129 results about "Global matching" patented technology

The invention discloses a spatial non-cooperative target pose estimation method based on model and point clod global matching. The method comprises the steps that target scene point cloud is acquired by using a depth camera, the target scene point cloud acts as data point cloud to be registered after being filtered, and three-dimensional distance transformation is carried out on the target model point cloud; deblurring main directional transformation is carried out on the initial data point cloud to be registered and the target model point cloud, a translation domain is determined, search and registration are carried out in the translation domain and a rotation domain by using a global ICP algorithm, and an initial transformation matrix from a model coordinate system to a camera coordinate system is acquired, namely, the initial pose of a target is acquired; a pose transformation matrix of the pervious frame is enabled to act on data point cloud of the current frame, and registration with a model is carried out by using the ICP algorithm so as to acquire the pose of the current frame; and a rotation angle and a translation amount are calculated from the pose transformation matrix. The method disclosed by the invention has good anti-noise performance and an ability of outputting the target pose in real time, geometric features such as the normal and the curvature of the data point cloud are not required to be calculated, the registration speed is high, and the precision is high.

A fingerprint recognition method includes iterative gamma correction that compensates moisture effect, feature extraction operations, directional morphological filtering that effectively links broken ridges and breaks smeared ridges, adaptive image alignment by local minutia matching, global matching by relaxed rigid transform, and statistical matching with Gaussian weighting functions.

The invention provides a method for detecting and identifying a station logo. The method comprises the following steps of: 1) setting a station logo example video frame, performing edge detection and filtering on the station logo example video frame, further positioning and segmenting a station logo area, and partitioning the station logo area to obtain a station logo template; 2) setting a video frame to be detected and identified, performing edge detection on the video frame to obtain an edge image of the video frame to be detected and identified; and 3) acquiring a station logo detection and identification result by a sliding window searching and block matching method according to the station logo template and the edge image of the video frame to be detected and identified. Compared with the conventional global matching-based method, the method of the invention can achieve higher station logo detection and identification accuracy rate and time efficiency.

The invention belongs to the technical field of three-dimensional imaging, and provides a three-dimensional imaging method and system for a human body and a simultaneous calibration method for multiple control base stations. Under the condition that spatial overlap does not exist in effective view fields of all distributed sensors, the structural parameters and the global matching parameters of all the sensors are calibrated. By using the corresponding point searching method with phase shift combined with random structured light, the image collection time for obtaining single-viewpoint depth data is shortened. By using the thought of time multiplexing, the whole data collection time of the system is compressed, due to the design of distributed computing, the computing power of the whole system is enhanced, and the speed for data acquisition is increased. According to the calibrated global matching parameters of all the different sensors, the depth data of the different sensors are matched automatically. The system can be debugged and expanded conveniently, the automation degree is high, and the three-dimension scanning process is easy to implement.

The invention relates to a floor distinguishing method based on the RSSI difference between floors. According to the method, in the indoor environment of multiple floors, the floor where a terminal is located is determined by the utilization of the characteristic that RSSI declines between the floors. Firstly, the floors where all APs are located are recorded, a mobile terminal with the WiFi function is used for collecting the RSSI of all the APs of the current position, and by counting the number of the APs of which the RSSI is greater than the threshold in each floor, the floor comprising the largest number of the APs of which the RSSI is greater than the threshold is determined to be the floor where the terminal is located currently. Compared with an ordinary global matching floor locating algorithm adopting signal fingerprints, the method has the advantages that a floor distinguishing algorithm based on the RSSI difference between the floors is added, the calculated quantity for locating the floors is decreased, and accuracy for distinguishing the floors is improved.

