109 results about "Scale variation" patented technology

A face feature analysis which begins by generating multiple face feature candidates, e.g., eyes and nose positions, using an isolated frame face analysis. Then, a nostril tracking window is defined around a nose candidate and tests are applied to the pixels therein based on percentages of skin color area pixels and nostril area pixels to determine whether the nose candidate represents an actual nose. Once actual nostrils are identified, size, separation and contiguity of the actual nostrils is determined by projecting the nostril pixels within the nostril tracking window. A mouth window is defined around the mouth region and mouth detail analysis is then applied to the pixels within the mouth window to identify inner mouth and teeth pixels and therefrom generate an inner mouth contour. The nostril position and inner mouth contour are used to generate a synthetic model head. A direct comparison is made between the inner mouth contour generated and that of a synthetic model head and the synthetic model head is adjusted accordingly. Vector quantization algorithms may be used to develop a codebook of face model parameters to improve processing efficiency. The face feature analysis is suitable regardless of noise, illumination variations, head tilt, scale variations and nostril shape.

A method of performing bilateral filtering and using the method for displaying high-dynamic-range images is presented. The method reduces the contrast of the image while preserving detail of the image. The presently disclosed method incorporates a two-scale decomposition of the image into a base layer encoding large-scale variations, and a detail layer. The base layer has its contrast reduced, thereby preserving detail. The base layer is obtained using an edge-preserving bilateral filter. The bilateral filter is a non-linear filter, where the weight of each pixel is computed using a Gaussian in the spatial domain multiplied by an influence function in the intensity domain that decreases the weight of pixels with large intensity differences. The bilateral filtering is accelerated by using a piecewise-linear approximation in the intensity domain and appropriate subsampling.

A methodology is described to reduce the complexity of filters for face recognition by reducing the memory requirement to, for example, 2 bits/pixel in the frequency domain. Reduced-complexity correlations are achieved by having quantized MACE, UMACE, OTSDF, UOTSDF, MACH, and other filters, in conjunction with a quantized Fourier transform of the input image. This reduces complexity in comparison to the advanced correlation filters using full-phase correlation. However, the verification performance of the reduced complexity filters is comparable to that of full-complexity filters. A special case of using 4-phases to represent both the filter and training/test images in the Fourier domain leads to further reductions in the computational formulations. This also enables the storage and synthesis of filters in limited-memory and limited-computational power platforms such as PDAs, cell phones, etc. An online training algorithm implemented on a face verification system is described for synthesizing correlation filters to handle pose/scale variations. A way to perform efficient face localization is also discussed. Because of the rules governing abstracts, this abstract should not be used to construe the claims.

There is provided an environment recognizing device for a vehicle capable of reducing false detections for an artificial object such as a utility pole, a guardrail and road paintings with smaller processing load, at the time of detecting a pedestrian by using a pattern matching method. The environment recognizing device for a vehicle includes an image acquisition unit (1011) for acquiring a picked up image in front of an own vehicle; a processing region setting unit (1021) for setting a processing region used for detecting a pedestrian from the image; a pedestrian candidate setting unit (1031) for setting a pedestrian candidate region used for determining an existence of the pedestrian from the image; and a pedestrian determination unit (1041) for determining whether the pedestrian candidate region is the pedestrian or an artificial object depending on a gray-scale variation rate in a predetermined direction within the pedestrian candidate region.

To stop the decline of the quality of image associated with encoding. The present invention multiplies a decoding rescaling factor (RFr) possibly used in the decoding process by only a transformation matrix (D), which is scale change, to calculate a rescaling factor (RF), which is a plurality of division factors, and then calculates, for each detection unit, the sum (ΣY) of evaluation values (Y) based on a residue (r) obtained as a result of dividing an element of a DCT coefficient by a plurality of rescaling factors (RF). Moreover, the present invention compares correlations of the sum (ΣY) of the evaluation values (Y) with a plurality of rescaling factors (RF), and detects, based on the rescaling factor (RF) whose sum (ΣY) of the evaluation values (Y) is a minimum value, a quantization factor used in the previous process of encoding the input image data.

The invention relates to a scale adaptive kernel correlation filter fast target tracking method. According to the scale adaptive kernel correlation filter fast target tracking method, separation of afront background is promoted and a scale estimation module is added through introducing a two-dimensional variable bandwidth Gaussian window into a KCF framework, scale variations of a target are estimated according to a matching result of feature points of an image block, and whether the target is sheltered is judged by combining displacement prediction based on the scale variations. Once the target is sheltered, the user considers that no scale conversion occurs at the moment and template update is not performed, further the tracking precision is enhanced in this way. The scale adaptive kernel correlation filter fast target tracking method has the advantages of simple and fast calculation, and can significantly improve the calculation efficiency on the basis of compatible scale adaptation.

