85results about How to "Guaranteed extraction accuracy" patented technology

The invention discloses a method for detecting vehicle door outer plate shape and size based on single-projection encoding structured light and aims to break through the condition that structured light visual detecting technology is rarely applied to online detection on the vehicle door outer plate shape and size. The method comprises the following steps of 1, calibrating a structured light detecting system, i.e. calibrating a camera (2) and a projector (4), and establishing corresponding non-linear relation between a space position and an image coordinate; 2, designing an encoding projection template; 3, decoding and identifying a detecting image, i.e. projecting the encoding projection template to the surface of a vehicle door outer plate by using the projector (4), shooting a surface image of the vehicle door outer plate by using the camera (2), and transmitting the shot image to a computer to be processed and identified; and 4, calculating a space point cloud coordinate and evaluating an error, i.e. 1) calculating a three-dimensional space coordinate of surface projection image characteristic points of the vehicle door outer plate, and 2) matching and rectifying a detecting point cloud obtained through calculation and a point cloud model measured by a three-coordinate measuring machine.

The invention relates to a hybrid image processing method of extracting the centric line of the structured-light stripe based on the ROI area, belonging to the field of machine vision technique. By combining the image threshold processing, communicated components extraction, shape factor constraint and expansion method, the invention automatically divides the structured light area as the ROI area of light stripe extraction, using the extraction method of fringe center based on Hessian matrix to realize the extraction of sub-pixel image space of the centric line of the structured-light stripe of the ROI area. The method has the advantages of high precision, good robustness, high automation degree and other advantages, avoiding the large formwork Gaussian convolution calculation of the pixel the non-light stripe area, thereby reducing the redundant calculation of the extraction of the structured light stripe and realizing the high speed and precision extraction of the centric line of the light stripe, laying a foundation for the real-time application of the structured light 3D vision measurement.

The invention discloses a multiscale hierarchical processing method for extracting object-oriented high-spatial resolution remote sensing information. The method comprises the following steps: at the global level, determining a texture sampling interval and a template window size by utilizing image semi-variance statistical calculation, and dividing the images into several local areas with smooth textures in a course scale on the basis of image texture characteristics; at the local level, setting local segmentation scale parameters through space statistical calculation by taking the local areas as units, and carrying out fine segmentation by utilizing the geometric and spectral information of the images so as to obtain refined image objects which are capable of embodying more details; and carrying out object-oriented remote sensing image classification by taking the global or local areas as units through utilizing a global or local area sample training classifier. According to the method provided by the invention, multiscale hierarchical processing is carried out, so that both the macro and micro characteristics of the images are considered and the ground features can be divided more accurately; and the specific characteristics of the remote sensing images are combined to decide whether to partition to carry out image classification, so that the classification accuracy of the whole image is improved.

The invention puts forward a virtual studio mobile robot visual localization method based on improved Apriltag tags, comprising the following steps: adding four uniformly-sized ring markers at the appropriate length of quadrilateral diagonal extension cords of a tag; calibrating a camera; acquiring tag images: preprocessing the acquired images; carrying out threshold processing on ROI region extraction; identifying elliptical markers; extracting the ROI region; carrying out threshold processing on the minimum rectangular area surrounding the four points which is the ROI tag area of the tag images with the coordinates of four ellipse centers on the improved tags as the benchmark; extracting tag corner points and fitting the extracted straight lines by the weighted least square method, calculating intersection points of straight lines to obtain subpixel-level corner point coordinate, and extracting image coordinates of the tag corner points; reading the coded message of the tags according to the extracted image coordinates of the corner points, obtaining world coordinates of the corner points, and obtaining the pose of a robot by coordinate transformation. The method of the inventionhelps reduce the data size of the identification tags and improve the real-time performance of identification.

The invention discloses a video splicing method, a device, equipment and a medium. The method comprises the following steps: adopting feature matching and/or adopting barrage information analysis to determine N splendid video segments from the target video, wherein N is greater than 1; extracting the N splendid video segments from the target video; splicing the N splendid video segments into one video to form a spliced video; when a request for obtaining the stitched video sent by a client is received, the stitched video is sent to the client for playback. The method, device, equipment, and medium provided by the present application can solve the technical problems existing in the prior art of watching historical game live videos, which waste viewers' viewing time and cause viewers to havea low probability of obtaining wonderful video clips. The technical effect of saving viewing time is realized.

