39 results about "Boundary extension" patented technology

The present invention discloses a three-dimensional model retrieval system and method based on parts division of a hand-drawn draft. The system comprises: a preprocessing module, a part marking module, a sampling point feature extracting module, a parts division module, and a similarity calculation and total score sequencing module, wherein the preprocessing module is used for denoising a hand-drawn query draft to obtain a grayscale image, and performing binarization processing, boundary extension processing and image hole filling processing on the grayscale image to obtain a processed image; the part marking module is used for performing equal-interval sampling on the processed image, and adding part tags at sampling points; the sampling point feature extracting module is used for extracting various feature vectors of the sampling points; the parts division module is used for performing division model training according to the various feature vectors of the sample points to which the part tags have been added; and the similarity calculation and total score sequencing module is used for performing local similarity calculation on parts, performing sequencing according to total scores, and feeding back a sequencing result to a client. Implementation of the embodiments of the present invention can make three-dimensional model retrieval based on the hand-drawn draft more accurate and effective.

The invention provides a lifting scheme-based optimization implementation method of 9/7 wavelet transformation. In the method, the times of shift operations in the computational process of a lifting scheme is effectively reduced by selecting a wavelet transformation lifting parameter which is suitable for being implemented on a high-speed hardware platform; and on the basis, an embedded boundary extension algorithm is adopted, an intermediate result cache region is established and the cache region initialized by using original data, so that stream processing capability is improved, accessing times are reduced, the time complexity of algorithm is reduced, saves memory space, and the space complexity of the algorithm is reduced. The experimental result shows that for the same image, the method and a CDF9/7 wavelet filter-based compression algorithm can achieve basically the same image recovery quality in a compression test, but the method is faster in operation.

A method and system for authenticated dynamic extension of security domain boundaries includes high security domain extension instructions for sequentially and dynamically forming an extended high security domain (133) through a protected communication path (128). The protected communication path (128) extends from a first computer (10) associated with a high security domain (80) into a second computer (10) associated with low security domain (120). The method and system establish the extended high security domain (133) within the second computer (10). A protected communication path (128) forms an isolation barrier (131) separating the extended high security domain (133) from other objects (126) within the low security domain (120). Authentication instructions (146) temporarily 20 authenticate at least one object (132) associated with the low security domain (120). Returning instructions (156) return the at least one object (132) processed within the extended high security domain (133) to said low security domain (120).

The invention discloses a method for extracting image characteristics by multivariate gray model-based bi-dimensional empirical mode decomposition, which belongs to the field of image processing and aims to solve the problem that image intrinsic information cannot be extracted effectively by the conventional bi-dimensional empirical mode decomposition method due to the influence of an end point effect. The method comprises the following steps of: performing boundary extension on a given image by using a multivariate gray model; performing the conventional bi-dimensional empirical mode decomposition on the extended image to acquire an extended bi-dimensional intrinsic mode function and a residual error; and extracting the bi-dimensional intrinsic mode function and the residual error at the position corresponding to the original image to be used as the final decomposition result. By the method, the advantages that a gray theory is low in data requirement quantity, high in short-term forecast accuracy and high in calculation speed and does not have particular requirements on the original data distribution are fully exerted, and the method is easy to implement, popularize and apply.

The invention provides an LED display control method, an image splicing edge optimization method and an image splicing processing device. Specifically, through adoption of a boundary extension mode, boundary extension is carried out an image interception range, so image data intercepted from input image data comprises the image data of the other side of the imaging splicing edge; the corresponding boundary extension is carried out on a screen mapping window, so the scaling of an image is kept unchanged; and through combination of an original image optimization algorithm of a video processing chip, integrated image optimization is carried out on the intercepted image data. In this way, the video processing chip can optimize the image data out of own display screen load area, thereby solving the color distortion condition of the splicing edge and improving the quality of the image of which splicing edge is processed.

The invention relates to a low power consumption parallel wavelet transforming VLSI structure, wherein the low power consumption parallel wavelet transforming VLSI structure is connected with two parallel delay cells and a data stream selector or two shunt-wound shift registers and a data stream selector in series before each adder of the data stream in a VLSI structure which directly realizes discrete wavelet transformation; an embedded boundary extension circuit is parallelly and synchronously connected with a structure in which two shunt-wound delay cells/shift registers and a data stream selector form a group. The invention can process two lines of data synchronously to realize time sharing multiplex of main arithmetic units with the hardware utilization rate up to 100%, and reduce the required cache on chips and the access to the exterior memory, thereby efficiently decreasing the power consumption of the whole design. Due to the simple hardware structure, the VLSI is easy to realize.

