33results about How to "Reduce the scope of calculation" patented technology

A method for compressing data in a video sequence, a system for carrying out the method, a computer program, and a computer program product are provided. In the described method, results of a motion estimation for a previous temporal decomposition stage are also used for motion compensation.

Provided is a new word discovery method and device. The method comprises the steps that pretreatment is conducted on received corpora, and text data are obtained; line division is conducted on text data, and statement data are obtained; word segmentation is conducted on the statement data according to individual words contained in a dictionary, and after word segmentation is conducted, word data are obtained; after word segmentation is conducted, combination is conducted on adjacent word data to generate candidate data strings; judgment processing is conducted on the candidate data strings to discover new words; judgment processing comprises the steps that information entropy of words in the candidate data strings and words outside the candidate data strings is calculated, and the candidate data strings of which the information entropy of the words and the outside words is out of the preset range are removed. By means of the new word discovery method and device, the accuracy of new word discovery can be enhanced.

The invention requests to protect a vehicle detection and tracking method based on monocular vision, and belongs to the field of machine vision. The method comprises the steps that in a vehicle detection stage, a rod driving region is determined by combining interest region extracting, self-adaptive Canny edge detection and lane line detection, then a vehicle bottom shadow is obtained by means of local tonal value statistics and a thresholding method in which a maximum between-class variance method is used twice in an integrated manner, furthermore, a supposed vehicle is proposed, and then the supposed vehicle is verified through a texture description co-occurrence matrix method; and in a vehicle tracking stage, an improved algorithm combining Kalman filtering with Cam shift is adopted to carry out multi-target tracking, then new target judging, whether searching is successful, and whether a vehicle is out of an edge are used as three standards, and if a new target is detected twice, the new target is processed as a new tracking target, and the tracking target is updated continuously. By adopting the method, vehicle detection and multi-target tracking under a dynamic background are realized, and the instantaneity, the accuracy and the reliability are relatively high.

Provided are dictionary updating method and apparatus. The dictionary updating method comprises: performing preprocessing on a received corpus to obtain text data; performing line segmentation on the text data to obtain sentence data; performing word segmentation on the sentence data according to individual words contained in a basic dictionary, so as to obtain word data after word segmentation; performing combination processing on the adjacent word data after word segmentation, so as to generate a candidate data string; performing determination processing on the candidate data string to discover a new word; and if the new word is discovered, adding the new word into the basic dictionary to update the basic dictionary. The dictionary updating method and apparatus can reduce dictionary maintenance costs and improve dictionary update efficiency.

The invention provides a pipeline leakage detection method and device and relates to the field of pipeline leakage monitoring. The pipeline leakage detection method comprises the steps of obtaining anegative pressure wave signal caused by pipeline leakage; preprocessing the negative pressure wave signal to obtain the dynamic change sequence of pressure along a pipeline; determining the interval of a leakage point on the basis of the dynamic change sequence of pressure along the pipeline; and obtaining the position of the leakage point on the basis of dynamic change sequences of pressure at the two ends of the interval of the leakage point. According to the method and the device, the pipeline is detected in a segmented manner by multiple sensors arranged in the pipeline so as to determinethe interval of the leakage point, then the leakage point is positioned on the basis of the interval of the leakage point, the positioning precision is high, the site operability is high, and the method and the device can be widely applied to leakage detection of long oil and gas transmission pipelines and city water supply and heat supply pipelines.

A map of area is created based on natural landmarks (14,16), and the preset route along which the vehicle (10) is to move is set. The landmarks that serve as a route along preset location is determined, and the environment (12) is scanned at different times using a sensor (22) for detecting the preset landmarks while the vehicle moves along preset route. The vehicle is located by comparing the detected landmarks with destination shown on map landmarks. The speed and / or direction of rotation of sensor motor (24) are controlled such that sensor is aligned on few preset landmarks.

The invention discloses a collaborative filtering processor. A method includes the following steps: adopting an improved K-means algorithm to perform clustering on users; selecting a cluster formed by the users similar to a target user in attribute; acquiring a nearest neighbour set of the target user; generating a recommendation set. The main idea includes that the users identical in feature are clustered into a class through clustering to lower dimensionality of a matrix and reduce spatial complexity, sparse matrices are filled through matrix decomposition, and then collaborative filtering is performed on the filled matrices.