The invention provides an unbalance excitation force compensation method of an LMS adaptive filtering bearingless motor. The method comprises the following steps: firstly, measuring the rotating mechanical angular velocity Omegar and the real-time radial displacement x(k) of a motor rotor, and constructing dual reference signals of an LMS filter according to the Omegar and the x(k); inputting the x(k) and the dual reference signals into the LMS filter, adjusting the weight of the filter by use of a weight adjustment algorithm containing a simple step length factor adjustment function, and extracting the optimal estimation value of an unbalance vibration displacement signal of the rotor; and performing rotor synchronous revolution transformation, closed-loop feedback and inverse rotor synchronous revolution transformation on the optimal estimation value in sequence, thereby obtaining an unbalance vibration compensation control force signal. The adaptive variable-step length factor adjustment function is capable of adjusting a critical tracking error absolute value between the adaptive fast adjustment stage and the slow adjustment stage of the step length factor by use of a parameter c. The unbalance excitation force compensation method is small in calculation amount, and capable of overcoming unbalance centrifugal excitation force generated due to the mass eccentricity of the rotor; as a result, the suspension control accuracy of the rotor and the unbalance displacement tracking extraction and convergence rates are increased.

The invention discloses an ORB feature extraction and matching method and device. The method comprises the following steps: firstly, converting an original image into a gray level image, performing FAST angular point and HARRIS corner point extraction, centroid calculation and Gaussian blur, and outputting the feature point key information and Gaussian blur images; performing down-sampling on theimage to obtain a plurality of sets of feature point key information and Gaussian blur images; calculating the descriptor information corresponding to the feature point, performing feature matching onthe descriptor information of the two frames of images to obtain the ORB feature matching result; and finally outputting the original image and the ORB feature matching result synchronized with the original image. The problem that a large amount of memory and CPU resources are consumed for identifying the same target image in the prior art is solved. According to the method and the device, the speed for identifying the same target image can be increased, and meanwhile, the real-time performance is guaranteed while the feature extraction and matching accuracy is guaranteed.

The invention discloses a weighted constraint-based star atlas segmentation method, which comprises the following steps of: (1) initializing a segmentation threshold; (2) optimizing an iterative segmentation threshold; (3) weighting the iterative segmentation threshold; and (4) segmenting a star atlas. According to the weighted constraint-based star atlas segmentation method, an optimized iterative algorithm and a weighted iterative algorithm are combined, an ideal segmentation threshold for the star atlas is obtained by adding constraint conditions in a weighting iterative process and using the autonomous adjustments and changes of two weighted factors, then, the star atlas is segmented, and star point targets in a perspective atlas obtained by segmenting can meet the practical requirements on the number of star points; and the weighted constraint-based star atlas segmentation method can effectively eliminate the influence on the identification and location of the star points caused by noises and a background, and the processing speed for the location of the star atlas is greatly increased under the premise that the accuracy in the extraction of the center of mass of following star points is ensured.

The invention relates to a semi-automatic antarctic subglacial lake recognition method based on an ice radar technology. The semi-automatic antarctic subglacial lake recognition method based on the ice radar technology has the advantages that in consideration of the characteristics that the interface between a subglacial lake and ice on an ice radar fracture surface image has high echo signal intensity and is flat, the initial computer recognition of a subglacial lake surface is realized by using an automatic extraction method in which the surface roughness and the radar reflection power are combined, so that the recognition efficiency and the recognition accuracy are greatly improved; the location of a subglacial lake is corrected through manual modification, no excessive workload is increased, errors caused by automatic extraction can also be made up, so that the final recognition result is relatively accurate, and the speed and the reliability are relatively high.

The invention discloses a crop effective grain number measuring method. According to the crop effective grain number measuring method, the grain distribution images and accurate grain weighing data of crops in different varieties are obtained through photographing and weighing respectively; the number of the grains of the crops is extracted preliminarily through independently-developed data processing software and based on a mathematical morphology principle and a digital image processing technology; and the number of the grains is corrected by using the weighing data of the grains, so that imperfect grains are eliminated, and therefore, the extraction accuracy of the number of effective grains can be improved. With the method of the invention adopted, image preprocessing, and grain number accurate and quick identification and batch extraction can be carried out on photos with different grain conditions, and crop grain extraction accuracy can be ensured, and extraction speed can be greatly improved.

The invention discloses a signal dedrying method based on time-frequency analysis. The method comprises the steps that 1, a to-be-processed signal is synchronously stored; 2, signal dedrying is conducted, wherein data processing equipment is adopted for dedrying the to-be-processed signal f(t), the process comprises the substeps that 201, signal sparse decomposition based on an optimization algorithm is conducted; when the data processing equipment is adopted for seeking psirn(t), and the time-frequency parameter rn is sought through the following processes of C1, time-frequency parameter optimization; C2, optimal time-frequency parameter determination; 202, signal reconstruction is conducted. The method is simple in step, reasonable in design, convenient to implement and good use effect,a signal sparse decomposition method based on the optimization algorithm is adopted for seeking the optimal matching atom, and meanwhile a fitness value and sparsity are combined for determining the optimal matching atom; the signal dedrying speed can be greatly increased, the dedrying effect can be effectively improved, and the accuracy of the signal obtained after dedrying is guaranteed.