The invention discloses a parallel real-time SAR image spot and noise reducing method based on multiple DSPs. The parallel real-time SAR image spot and noise reducing method based on the multiple DSPs mainly solves the problems that large-scale SAR image spot reducing operation is large in time complexity and low in processing speed. The parallel real-time SAR image spot and noise reducing method based on the multiple DSPs comprises the implementation steps that (1) a main processor and an auxiliary processor are initialized; (2) an upper computer divides a large-scale SAR image into a plurality of sub-images and transmits the sub-images to the main processor and the auxiliary processor; (3) the main processor and the auxiliary processor carry out boundary extension and floating-point type transformation, then calculate the expectation filter weight of a central image block and the expectation filter weights of neighborhood image blocks and obtain an estimation result of the central image block according to the weight; (4) aggregation, boundary clipping and integer transformation are carried out on the estimation result in sequence, and sub-image spot reduction results are obtained; (5) all the sub-image spot reduction results are transmitted back to the upper computer and spliced, and thus a spot reduction result image of the original large-scale SAR image is obtained. The parallel real-time SAR image spot and noise reducing method based on the multiple DSPs greatly improves spot reduction speed, can meet the real-time performance requirement of image spot reduction, and can be used for large-scale SAR image fast spot reduction processing and SAR image target recognition.

The invention relates to a real-time evaluation method of human body motion sequences. Boundary extension is carried out on the motion sequences by employing interpolating wavelet based on difference, thereby reducing boundary effect; a wavelet coefficient is calculated, thereby obtaining key frames; features of the motions are described through the key frames; the distance between two motion sequences is calculated by employing a DTW distance measurement function; finally the distance is normalized, thereby obtaining a similarity score of the two motions. The method has the advantages that the boundary effect is reduced by employing the interval interpolating wavelet, thereby obtaining the more accurate key frames, and further the more accurate motion evaluation is realized.

The present invention provides an inscription oriented Chinese character extracting method, mainly comprising two parts: a Chinese character region positioning method based on Area Voronoi and a single character segmentation method based on a communication field and a bounding box; and the method comprises the steps of: firstly, for a distribution rule and feature of Chinese characters in an inscription image, acquiring communication fields in the image as generators, and representing the contour of each communication field by using a Freeman encoding chain; constructing an Area Voronoi diagram of the image by using a contour boundary extension method; then according to a standard deviation between the communication fields, merging adjacent Area Voronoi regions so as to complete the positioning operation of a Chinese character region; after acquiring an image of the Chinese character region, performing pre-processing on the image and removing a certain noise influence; and by concluding structure features of the Chinese characters, combining a communication field analysis method with a bounding box overlapping judgment method, and performing single character extraction on the Chinese characters. According to the inscription oriented Chinese character extracting method provided by the present invention, for an arrangement rule of ancient inscriptions, the extraction work for a single Chinese character can be better completed, and the method has a certain noise immunity.

The invention relates to the technical field of smart home, and discloses a partition boundary extension method, device and equipment of a cleaning robot and a storage medium, which are used for improving the accuracy of partition boundary extension. The partition boundary extension method of the cleaning robot comprises the steps: establishing a preset partition in an environment map; determininga preset partition cleaning path according to the partition boundary of the preset partition; when the cleaning robot advances to the current partition boundary according to a preset partition cleaning path, obtaining distance data, map data and state data of an adjacent region of the current partition boundary; judging whether the adjacent region is a target region or not according to the distance data and the map data; if the adjacent region is not the target region, updating the state data, and continuously performing boundary extension detection according to the preset partition cleaningpath; and if the adjacent region is the target region, judging whether a preset extension condition is met or not according to the distance data, the map data and the state data, and if YES, carryingout boundary extension on the current partition boundary.