The invention relates to a money checking method and system. The method comprises the following steps: converting a UV image of paper money into a binary image; inquiring a maximum value position column I in binary image column projection, and substituting first pixels in the maximum value position column I with second pixels; circularly inquiring a maximum value position column II in the binary image column projection within a preset inquiring frequency threshold value, and substituting first pixels in the maximum value position column II with second pixels if the quantity of the first pixels is greater than a quantity threshold value, wherein the maximum value position column I is a position column with the projection in a magnetic safety line set zone line being a maximum value; the maximum value position column II is a position column with the projection within preset zones on the two sides of the maximum value position column I being a maximum value. By locking possible zones of a magnetic safety line, the system reduces calculated amount, reduces calculation difficulty and the error rate, solves the problem of the interference of splicing traces, and improves the accuracy.

The embodiment of the invention discloses an automatic network topology discovery method. The method comprises: devices in a network are grouped based on a switching domain, wherein the devices in a same switching domain belong to the same group; a hierarchical relation between all devices in each group is determined based on the groups; and on the basis of the hierarchical relation, a connection relationship between a port of a lower-level device and a port and an upper-level device is calculated and a network topology is established and updated based on the connection relationship. In addition, the embodiment of the invention also discloses an automatic network topology discovery apparatus.

The invention provides a method and a device for fast obtaining a course angle of a trackless vehicle in an underground roadway environment. The method comprises the steps that in a self coordinate system of the trackless vehicle, a plurality of corresponding distances between a coordinate origin and a plurality of points on the peripheral roadway wall in different angle directions are measured; the course angle of the trackless vehicle is calculated according to measurement results. The device comprises a distance detecting module and a course angle calculating module, wherein the distance detecting module is used for measuring the corresponding distances between the coordinate origin and the points on the peripheral roadway wall in different angle directions in the self coordinate system of the trackless vehicle, and the course angle calculating module is used for calculating the course angle of the trackless vehicle according to the measurement results. The method and the device can realize the self obtaining of the course angle of the trackless vehicle in the underground roadway environment, and are particularly applicable to unmanned autonomous running vehicles.

The invention discloses an underwater moving target detection device and method based on active polarization imaging, and belongs to the technical field of underwater moving target detection. According to the invention, the polarized light source is used for active illumination, the focal plane polarization camera is used for obtaining image polarization information, and underwater moving target detection is carried out. The method combines a Gaussian mixture background model algorithm with a pyramid optical flow method by using target polarization characteristics, overcomes the light influence and shadow mutation of an underwater imaging environment, eliminates the influence of shadow on target extraction, and can accurately detect the attitude, the movement direction and the movement track of the moving target. The calculation range of the optical flow method is reduced, and the target detection speed is improved.

The invention discloses a turnout steel rail variable cross-section-based wheel rail geometric contact point calculation method, which comprises the following steps of constructing a wheel coordinatesystem, making a plane x' = 0 in the wheel coordinate system to cut a wheel, and taking an intersection line of the plane x' = 0 and the periphery of the wheel as a main contour line; equally dividingthe wheel in the normal angle range of the two sides of the main contour line into n parts in the normal direction, and obtaining n sub-contour lines; cutting the steel rail through a plane x' = 0 toobtain a profile data set of a stock rail side and a turnout zone; and according to the main contour line of the wheel, the sub-contour line of the wheel and the contour data set of the stock rail side and the turnout zone, calculating the wheel rail geometric contact point to obtain a wheel rail geometric contact point calculation result. According to the method, the variable cross-section characteristic of the turnout zone steel rail is considered, large errors generated by turnout zone wheel rail geometric contact point calculation in the prior art are avoided, and the calculation result is accurate.

The present invention provides a method for estimating the number of superpoint connections based on GPU network access under a sliding window. The data structures are all stored in the video memory of the GPU, and the estimation of the number of superpoint connections is performed in parallel by the GPU, including the following steps: initializing the sliding estimation on the GPU Filter matrix and sliding candidate superpoint list, scan all messages in a time slice, estimate the number of superpoint connections at the end of the time slice, and slide the window. The invention can detect superpoints and estimate the connection number of the device under the sliding window, and the estimated value is more accurate, not affected by the window start time, and superpoints crossing the time boundary will not be missed; it has lower time delay than the existing method , can estimate the number of super-point connections in time; can process high-speed network traffic in parallel in real time, reduces the calculation range, and improves the overall speed of the algorithm. The sliding estimator matrix can be updated by multiple IP address pairs at the same time without generating errors .