The invention discloses a method for reconstructing real-time three-dimensional face data with multiple images. The method includes the following steps: constructing shooting systems of eight cameras; performing multi-camera calibration; performing image matching, to be more specific, calculating global matching points of an ith group of images, and calculating accurate matching points of the ith group of images; calculating three-dimensional face coordinates via the matching points and a projection matrix of the cameras, to be more specific, imaging an object point on the third image of the ith group, and taking an intersection of three projection lines as the object point so as to complete reconstructing of the real-time three-dimensional face data. The face data obtained in the method are global, dense and real-time and can be used directly, and the method can be widely applied to the fields such as the animation industry, computer 3D (three-dimensional) games, human-computer interaction, identity recognition, medical virtual surgery, videophones, face recognition, face expression, age simulation, film and TV advertising production and computer cognition.

Embodiments of the present invention relate to a sensing information fusion method. The method comprises the steps of: acquiring first obstacle information, second obstacle information, third obstacleinformation, vehicle positioning information and vehicle speed information collected by different sensors; performing time synchronization of the first obstacle information, the second obstacle information, the third obstacle information, the vehicle positioning information and the vehicle speed information; performing preprocessing of the first obstacle information, the second obstacle information, the third obstacle information, the vehicle positioning information and the vehicle speed information after time synchronization; performing global matching of the information collected by each sensor after preprocessing to obtain a matching result; and according to the matching result, performing multi-target Kalman filtering tracking, and obtaining obstacle fusion data. The sensing information fusion method can achieve perception of unmanned sweeper for the environment road condition can make more accurate obstacle avoidance decision and obstacle avoidance control so as to enhance the intelligence of the vehicles and facilitate achieving of unmanned automatic driving sweeping.

The invention relates to a computer binocular vision matching method based on global and local algorithms, which comprises the following steps: (1) obtaining a layered matching result from a left image and a right image which are obtained by a binocular three-dimensional vision sensor by using an image cutting method; (2) obtaining different local matching results in a local matching result by using a window matching method with different windows; (3) putting the local matching results into a markov random field with a second-order smooth energy function for iterating and optimizing to obtain the matching result. The computer binocular vision matching method can integrate global optimized and local characteristics and obtain a more accurate matching result.

The invention provides a twin network target tracking method and system based on a convolutional self-attention module. The method comprises the following steps of constructing a feature fusion network model; learning the target image features on the template branches and local regions in the target image features of a search region through a convolutional neural network model to obtain the corresponding local semantic information, and aggregating the local semantic information to obtain the global context related information; pre-training the feature fusion network model; using the pre-trained feature fusion network model to extract the target image features and the search area target image features in the template branches, and introducing the features into the classification branches and regression branches of an area suggestion network with the anchor points; carrying out depth cross-correlation convolution calculation to obtain similarity scores; and performing target tracking on the target candidate block with the maximum similarity score. According to the invention, the global matching accuracy of the target image and the target image of the search area can be improved, and the more accurate tracking is realized.

The invention provides a method for searching keywords based on image content, and a device thereof. The device for searching keywords searches and locates an input keyword in an input file image and comprises a global matching unit, a checking unit and a filtering unit; wherein the global matching unit extracts a plurality of candidate keyword image areas from the file image, extracts the image characteristics of the plurality of candidate keyword image areas, and matches the image characteristics with the keyword so as to obtain a matching distance corresponding to the plurality of candidate keyword image areas; the checking unit carries out identification to the front N candidate keyword image areas with small matching distance and calculates the checking distance between an identification candidate and the keyword; the filtering unit calculates the combining distance of the matching distance and the checking distance and filters the candidate keyword image areas with large combining distance according to the combining distance.