The invention discloses a device and a method for positioning a wheel hub work piece. The device for positioning the wheel hub work piece comprises an image collection module, a wheel hub template information extraction module, a feature point extraction module of a wheel hub to be detected, a feature point matching module and a wheel hub positioning module, wherein the wheel hub template information extraction module is used to extract SIFI (scale invariant feature transform) feature points, a circle centre and positions of air nozzles and four points on the circumference of the outer edge of the wheel hub from a wheel hub template image. The device and the method for positioning the wheel hub work piece are in consideration of illumination influences and the problems of translation, rotation, scale variation and the like, which are encountered in the wheel hub image matching process, use an SIFI feature point matching method to match point pairs of corresponding spaces in the wheel hub template image and an image to be detected, then judge a space corresponding relation in wheel hub area images in the wheel hub template image and the image to be detected through the point pairs, and finally figure out points in a wheel hub area in the image to be detected, which are corresponding to known calibration points in the wheel hub template image, through the space corresponding relation between the wheel hub area images in the wheel hub template image and the image to be detected, and thereby achieve the purpose of positioning the wheel hub.

The invention discloses an image segmentation method based on a rapid density clustering algorithm. The image segmentation method comprises the following steps: 1) for a natural image to be processed, firstly carrying out preprocessing and initialization, comprising filtering noise reduction, gray-level registration, area dividing, scale zooming and the like; 2) then carrying out calculation of similarity distance between data points on sub-graphs completing scale variation, and obtaining correlation between pixel points; 3) then carrying out concurrent segmentation processing in each sub-graph, comprising drawing a decision graph based on the density clustering algorithm, carrying out residual analysis to determine a clustering center based on the decision graph and comparing based on the similarity distance to classify remaining points on an original scale sub-graph; and 4) then merging the sub-graphs after the segmentation is completed, and carrying out secondary re-clustering to obtain a segmentation result graph with original size dimensions. The image segmentation method based on the rapid density clustering algorithm for parameter robust provided by the invention can automatically determine the number of segmented classes to realize relatively high segmentation accuracy rate.

The invention discloses a logo detection method based on spatial connected domain prepositioning. The logo detection method disclosed by the invention comprises the following steps: creating a logo picture sample library comprising multiple logo pictures; creating spatial connected domain descriptor (SCCD) characteristics for the logo pictures in the logo picture sample library; for a test picture containing a target logo, creating connected domain descriptor (CCD) characteristics of the test picture; acquiring logo prepositioning regions (LPRs) in the test picture by utilizing the SCCD characteristics of the logo pictures in the logo picture sample library; and matching the LPRs with the logo pictures in the logo picture data library by utilizing the characteristics which are created based on logo colour and shape information, wherein the successfully matched logo prepositioning region is namely a finally detected logo. The logo detection method disclosed by the invention fully directs at the characteristics of a logo, speed and accuracy of logo detection and positioning identification are further improved, and detection on the logo, which is influenced by scale variation, visual angle transformation, illumination, overlap and background interference, can have good effect.

The invention relates to a self-adaptation mean-shift target tracking method based on optical flow field estimation. Aiming to solve the problem that the tracking fails due to high motion speed of the target or obvious scale variation and shielding of the target during target tracking of the mean-drift shift algorithm, a light stream method is introduced on the basis of a traditional mean-shift vector method, feature points are searched on the target, the centre position and size of a window are modified and tracked based on the variation information of the feature point, and more accurate length and width of the window can be obtained through self-adaptation by a Bhattacharyya coefficient bisection method. The area of the object shielded by a stationary object can be observed through aberration analysis, and the object can be recaptured by using the Bhattacharyya coefficient.

The invention relates to a method for image mosaic based on a feature detection operator of second order difference of Gaussian, which carries out mosaic on sequence images which vary in viewpoint, rotation, proportion, illumination and the like to a certain degree. The method provided by the invention adopts zero crossing point detection of second order difference of Gaussian (D<2>oG) pyramid to replace extreme point detection of the original difference of Gaussian (DoG) pyramid so as to extract scale invariant feature points, thereby effectively simplifying the structure of the Gaussian pyramid. The method comprises the steps of: first, extracting image feature points by using an improved SIFT (Scale Invariant Feature Transform) algorithm; then, searching a rough matching point pairs for the extracted feature points through a BBF (Best-Bin-First) algorithm, and purifying the feature point matching pairs by adopting an RANSAC algorithm so as to calculate an invariant transformation matrix H; and finally, completing seamless mosaic for the images by adopting a fading-in-and-out smoothing algorithm. Experimental results show that the method improves the accuracy and the real-time performance of image mosaic, can well solve problems such as illumination, rotation, scale variation, affine and the like, and realizes automatic mosaic without manual intervention.