The invention provides a multimedia information recommendation method and device and a storage medium, and the method comprises the steps: determining sample multimedia information which comprises the sample multimedia information and user feedback information corresponding to the sample multimedia information; clustering the sample multimedia information to generate N clusters; storing the sample multimedia information meeting a preset condition in the N clusters into a target matrix; determining a neighbor multimedia information set according to a similarity value between the sample multimedia information in the target matrix and the target multimedia information; the neighbor multimedia information set comprises at least one piece of sample multimedia information of which the similarity value is greater than a first preset value; determining the matching degree between the neighbor multimedia information in the neighbor multimedia information set and the target multimedia information; determining the first preset number of neighbor multimedia information corresponding to the maximum matching degree as recommended multimedia information; the recommended multimedia information is pushed to a target user, and the target user is a user who pays attention to the target multimedia information.

The invention belongs to the technical field of content similarity sorting of mass data, relates to content similarity sorting based on text classification probability vectors and text semantic vectors in mass data, in particular to a content similarity sorting algorithm based on text classification probability vectors and text semantic vectors in mass data. The algorithm can be divided into two layers: the first layer is a similarity calculation recall layer based on a text classification probability vector, and the layer calculates a local optimal similar text pool through the classificationprobability vector, so the layer is called as a text similarity recall layer; and the second layer performs content similarity calculation based on the local optimal similar text pool of the first layer, so the second layer is called as a text semantic similarity sorting layer. According to the method, a text similarity coarse-grained sorting layer is embedded on the basis of traditional text similarity calculation, so the optimal similar text pool is screened, and the calculation range of the text similarity is reduced through the algorithm. And when facing massive data, the method not onlysaves a large number of computing resources, but also reduces the computing time consumption.

The invention discloses an equal-sheet banknote distribution method for a plurality of banknote boxes for storing banknotes with different denominations in a withdrawal device. The method comprises the following steps: acquiring information of each banknote box for distributing banknotes and storing banknotes with different denominations respectively, and recording the information; then, according to the number of banknotes in each banknote box, sorting the banknote boxes from less to more; calculating an initial optimal banknote distribution sheet number for realizing an equal-sheet effect after the banknotes are discharged from each banknote box storing the banknotes with different denominations; sorting the cash boxes from large to small according to denominations and from small to large according to the same denominations; calculating the maximum lower limit value and the minimum upper limit value of each cash box for cash distribution of the current cash distribution amount; and finally, iteratively solving the banknote distribution number of banknotes of each denomination. According to the method, the calculation range of the exhaustion method can be greatly reduced, the number of iterations is reduced, the solving speed of the algorithm is improved, and the calculated solution variance is small. The method is suitable for cross-platform systems such as WINDOWS, LINUX and the like, is also suitable for systems such as PK, credential and the like, and is suitable for large-scale popularization.

The invention relates to the mining engineering, and particularly provides an underground trackless vehicle positioning method and system. The underground trackless vehicle positioning method comprises the steps that multiple label positions of which the intervals do not exceed the preset value along the running direction of a trackless vehicle under a mine are selected; an information label with a corresponding code is arranged on the roadway wall at each label position, and the corresponding relation is stored in a label information base; the trackless vehicle in the running process obtains the position information of the information label from the label information base according to the corresponding code by scanning the information label, and obtains the position information of the trackless vehicle relative to the information label at the moment through measuring the interval between the information label and the trackless vehicle; the position where the trackless vehicle is located at the moment is calculated by combining the position information of the information label and the position information of the trackless vehicle relative to the information label at the moment. By applying the underground trackless vehicle positioning method and system, the global coordinates of the trackless vehicle running under the mine can be easily obtained.

The invention provides a method and a device for fast obtaining a course angle of a trackless vehicle in an underground roadway environment. The method comprises the steps that in a self coordinate system of the trackless vehicle, a plurality of corresponding distances between a coordinate origin and a plurality of points on the peripheral roadway wall in different angle directions are measured; the course angle of the trackless vehicle is calculated according to measurement results. The device comprises a distance detecting module and a course angle calculating module, wherein the distance detecting module is used for measuring the corresponding distances between the coordinate origin and the points on the peripheral roadway wall in different angle directions in the self coordinate system of the trackless vehicle, and the course angle calculating module is used for calculating the course angle of the trackless vehicle according to the measurement results. The method and the device can realize the self obtaining of the course angle of the trackless vehicle in the underground roadway environment, and are particularly applicable to unmanned autonomous running vehicles.

The invention discloses a real-time large-scale image synthesis algorithm of a microscopic system, which relates to the technical field of microscopic optical imaging and image processing. The algorithm adopts a multi-threaded parallel image acquisition method, which is convenient and fast; Extracting features from the region of interest ROI avoids the calculation of feature points that do not contribute to the composite image, greatly reduces the calculation time and steps, avoids the influence of wrong feature points, and improves the accuracy and efficiency of the extracted feature points ;Under the GPU acceleration environment, the matching pair sets of each image are calculated in parallel according to the method from dense to sparse, which greatly reduces the calculation of the sparse area of ​​​​feature points, and gives priority to the matching of dense feature sets, which improves the processing efficiency and Accuracy; the optimal matching pair is screened based on the length and slope of the reference line, which simplifies the calculation process; the algorithm is easy to realize real-time large-scale image synthesis of the microscopic system.