The invention discloses a fingerprint matching method based on multiple reference point pairs, which comprises the following steps: determining template fingerprints and input fingerprints, acquiring the matching scores of every detailed point pair by a fingerprint matching algorithm based on three-value characteristic vector, and ranking all the detailed point pairs according to the matching scores in a descending order; selecting the first m detailed point pairs as the initial reference point pairs, and computing the rotation / translation parameters of the initial reference point pairs; selecting the reference point pairs according to the conformity, carrying out global matching by using each reference point pair as the reference to acquire a matched detailed point pair set, and fusing information of all the matched detailed point pair sets by using a voting fusion strategy; and acquiring the number of final matched detailed point pairs, computing the matching scores according to thenumber of final matched detailed point pairs, and judging whether the template fingerprints are matched with the input fingerprints according to the matching scores and the matching threshold. The invention overcomes and solves the problem of nonlinear deformation of fingerprints in the existing fingerprint matching algorithm, and has favorable fingerprint matching performance.

A super-pixel-level image global matching method comprises the steps of obtaining an input image pair corrected by polar lines through a binocular stereo camera, calculating a self-adaption cross of each pixel of the input image pair so as to obtain a self-adaption window of the current pixel, calculating the matching cost of the pixel, adopting a replacement strategy to process a coverage area and adopting a suboptimum strategy to process an image boundary; establishing super-pixels for images, conducting planar fitting on a parallax value of each super-pixel area to determine reliable pixels and deleting obvious wrong planes so as to determine initial parallax plane set; calculating the matching cost of the super-pixels according to the obtained matching cost of the pixel, establishing a data item and a smoothing item and utilizing a Graph-Cut optimization algorithm to conduct continuous iteration on an energy equation so as to obtain a final parallax plane. The super-pixel-level image global matching method can effectively avoid the problem that image noise, distortion or pixel value abnormity and other situations easily occur in a weak texture area, a discontinuous-parallax area and the coverage area, is good in robustness and can obtain depth information more approximating to real scenes.

The invention discloses a Wi-Fi fingerprint indoor positioning method low in complexity and belongs to the field of Wi-Fi indoor positioning. In an indoor environment, the position of a terminal is determined by utilizing Wi-Fi signal intensity RSSIs, received by the terminal, of multiple APs and the direction of a magnetometer. In the fingerprint database building stage, a fingerprint database is built through a sampling and averaging method. In the real-time positioning stage, according to the direction of the terminal and the position of the terminal at the previous time, a fingerprint database subset is obtained for calculating the position, and the calculation complexity of a matching algorithm is reduced. Compared with a traditional fingerprint global matching algorithm, the Wi-Fi fingerprint indoor positioning method low in complexity can reduce the calculation complexity and improve the positioning precision.

The invention discloses a local matching method of a three-dimensional model. The method comprises the following steps of: calculating the bending salience of any vertex of the three-dimensional model; generating a sequencing list according to the bending salience and carrying out region growing to form a local vertex set; carrying out secondary surface fitting according to the local vertex set formed by the region growing; identifying a fitting region to be a three-dimensional model local subblock and extracting the features of the three-dimensional model local subblock; carrying out featurecomparison and local matching on the three-dimensional model local subblock; and carrying out overall feature comparison and overall matching on the three-dimensional model. The method disclosed by the invention can be used for matching locals of the three-dimensional model, realizing the local-to-local retrieval and the local-to-overall retrieval of the three-dimensional model and extracting obvious region features of the locals, thereby increasing the matching precision of the method; and shape normalization is not needed any more in a feature extracting process, thereby the stability of feature extraction is guaranteed.

The invention relates to a stereo matching method based on a digital score delay technology, comprising the following specific steps: correcting the obtained left and right images by the correction algorithm and converting the left and right images into grayscale images; then segmenting the target objects which need the ranging in the images after conversion by the segmenting method, and then projecting the targets in the left and right images respectively in the vertical direction to convert into a one-dimensional signal, thereby converting the stereo matching of the images into the delay detection of the corresponding one-dimensional signal; minimizing the error through the global matching to obtain the parallax of the entire pixel; then calculating the parallax in sub-pixel level by using the method of digital score delayer; and finally calculating the depths of the target objects by the method of binocular vision. The method provided by the invention can be used for calculating the depth of far targets, such as buildings, bridges and the like in long distance.