The invention relates to a long-time sheltering robust tracking method based on convolutional features and global search detection. Through adopting a convolutional feature and multi-scale relative filtering method in a tracking module, the feature representation capacity of a tracking object appearance model is enhanced, so that a tracking result is very robust to factors such as illumination variation, object scale variation and object rotation. In addition, through the introduced global search detection mechanism, a detection module can detect the object again when the tracking fails because the object is sheltered for a long time, thus the tracking module can recover from error, and the object can be continuously tracked for a long time even in the case of object appearance changes.

The invention discloses a multiscale invariable oriented FAST and rotated BRIE (ORB) algorithm used for stitching images. The algorithm mainly comprises the following steps of firstly, using a featuredetection algorithm based on scale invariability to detect feature points of the image; then using an ORB descriptor to describe features of the detected feature points, then using an ORB matching algorithm to perform rough matching, and using a bidirectional matching method and a RANSAC (random sampling consistency) algorithm to perform fine matching and purifying on matched points, thus furtherimproving accuracy; and at last completing image stitching by using gradual weighted fusion. The method provided by the invention effectively solves that the problem that image feature points detected by the ORB algorithm is lack of the scale invariant feature and high in mismatching rate, and thus the stitched image is poor in quality, and has good robustness and stability for different light conditions, angle rotation, low resolution and scale variation of the image, and is the image stitching method that is short in consumed time, high in accuracy and good in stitching effect.

A trademark density based personalized trademark matching recognition method includes: establishing a trademark image sample library, extracting SIFT (scale invariant feature transform) features, calculating the trademark density, and setting personalized matching strategies according to the trademark density; extracting the SIFT features and GSC (generalized shape contexts) features for trademark images in the sample library and establishing an SIFT word bag model and a GSC word bag model; extracting the SIFT features and the GSC features for a to-be-recognized image area, and respectively calculating similarity lists of remarks in the SIFT word bag model and the GSC word bag model; combining the two similarity lists of the trademarks according to the personalized matching strategies to acquire the all the similarity lists of the trademarks of the area, and taking the trademarks with the largest similarity and meeting the similarity thresholds as the recognition result of the area. The trademarks are recognized rapidly and accurately, and good effect of detection under the influence of scale variation, focus change, illumination, shielding, background interference and the like on the trademark can be achieved.

The invention discloses a method for self-adaptation geometric correction of a projector picture. The method for self-adaptation geometric correction of the projector picture comprises the steps of (1) shooting a projection plane image without a projection picture, (2) projecting a target original image to be corrected on a projection plane to obtain a current projection picture, (3) shooting a projection plane image of the current projection picture, (4) collecting point cloud sample data of the projection plane, (5) obtaining a homography matrix from a camera plane to the projection plane, (6) identifying a frame number which corresponds to the current projection picture, searching for an original projection image corresponding to the current projection picture, (7) obtaining the corresponding relation, between the projection plane image of the currently existing projection picture and the original projection image of the currently existing projection picture about two-dimensional position coordinates with the matching characteristics, obtaining a homography matrix from the projector plane to the camera plane, obtaining a homography matrix from the projector plane to the projection plane, (8) carrying out scale variation processing on a target original image, to be corrected, of a next frame to generate a variation projection image, and (9) projecting the variation projection image on the projection plane to obtain a new current projection picture.

The invention belongs to video and image processing technology and particularly relates to a method for accurately tracking characteristic points of a quick movement target. The method comprises the following steps of: selecting M characteristic points in a tracked target selection rectangular region at the moment of t-1 according to characteristic matrixes; iterating the M characteristic points at the moment of t by using a KLT luminous flux vector formula to solve the optimal solution so as to acquire the new positions of the current frames of all the characteristic points; and estimating the proximity positions of all the characteristic points by using a checker in a tracked target searching rectangular region if the optimal solution cannot be solved so as to acquire new positions. Due to the adoption of the method, the proximity positions can be found, then the optimal position can be solved by the luminous flux vector calculation formula, and a search region is partitioned by using the checker, so that search time is saved greatly, and the proximity positions with approximate gray scale variation to that of the tracked target can be searched. The method avoids the situation of losing the tracking of the characteristic points and is very suitable for tracking the characteristic points of the quick movement target.