The invention provides a method and a device for generating coronary artery center line extraction model, and the method comprises the steps: carrying out the segmentation processing of a coronary artery sample image, and obtaining a coronary artery sample segmentation image; performing coronary artery center line marking on the coronary artery sample segmentation image to obtain a coronary artery marking image; obtaining an output image of a first generator in a coronary artery center line training model trained last time, inputting the coronary artery sample segmentation image and the output image into the coronary artery center line training model, obtaining a coronary artery center line prediction result image, and the coronary artery center line training model comprises a plurality of generators connected in sequence; inputting the coronary artery center line prediction result image and the corresponding coronary artery marking image into a discriminator, performing loss calculation according to the judgment result of the discriminator, and if the loss meets a preset loss threshold value, obtaining a coronary artery center line training model meeting the loss threshold value. The complexity of a coronary artery center line extraction model can be reduced.

The invention discloses a defect detecting method based on ultrasonic flaw detection. The method comprises the following steps: 1, acquiring an ultrasonic echo signal and uploading the ultrasonic echosignal synchronously; 2, determining wave crests and wave troughs; 3, eliminating extreme points; 4, dividing the signal: determining a time interval between every two adjacent extreme points, judging dividing points, determining sampling moments of the dividing points, judging signal division, sorting the dividing points and dividing the signal; 5, judging signal separation and performing signalseparation; and 6, identifying a defect. The method is simple in steps, reasonable in design, convenient to implement and good in using effect; the judgment of the dividing points and the determination of the sampling moments of the dividing points are achieved by performing threshold judgment on the time interval between every two adjacent extreme points, and the ultrasonic echo signal is divided; then signal separation judgment and signal separation are respectively performed on every divided signal, and independent ultrasonic echo signals at multiple defect positions in the same ultrasonicecho signal can be quickly and accurately identified.

The invention relates to an extraction method for target movement information based on a single remote sensing image. The extraction method comprise the following steps of (1) laying a plurality of linear array imaging sensors or multiband planar array imaging sensors on the same optical remote sensor imaging surface, and acquiring pattern parameters of the above imaging sensors; (2) acquiring temporal and spatial parameters of imaging the remote sensing images by a remote sensing imaging controller, a remote sensing platform position measurement sensor and a remote sensing platform attitude measurement sensor; (3) acquiring single remote sensing image data of the same target object by registration synthesis of the remote sensing images; and (4) extracting target characteristics of the acquired single remote sensing image, determining a scanning line of the target, and solving the movement information of the target object according to tiny time-space shift characteristics (i.e. space shift parameters and time shift parameters) of the moving target object on pixels of the imaging sensing sensors. The extraction method can be widely applied in the fields such as large-scale remote sensing application, traffic status monitoring, target movement information extraction, etc.

The invention provides a traffic intersection RPP point automatic extraction method and device. The method comprises the steps: acquiring acquired laser point cloud data, vehicle trajectory data and binocular image data; extracting a bounding box of an intersection lane divergence or convergence area in the binocular image through a target detection algorithm, and marking an image containing the bounding box; performing depth estimation on the binocular image corresponding to the position of the track point, and predicting the depth of the binocular image; constructing a three-dimensional point cloud according to the binocular image corresponding to the position of the track point and the depth of the binocular image, fusing the three-dimensional point cloud and the laser point cloud data,and predicting a region of interest of an RPP point in the fused three-dimensional point cloud; and generating a top view of the point cloud near the RPP point based on the fused three-dimensional point cloud, and extracting the RPP base point from the region of interest of the top view. According to the scheme, automatic extraction of the intersection RPP points can be achieved, manual participation is not needed, the intersection RPP point element extraction efficiency is high, the precision is high, and the high-precision map manufacturing cost can be effectively reduced.

The invention relates to the technical field of computer natural language processing, and discloses an entity relationship extraction method and device for an open domain and terminal equipment. The method comprises the steps of obtaining to-be-processed open domain data; inputting the open domain data into a trained entity relationship extraction model to obtain an entity relationship in the opendomain data; wherein the entity relationship extraction model is trained by adopting a supervised learning algorithm. According to the invention, the accuracy of an open domain-oriented entity relationship extraction result can be improved.

The invention discloses a highly adaptive time delay target signal automatic extraction method, which designs general atoms, does not need to construct a huge redundant dictionary in advance during sparse decomposition, sets atomic characteristic parameters as bat individual position vectors, and constructs flexible atoms by using an improved bat algorithm to perform optimal position search, and further achieves standardized optimal matching atom search; a sparse decomposition dual-threshold iteration stop criterion is adopted, wherein a residual l-2 norm threshold is adaptively determined by background noise intensity, namely, a l-2 norm not containing a target signal sequence ns; the iteration number threshold value is set by the user according to the real-time requirement. The application of double criteria can ensure the signal extraction precision, improve the iteration speed and effectively avoid excessive decomposition. The method does not depend on complex prior knowledge and subsequent manual judgment and screening, other components of the signal can be reserved to the maximum extent while the target signal is effectively extracted, and the method is a fully-automatic non-stationary time delay signal extraction method.