In order to meeting the requirement of binocular distance-measurement at night, the invention provides a distance-measuring system and a distance-measuring method based on a shimmer binocular camera.Considering the defect of low signal noise ratio of shimmer images, the distance-measuring method comprises the following steps: performing pretreatment to reduce image noise, fully utilizing the color and texture information of global image in a feature matching and optimizing process, reducing the error matching effect caused by the local noise; combining the calculation result of the Census descriptor and color feature with a semi-global matching optimization algorithm, the robustness of feature extraction and the accuracy of feature matching are enhanced. The distance-measuring system andthe distance-measuring method have night shimmer distance-measuring ability at night and can realize the night accurate distance-measurement.

When position matching is performed on two images in which the soft tissue structures and the bone tissue structures are displaced in different directions, artifacts appearing in the subtraction image are reduced, and it becomes possible to perform the position matching according to the diagnostic objective or the preferences of the operator. A global position matching means that globally matches the positions of the entirety of two inputted images that have been obtained of the same subject, and a local position matching means that locally matches the positions of the two images that have been globally matched by the global position matching means are provided. The global position matching means has a function for performing a bone tissue structure global position matching process and a function for performing a non-bone tissue structure global position matching process. One of these two functions is selected and the global position matching is performed.

The invention relates to a multi-platform point cloud intelligent processing method for holographic mapping. The method comprises the steps of organizing and scheduling mass point cloud data; carryingout point cloud data quality control and improvement, and achieving automatic correction of point cloud position consistency under the condition of no control point; carrying out high-precision fusion of multi-platform laser point cloud data, which comprises neighbor point cloud search, global matching energy equation construction and bipartite graph minimum cost matching; automatically extracting the ground object targets, namely realizing high-precision extraction and vectorization of the full-type ground object targets through geometric semantic consistency extraction of the ground objecttargets; carrying out the ground object target multi-detail-level model reconstruction based on the Grignard law, which comprises the step of establishing a multi-detail-level three-dimensional modelof the ground object target through a Grignard mathematical model and a topological relation graph. The holographic surveying and mapping product production process based on the multi-platform point cloud data is achieved, the method is easy to operate and implement, the manual workload of data processing can be greatly reduced, the working production efficiency is improved, and the product updating period is prolonged.

The invention relates to a bus station arrival location matching method and system. The method includes the following steps that: step 1, bus position information in an analysis period is acquired, all data to be analyzed are marked as unvisited, and one piece of unvisited data p is extracted out and is marked as visited, the attributes of the IDs of sub-segments of a road segment to which sub-segments belong, and the attributes of the ID of the rod segment of the record are obtained, step 2 is carried out; step 2, bus station site codes corresponding to the road segment ID which is equal to a p. road segment ID and a reference point which is equal to sub-segment ID+-10 belonging to a P. road segment, namely, the real-time arrival station locations of a bus are obtained; step 3, the GPS time of the p is adopted as station arrival time; and step 4, the method returns to the step 1, and analysis in the step 2 is carried out until all records are marked as visited, and calculation is terminated. According to the method and system of the invention, comprehensive analysis and processing are performed on road segments, the sub-segments of the road sections, running directions, GPS positioning the positioning time, and accurate and global matching is performed on the locations of the bus, and therefore, subsequent bus scheduling and planning can be facilitated, and the application of public transportation resources can be maximized.

Disclosed are systems for and methods of registering a deformable image with a reference image subject to a local or global quality of match. These systems and methods may use the quality of match to reduce effects of artifacts within the deformable image or the reference image on a deformation algorithm. The registration includes modifying the deformable image to match similar features in the reference image. The systems include a deformation engine configured for performing the deformation algorithm subject to the local or global quality of match. The systems and methods allow for correcting for cone beam computed tomography artifacts in one or more of the images.