The invention provides a fast document type recognition method based on full-sized feature extraction. The method comprises the following steps of document image preprocessing, including image zooming, graying and noise filtering; document image feature extraction, including Hessian matrix building, scale space generation, primary determination of feature points, precise positioning of the feature points, main direction determination of selected feature points, feature point descriptor construction and feature value string generation; document image feature value comparison, including document similarity calculation and comparison algorithm optimization. According to the method, an image is preprocessed through software, and additional hardware equipment does not need to be added. According to the method, the scale-invariant feature is creatively introduced for improving a typical SURF feature extraction algorithm, so that the problem of matching failure due to error amplification of the SURF algorithm caused by scale variations is fundamentally solved. The method has the advantage that a multi-thread technology and a large cache are used for solving the problems of large data volume calculation during comparison and the harsh time requirement of a user on an electronic government affair platform.

The invention discloses a high-precision real-time stereoscopic visual positioning method utilizing the parallax space bundle adjustment. According to the invention, only a video streaming obtained by a stereo camera is utilized to carry out navigation positioning of a mobile robot, so that the high-precision real-time stereoscopic visual positioning method has complete autonomy; a U-SURF descriptor is utilized to carry out feature matching, so that the high-precision real-time stereoscopic visual positioning method has strong robustness on illumination variation, motion blur and large rotation and scale variation which are generated in the moving process; and due to the utilization of a parallax space bundle adjustment optimizing method, an autonomous navigation vehicle can also be subjected to accurate real-time positioning when traveling for a long distance, so that the high-precision real-time stereoscopic visual positioning method has the advantages of low degree of dependence on an optimal initial value, high convergence rate, high value stability and the like and compared with a stereoscopic visual positioning method utilizing the bundle adjustment, the high-precision real-time stereoscopic visual positioning method has higher positioning precision and higher speed. The integral stereoscopic visual positioning method disclosed by the invention has high calculating speed and can be operated on line in real time.

The invention provides a liquid crystal display device. The liquid crystal display device comprises a pixel array consisting of a plurality of sub-pixel arrays arranged in an array way; each sub-pixel array comprises a structure which contains two rows and two columns of pixels and consists of four pixel electrodes, four thin film transistors, a data line and four sequentially arranged scanning lines; the sources of the four thin film transistors share the data line; the drain of each thin film transistor is connected with a pixel electrode; the grids of the two thin film transistors of one column are connected with the two scanning lines in the middle respectively, and the grids of the two thin film transistors of the other column are connected with the other two scanning lines; and whenthe liquid crystal display device is driven, the four scanning lines are used for scanning sequentially. As the pixel electrodes having gray scale variation are arranged on the different columns, a vertical bar phenomenon due to non-uniform periodic gray scales can be avoided; and the aperture opening ratio can be guaranteed.

The invention discloses a DHOG and discrete cosine transform-based overhead line system equilibrium line fault detection method. The method includes an equilibrium line and equilibrium line supporting seat recognition and positioning process and an equilibrium line default fault detection process. The equilibrium line and equilibrium line supporting seat recognition and positioning process includes the following steps that: DHOG feature extraction is performed on a positive sample and a negative sample; and an AdaBoost classifier is trained; and a classifier obtained through training is utilized to accurately judge whether a target area is contained in a slide detection window. The equilibrium line default fault detection process includes the following steps that: the target feature information of the equilibrium line and a supporting seat thereof is enhanced through utilizing discrete cosine transform; a two-step method which performs circle detection based on arc piecewise fitting is adopted to perform statistics on circles in an image; and an overhead line system equilibrium line default fault can be judged according the number of the circles. The DHOG and discrete cosine transform-based overhead line system equilibrium line fault detection method has the advantages of high detection accuracy and excellent robustness under a situation in which rotation and scale variation occur on the target image, and can provide a certain technical reference for intelligent inspection of high-speed rail overhead line systems.

The invention discloses a synchronization device and method of a high dynamic UAV data link. An input cache module receives and caches data. Related device group modules carry out part correlation on the cached data. A parallel N-FFT module carries out FFT on a related result. A threshold judgment module carries out threshold judgment on a parallel output spectrum value. A Doppler estimation module estimates a Doppler frequency offset amount and a signal scale variation amount. A data interception module intercepts input data from a valid data start bit. An interpolation/extraction module carries out interpolation or extraction compensation to complete Doppler time domain compensation, and a reference phase storage module compensates an instantaneous phase deviation caused by the Doppler frequency offset. The device and the method can be used for the communication synchronization between a UAV and a ground station and between UAVs, the change of a Doppler effect to frequency offset and signal scale changes can be estimated and compensated in real time, and the reliability of communication link synchronization establishment can be improved to a maximum degree.