The invention discloses a motion compensation reference block data reading device and a motion compensation reference block data reading method for embedded video decoding. The device and the method are used for reading the reference block data of a current decoding macro block from an off-chip reference frame buffer into an on-chip memory space in the embedded video decoding. In the device and the method, inter-frame macro blocks are classified according to types and motion vectors of the macro blocks, different reference block data ranges are determined respectively, and the reference block data is read into the on-chip memory space once. Simultaneously, boundary extension is performed on data out of a reference frame range, so the reference block data is not required to be judged whether to exceed a reference frame boundary or not in subsequent sub-pixel interpolation. By the device and the method, on-chip and off-chip data interaction times can be remarkably decreased, the overlapped part of data of adjacent 4*4 reference blocks can be reduced, and an on-chip and off-chip interactive data volume can be decreased.

The invention discloses an image processing implementation method and system, and the method comprises the steps: obtaining a to-be-filtered original image which is H in height and W in width; whereinR is the radius corresponding to the mean value, the expansion operation core and the corrosion operation core, the window size of the corresponding core is 2R + 1, and R is greater than 0; performing upper and lower boundary extension processing on the original image according to the radius R; performing left and right boundary extension processing on the image after the upper and lower boundaryextension processing; and outputting the target image according to a preset performance function. According to the image algorithm provided by the invention, except that the column direction and theradius R are related to edge extension and first calculation, the rest mean filtering and expansion corrosion operations are irrelevant to the radius; according to the method, the required memory amount is small, the algorithm flow storage data is less, mean filtering, expansion and corrosion operations are realized by adopting a unified framework, and the performance and the application value ofthe graphic image algorithm are greatly improved.

In a forging process analyzing method, many trace points are set in a workpiece. Lines connecting the trace points express fiber flow lines. The fiber flow lines may be imaginary ones that are irrelevant to real fiber flow lines. By applying a technique of a finite element method to the fiber flow lines, deformation calculation is performed. A new trace point is added between trace points when the distance therebetween is great. The velocity of each trace point cause by deformation is calculated, and the position of each trace point is updated. When any trace point in an element is left behind in a die, the left-behind phenomenon is prevented by converting the shape of the element while the crossing of fiber flow lines is prevented. A radius is set for each trace point. The radius of each trace point is adjusted so as to eliminate off-boundary extension and overlap of fiber flow lines.

The invention discloses an interval threshold image denoising method based on BEMD, which is an interval threshold image denoising algorithm based on two-dimensional empirical mode decomposition. According to the method, boundary extension is carried out on an image to be denoised by using the self-similarity of the image, the end effect existing in a BEMD algorithm is effectively inhibited, and multi-scale decomposition is carried out on the extended image by using the BEMD algorithm to obtain a series of two-dimensional intrinsic mode functions. Interval threshold denoising is conducted on the BIMF components with the noise occupying the main components, and finally all orders of BIMF components are added and reconstructed to achieve the purpose of denoising. Experiments show that the denoising effect of the method is superior to that of an existing BEMD denoising algorithm, and compared with a traditional denoising algorithm, the method also has advantages.

The invention discloses a method and a device for automatically generating test cases, relates to the technical field of automatic testing, and can automatically expand and generate a plurality of test cases for test requirements containing boundary values. The method comprises the following steps: compiling a demand document, wherein the demand document comprises a test case boundary value; identifying the content of the boundary value of the test case in the demand document and storing the content in a key value pair mode, wherein the content comprises at least one test function name and a corresponding boundary value; reading a key value pair, and generating an extended value group corresponding to each boundary value after boundary extension is performed on each boundary value in the key value pair; and generating a plurality of test cases based on each test function name and the corresponding boundary value and extension value group. The system applies the method provided by the scheme.

The invention discloses a fast realization method for digital image non-local average denoising. The fast realization method specifically comprises the steps of: 1) selecting an appropriate searching window and similar windows, and determining a radius M of the searching window and a radius R of the similar windows; 2) performing boundary extension on an image according to the radii of the searching window and the similar windows as well as the size of R+M; 3) forming a large matrix A and large matrix B, wherein the matrix A stores results of the similar windows searching within the searching window, and the similar window in which a current central pixel locates is extended to form the matrix B according to the size of the matrix A; 4) estimating weights of the large matrixes, calculating the matrix A and the matrix B to determine similarity of the similar windows according to Euclidean distance, and further determining a weight allocated to a current pixel; 5) and performing weighted averaging on the large matrixes, and performing weighted averaging on the calculated weights to obtain a final estimation result of the current pixel. By adopting the fast realization method image non-local average denoising can be realized quickly, and the program operation time is greatly saved.