A real-time moving target tracking method based on a global matching similarity function increases reliability of target color distribution by building improved color histogram models on the basis of real-time detecting moving targets, and constructs the global matching similarity function according to positions, sizes and colors of the moving targets so as to track the moving targets reliably. The specific steps includes: firstly, initializing tracking parameters of the real-time moving targets, secondarily, building the improved color histogram model for each detected real-time moving target, thirdly, forming to-be-matched moving target pairs by the current detected moving targets and the previously detected moving targets, building a color histogram similarity function, a central position closeness degree function and a pixel point sum closeness degree function for each to-be-matched moving target, and finally, building the global matching similarity function to track the real-time moving targets. The real-time moving target tracking method increases reliability of tracking moving targets, can effectively resolve a problem that the targets are shielded and an automatic correlation problem when the targets are blended and then separated again.

The invention discloses a method for implementing the SHA-1 algorithm by using a GPU, which includes the following steps: (1) implementing the SHA-1 algorithm on a general-purpose CPU; (2) transplanting the program to the GPU, and performing preprocessing according to the following method Processing: a. Create an empty function body; B. Copy the parameters from the main memory to the global matching storage space of the GPU through the parameter pointer, and copy the result from the global matching storage space of the GPU back to the main memory; C. Add device configuration modifier; D. Synchronize. The present invention provides a method for implementing the SHA-1 algorithm by using the GPU, utilizing the rapidly developing graphics hardware technology, adopting the CUDA parallel computing architecture of NVIDIA Corporation for parallel implementation, and using the GPU as a parallel computing device for program release and management operations, with flexible It has good performance, reduces CPU load, improves the processing performance of information security equipment, improves computing efficiency, and has the advantages of high security in the computing process.

The invention discloses a point registration algorithm capable of keeping local topology invariance. The point registration algorithm comprises the following steps: converting the point registration problem into an image matching problem, wherein points in a point set serve as nodes in an image; defining a neighbourhood region of each node in the image, wherein the node and the neighbourhood region thereof form a local subgraph, the node and one point in the neighbourhood region thereof are connected to form one edge in the local subgraph; carrying out dynamic planning on the local subgraphs to obtain matching probability between the local subgraphs; carrying out processing on global matching probability through the matching probability between the local subgraphs and a relaxation labeling method to obtain correspondence relation of point pairs; and according to the correspondence relation of the point pairs, carrying out elastic deformation regularization processing on the point set through a thin plate spline function to obtain model parameters. The point registration algorithm capable of keeping local topology invariance accurately quantizes local topology difference, improves accuracy of non-rigid registration when large deformation occurs and improves algorithm robustness when an abnormity point exists.

The invention discloses an image matching method based on a corner point and single-line segment marshalling feature description operator. The method comprises the following steps: firstly, extractingline segments and Harris corner points from an image; constructing a line segment, searching and grouping the line segment, constructing an angular point-single line segment texture descriptor with scale, rotation and illumination invariance by using a detected angular point and the line segment, the Harris angular point having the advantage of rotation invariance, and the line segment adopting ahalf-width descriptor can improve the reliability in a parallax change scene; performing spatial weighted shortest distance measurement to obtain a local matching result; finally, determining the candidate matching of each line segment, establishing a matching matrix and solving a global matching result through spectral analysis. The image matching description operator has the characteristics ofscale, rotation and illumination invariance. According to the invention, image pyramids are respectively established for three-dimensional images. The pyramids on different layers are matched one byone, so that the scale influence can be eliminated. The defects of large marshalling calculation amount and long consumed time in multi-line-segment matching can be overcome.