The present invention belongs to the technical field of computer visual measurement and relate to an energy linear variation mark point and an accurate and stable mark point extraction method. The mark point is circular as a whole, the gray scale is linearly varied from the center to the edge, and the center has highest brightness and highest gray scale. The mark point extraction method comprises: firstly, performing crude extraction on the center of the mark point, and by utilizing the symmetry of a graph of the whole mark point, performing crude extraction with a gray-scale centroid method to extract a centroid of the mark point; secondly, performing elliptical cone fitting by using gray-scale linear information of the whole mark point rather than single boundary information; and thirdly, fitting out an optimal elliptical cone with a least square method to obtain precise center coordinates of an energy mark point. According to the energy linear variation round mark point and the extraction method, two extraction methods are used to ensure extraction precision; and compared with a conventional mark point, the energy linear variation round mark point and the fitting method have wider and preciser applications, and mark point extraction precision and robustness are improved.

The invention relates to a forest land remote sensing image extraction method and system based on a region-of-interest network. The method comprises steps of acquiring forest land survey data and multispectral wide-coverage image data; performing data preprocessing on the forest land survey data and the multispectral wide-coverage image data to obtain preprocessed forest land survey data and preprocessed multispectral wide-coverage image data; according to the preprocessed forest land survey data and the preprocessed multispectral wide-coverage image data, obtaining a sample set based on a geographic space corresponding relationship; constructing a region-of-interest network; training the region-of-interest network according to the sample set to obtain a trained region-of-interest network;acquiring a remote sensing image to be extracted; taking a remote sensing image to be extracted as input, and inputting the remote sensing image to be extracted into the trained RON to obtain a forest land extraction image; and optimizing the image to obtain an optimized forest land extraction image. According to the invention, high extraction precision of the forest land in different time phaseimages can be ensured.

The embodiment of the invention discloses a target detection method and device, electronic equipment and a storage medium. The method comprises: acquiring original point clouds and voxel grids corresponding to the original point clouds, matching original points in the original point clouds to corresponding voxel units in the voxel grids respectively, and storing point cloud information of the original points matched to the voxel units in each voxel unit; determining a to-be-extracted point cloud according to the size information of the voxel grid and the information of each point cloud in each voxel unit, and inputting the to-be-extracted point cloud into a trained point cloud feature extraction network; according to an output result of the point cloud feature extraction network, obtaining a point cloud feature of the original point cloud; and obtaining a target detection result of the to-be-detected target in the original point cloud according to the point cloud features. According to the technical scheme provided by the embodiment of the invention, the to-be-extracted point clouds containing the point cloud information in the voxel grids are taken as a whole for point cloud feature extraction, and the effect of quickly extracting the point cloud features is achieved.

The invention discloses an echo signal separation method for ultrasonic flaw detection. The method comprises steps of: first, acquisition and synchronous uploading of an ultrasonic echo signal: an ultrasonic flaw detection device performs ultrasonic detection on a tested object, acquires an ultrasonic echo signal F(t) of the tested object, and synchronously transfers the acquired ultrasonic echo signal F(t) to a data processing device ,wherein the ultrasonic echo signal F(t) comprises ultrasonic echo signals of positions of K flaws of the tested object; second, signal sparse and decomposition;and third, echo signal separation: the data processing device separates the ultrasonic echo signal F(t) according to K optimal matched atoms in an iteration decomposition optimal atom set, and acquires K separation signals. The method has simple steps, is reasonable in design, is easy to realize, has good use effect, and conveniently, rapidly separates overlaid signals in multiple flaw positionsin the same ultrasonic echo signal through signal sparse decomposition.

The invention is applicable to the technical field of digital image processing, and provides a light bar center extraction method, terminal equipment and a storage medium, and the method comprises the steps: obtaining an original image comprising a light bar; extracting the maximum gray value of each row or each column in the original image, and determining an initial coordinate of a light bar center according to the maximum gray value; calculating the offset of the light bar center initial coordinate in the original image according to a preset model; and calculating the actual coordinates of the light bar center in the original image according to the offset and the initial coordinates of the light bar center. According to the light bar center extraction method, the terminal equipment and the storage medium provided by the embodiment of the invention, the detection time can be shortened while the detection precision is improved, and the problem that the detection efficiency and the detection precision cannot be considered at the same time when the light bar center coordinates are extracted at present is solved.