The invention specifically relates to a distortion-free boundary extension method for wavelet online denoising. The distortion-free boundary extension method comprises the steps of continuously intercepting a latest segment of data xn by adopting a sliding window to perform denoising, wherein xn is finite-length data with the length being N, temporarily storing the data xn, continuously sampling a length of time T, acquiring L pieces of future data f1 at the right boundary of the data xn so as to act as distortion-free extension of the right boundary of the data xn, acquiring M pieces of historical data hm of the left boundary of the data xn through pre-storage so as to act as distortion-free extension of the left boundary of the data xn, and completing information of the right boundary and the left boundary of the data xn. The distortion-free boundary extension method not only can suppress boundary effects in the segmented wavelet denoising process, but also can avoid artificially introduced algorithm interference through simple cache operations and time sequence arrangement, thereby effectively solving a problem of boundary distortion in the online implementing process of wavelet denoising.

An N type longitudinal silicon carbide MOS tube with high breakdown voltage comprises an N type substrate, wherein drain electrode metal and an N type drift region are arranged at two ends of the N type substrate respectively; a P type base region layer is arranged on the N type drift region; N type source regions and P type body contact regions distributed according to an array are arranged on the P type base region layer; N type drift region bulges are arranged between adjacent N type source regions; boundaries of the N type source regions extend into corresponding breakdown voltage improving regions; the breakdown voltage improving regions are rectangular regions formed by inner side boundary extension cords of four adjacent N type source regions; gate oxide layers are arranged on the N type drift region bulges, and boundaries of the gate oxide layers extend outwards and stop at the boundaries of the N type source regions; poly-silicon gates are arranged on the gate oxide layers; field oxide layers arranged on the poly-silicon gates and the N type source regions; the N type source regions and the P type body contact regions are connected with source electrode metal; and the surfaces of the poly-silicon gates are connected with gate metal.

The present invention discloses a liquid crystal display panel and a driving method thereof, and a liquid crystal display, which the driving method comprises: normalizing an image to be displayed first, to obtain characteristic data of the input image; extending backlight boundary of each backlight partition according to a default condition, and determining a backlight opening coefficient of each backlight partition by data information after the backlight boundary extension; the backlight opening coefficient of each backlight partition subjected to a fusion process, to obtain a driving current of the liquid crystal display panel; obtaining characteristic data of output image of each backlight partition based on the characteristic data of the input image and the driving current; displaying the output image by the driving current and the characteristic data. It is possible to display high-quality picture information in the case of optical crosstalk existing in the liquid crystal display panel.

A technique for coarsening a finite element mesh (“FEM”) is described. This technique includes identifying a coarsening region within the FEM to be coarsened. Perimeter chords running along perimeter boundaries of the coarsening region are identified. The perimeter chords are redirected to create an adaptive chord separating the coarsening region from a remainder of the FEM. The adaptive chord runs through mesh elements residing along the perimeter boundaries of the coarsening region. The adaptive chord is then extracted to coarsen the FEM.

Systems and methods for computer simulation for determining a field generated from a source, with the field interacting with one or more structures. The systems and methods comprise dividing a domain into subdomains, solving iteratively for the field in a subset of the subdomains by solving for a residual field within an extended subdomain around each subdomain within the subset. If the subdomain comprises a structure, the boundary of the structure extends beyond the boundary of the extended subdomain to a second extended subdomain.

The invention discloses a novel hoisting and dropping machine. The novel hoisting and dropping machine is characterized in that the mounting position of an electric control box body is changed, and the electric control box body is installed above an air inlet grid; during maintenance, the whole electric control box body does not need to be disassembled, and the electric control box body can be maintained only by removing the air inlet grid; and by designing the upper boundary extension line A of a volute air outlet to coincide with the upper edge of an air outlet of the hoisting and dropping machine and designing the lower boundary extension line B of the volute air outlet to coincide with the lower edge of the air outlet of the hoisting and dropping machine, the interference between the upper boundary extension line A and a shell of the hoisting and dropping machine and between the lower boundary extension line B and a water receiving tray can be effectively avoided, and vortex formation can be avoided, so that the noise is reduced effectively, and the air outlet efficiency of the whole machine can be improved.