the moving objects through a SURF algorithm; an SURF feature description vector quantity adopted has the property of dimension and rotation invariance and is good in discrimination so that reliability and accuracy of image registration can be improved effectively according to the new algorithm. A cross matching algorithm is further adopted so that a fault matching ratio is reduced, and therefore robustness of the algorithm is strong and a solution of a motion parameter is accurate.

The invention belongs to the radar technical field and relates to a marine navigation radar data interconnection-based multi-target tracking method and system. The method includes the following steps that: a plurality of groups of radar data are collected from a radar; appropriate wave gate parameters are selected according to a tracking duration and a tracking state; the overlap ratio, distance ratio, shape ratio and energy ratio of navigation trajectory data and a plurality of point trajectory data are calculated through a point trajectory data and navigation trajectory data processing unit, so that an interconnection condition can be obtained; and a target point trajectory data tracking result is obtained through a target point trajectory output point. According to the marine navigation radar data interconnection-based multi-target tracking method and system of the invention, besides position information, the features of targets, such as the shape ratio and energy ratio are utilized; and multidimensional information judgment is utilized, and therefore, when multi-target tracking is performed in a dense clutter environment, the tracking accuracy of the targets can be ensured, and the probability of false matching can be decreased.

The invention provides an enhanced hidden Markov model-based map matching method. The invention aims to solve a technical problem that an existing map matching method fails to guarantee the accuracy of matching results and effectively reduce operation time. According to the enhanced hidden Markov model-based map matching method of the invention, traffic rules are introduced, so that the mismatch rate of the track points of a floating car on a retrograding road segment can be effectively reduced; the matching accuracy of the track points on a complex road segment can be improved; the matching time of the track points can be shortened; the overall operation time of a matching process can be effectively reduced; the self-correction of matching errors can be realized; and accurate track pointcorrection data can be provided for later-stage floating car data-based application.

Embodiments of the present invention provide a method and an apparatus for predicting a motion parameter of a target object. According to the embodiments of the present invention, the method for predicting the motion parameter of the target object comprises: according to detection data, acquiring a first motion parameter of the target object; according to the detection data, determining a neighborhood object positioned around the target object; obtaining a distance between the target object and the neighborhood object; and according to the first motion parameter and the distance between the target object and the neighborhood object, estimating a probability density function of a second motion parameter of the target object at the next moment.

The invention discloses a neuromorphic inspiring robot cognitive map building method, and belongs to the technical field of robot environment cognition and motion navigation. A cognitive map can be built to assist a robot in navigation. Firstly, the robot acquires image information surrounding the environment through a robot. Then, a visual odometer processes the image information to obtain the rotating angle speed and the line speed of the robot, the rotating angle speed and the line speed are input to a head orientation cell model and a position cell model, and cognition of the robot to theposition is formed. A visual feature extraction module processes images through a deep learning network and a principal component analysis algorithm, and outside cognition information of the robot tothe environment is obtained; and finally, information is combined through the experience map, and the map is updated through the closed-loop detection and update algorithm. By means of the method, theproblems that in the current neuromorphic cognitive map building method, the robustness of a visual odometer is poor, and accuracy of closed-loop detection is not high enough can be solved, detectionand updating to closed-loop points in the environment can be completed, and global consistency can be guaranteed.

The invention discloses a loopback detection method based on a convolutional neural network and ORB features. The method comprises the following steps: adding a new image i into an image sequence; Extracting feature vectors of the image i and other images in the image sequence by using a convolutional neural network, and calculating the similarity of other chords; Carrying out ORB feature extraction on the image i and the image j of which the similarity is greater than a threshold value; And carrying out feature matching on the image i and the image j, and if the correct logarithm of the finally matched feature points of the two images is greater than a set threshold value, considering that loop-back occurs. Due to the fact that the convolutional neural network is used for replacing a traditional word bag method, the speed and accuracy of loopback detection are improved. According to the invention, the convolutional neural network and ORB features are combined, so that the mismatchingprobability is reduced.

The invention discloses a data benchmarking method and device and a storage device. The data benchmarking method comprises the steps that original data information is extracted from a data table to bebenchmarked, and the original data information comprises field names and field annotations corresponding to the field names; identifying the field annotation based on a sequence annotation model of deep learning to obtain a characteristic word corresponding to the field name; carrying out first text matching on the characteristic words corresponding to the field names and standard data elements in a standard library; and verifying a result output after the first text is matched. By means of the mode, text matching is conducted on the basis that the feature words are recognized, the credibility of a text matching result is improved, and the mismatching rate in the benchmarking process is reduced.

The invention discloses a monocular vision mileage measuring method based on image color characteristics. The method comprises the steps of using an image acquisition device for transmitting the acquired color image sequence to a computer through a transmission interface; extracting features from an input image frame, carrying out feature matching and calculating a frame descriptor of the image; establishing an initial map point meanwhile, using two initial frames as key frames, projecting the successfully initialized map point into the current frame; carrying out bundling optimization based on a reprojection error between the two frames of images, obtaining a pose transformation relation of a camera between the two frames to track the motion of the camera, finally selecting a new key frame according to the similarity with a previous key frame and an existing key frame in a map, adding the new key frame into the map, and updating the map. The method has a more robust feature matching effect, the used equipment is simple, the cost is relatively lower, and the method has relatively higher engineering application value and a relatively wide application scene.

The invention discloses a rock slice image alignment method combining single polarized light and orthogonal polarized light images. The method comprises the following steps: S1, preprocessing a single polarized light image and an orthogonal polarized light image of a rock slice; s2, extracting feature points in the preprocessed image, and performing rough matching on the extracted feature points; s3, based on the rough matching result of the feature points, obtaining a homography matrix for pixel mapping by utilizing a bidirectional cross matching method, and taking the homography matrix as an accurate matching result; and S4, mapping all pixels of the orthogonal polarized image into the single polarized image based on an accurate matching result to achieve rock slice image alignment. According to the orthogonal polarized image and the single polarized image which are aligned through the method, a serious distortion phenomenon does not occur in a subsequent fusion process, and a relatively good image alignment effect is achieved; according to the method, the calculation time of image alignment can be reduced, the real-time performance is achieved, the influence of noise on feature extraction can be inhibited, and the method has stability when feature points are extracted.