The present invention discloses a non-cooperative navigation satellite identification method based on a Doppler frequency shift characteristic, and belongs to the non-cooperative spacecraft orbit determination technology field. The method of the present invention comprises the steps of obtaining the single carrier signals of different frequencies and the Doppler frequency shift measurement values by the codeless carrier recovery; calculating the Doppler frequency shift prior values according to the initial orbit parameters of a low-orbit user satellite and the navigation star ephemerides marked on a cycle; arranging the Doppler frequency shift measurement values containing the receiver frequency difference and the prior values into two sets of samples according to a descending order, and using a dichotomy to find a frequency difference estimated value efficiently to realize the two sets of consistent and corresponding frequencies; according to the proportion of the Doppler frequency shift matching residuals, adopting a global matching and local rejecting method to improve the global credibility of the satellite identification. The method provided by the present invention can identify a navigation satellite, and also can obtain the more accurate receiver frequency difference estimated value.

The invention relates to the field of oblique photography three-dimensional reconstruction of unmanned aerial vehicles, and discloses an oblique photography three-dimensional reconstruction matching method, which comprises the steps of extracting an image position and a pose feature component; extracting an image color feature component; extracting texture feature components; standardizing the three components, and performing combining to obtain a fusion feature vector; constructing a hash function to map the obtained fusion feature vector; and further calculating the Euclidean distance, performing sorting to form a to-be-matched image set, and then selecting a single image matching object from the to-be-matched image set in the structure motion recovery process. Compared with the prior art, POS information and color texture information of an unmanned aerial vehicle image are fused; the images before reconstruction are clustered according to the characteristics of the unmanned aerial vehicle images, the global matching process of a traditional reconstruction algorithm is avoided, the efficiency in oblique photography image three-dimensional reconstruction is remarkably improved, and the reconstruction dense point cloud can keep good precision while the reconstruction efficiency is improved.

The invention relates to the field of artificial intelligence, and provides a PPT video segment extraction method and device, equipment and a medium. According to the PPT video segment extraction method and device, the equipment and the medium, one or more key point short videos can be automatically clipped; users can be facilitated to quickly understand video subjects; important reference is provided for later forecast production and the like; a machine learning method is effectively used to establish feature engineering for video analysis; on the basis, coarse segmentation is carried out onan audio file; feature weights are increased through global matching and local matching; fine-grained division is carried out on chapters; and therefore, a large number of training models are not needed, personalized customization can be carried out for a certain vertical field, and operability is higher. The invention also relates to a block chain technology. At least one short video can be stored in a block chain.

The invention provides an interframe motion estimation method and device and electronic equipment. The method includes the steps of according to a current frame image and a known coding frame image, building a background model; according to the background model, determining a moving target area of the current frame image; tracking each moving target in the moving target area, and determining a motion vector of each moving target; according to the motion vectors, obtaining an initial motion vector value of each pixel block in the moving target area; as for each pixel block, searching for the optimal global matching reference block of the pixel block in a preset domain of a searching point which the initial motion vector value of the pixel block in the reference frame image points to, and calculating a motion vector estimation value of the pixel block; adopting the motion vector estimation value of each pixel block and the corresponding matching reference block as interframe motion estimation parameters of the current frame image. In this way, time consumption for interframe motion estimation can be reduced, real-time coding demands are met, and the motion estimation precision is improved so that the quality of coding videos is improved.

The invention relates to a single-line laser radar two-dimensional positioning method based on a light reflection column or a light reflection panel. The method comprises the following steps: S1) carrying out horizontal distortion correction on original measurement data of a single-line laser radar, and calculating actual position of a horizontal target bearing; S2) carrying out motion distortioncorrection on the original measurement data of the single-line laser radar; and S3) carrying out global matching on the measurement data of the single-line laser radar obtained after distortion correction in the steps S2) and S3) and map data of the actual position of the target bearing in the step S1). According to the single-line laser radar two-dimensional positioning method based on the lightreflection column or the light reflection panel, through the horizontal distortion correction and the motion distortion correction, errors of the measurement data itself are greatly reduced, and the error is smaller; and compared with a triangulation location method, the method enables all of the target bearings to participate in calculation to obtain an optimal solution instead of enabling threepoints to participate in calculation in triangulation location, and the error and precision of the three points directly affect the final calculation result, so that precision is higher.