The invention discloses an equal-space banknote distribution method for multiple banknote boxes for storing banknotes with different denominations in a withdrawal device. The equal-space banknote distribution method comprises the following steps of: acquiring information of each banknote box for distributing banknotes and storing banknotes with different denominations respectively, and recording the information; then, according to the number of banknotes in each banknote box, sorting the banknote boxes according to a sequence from more to less; calculating an initial optimal number of reserved banknotes for realizing an equal-space effect after the banknotes are discharged from each banknote box for storing the banknotes with different denominations; sorting the banknote boxes according to denominations from large to small, and sorting them according to the number of banknotes from small to large when the denominations are the same; calculating the maximum lower limit value and the minimum upper limit value of each banknote box for banknote distribution of a current banknote distribution amount; and finally, iteratively solving the banknote distribution number of banknotes of each denomination. According to the equal-space banknote distribution method, the calculation range of an exhaustion method can be greatly reduced, the number of iterations is reduced, the solving speed of the algorithm is increased, the calculated solution variance is small, and the method is suitable for cross-platform systems such as WINDOWS and LINUX, is further suitable for systems such as PK and creative systems and is suitable for large-scale popularization.

The invention discloses a real-time large-scale image synthesis algorithm of a microscopic system, and relates to the technical field of microscopic optical imaging and image processing, the algorithm adopts a multi-thread parallel image acquisition mode, and is convenient and rapid; according to the method, features are only extracted from a region of interest (ROI) by using a multi-thread technology, so that calculation of feature points which do not contribute to a synthesized image is avoided, the calculation time and the calculation steps are greatly reduced, the influence of wrong feature points is avoided, and the accuracy and the efficiency of extracting the feature points are improved; in a GPU acceleration environment, a matching pair set of each image is solved in parallel according to a dense-to-sparse method, so that calculation of sparse regions of feature points is greatly reduced, matching of dense feature sets is preferentially ensured, and processing efficiency and precision are improved; the optimal matching pair is screened based on the length and slope of the reference line, and the calculation process is simplified; the algorithm is easy to realize real-time large-scale image synthesis of the microscopic system.

The invention provides a road point cloud error scene analysis method based on a space plane offset analysis model. The method is characterized by comprising the following steps: S1, collecting point cloud data of a field road, and selecting local plane features on the road as a reference plane; S2, randomly performing control point coordinate acquisition on the selected reference plane according to the reference plane selected in the S1; S3, performing discreteness analysis on the control point data, and salient outliers in the control point data are eliminated by using a proposed salient outlier rapid screening algorithm; S4, constructing a point cloud local calculation space based on a discreteness analysis result in the S3; S5, acquiring a reference plane equation based on least square space plane fitting; S6, calculating the distance distribution condition from the laser foot points of the road point cloud to the reference plane; and S7, error scene analysis based on the space plane offset model. According to the method, qualitative analysis and systematic induction can be carried out on the road point cloud data deviation condition, and the mastering degree of the local data deviation condition is improved.

The present invention relates to a monocular vision and motor imagery-based air vehicle three-dimensional space target searching system. The system comprises: a monocular vision navigation sub-system:used for obtaining real-time foresight environment barrier data of an air vehicle, estimating a feasible flight direction, and transmitting the data to a main flight control panel of the air vehiclethrough a COM interface; and a decision sub-system: 15 leading left hand, right hand, both feet and tongue motor imagery is collected according to analysis, and a PC end ground control system convertsa final decision into a control instruction through WIFI and transmits the final decision to the air vehicle. The system is based on monocular vision and motor imagery, a brain machine interface forcollecting 15 leading left hand, right hand, both feet and tongue motor imagery and monocular vision-based navigation are combined to implement multi-rotor air vehicle indoor three-dimensional space target searching, and the system has advantages of being high in intelligence degree, low in calculation overhead, easy for operation, stable in control and capable of enabling a user to simultaneouslyimplement more control tasks and the like.

The embodiment of the invention discloses an automatic network topology discovery method. The method comprises: devices in a network are grouped based on a switching domain, wherein the devices in a same switching domain belong to the same group; a hierarchical relation between all devices in each group is determined based on the groups; and on the basis of the hierarchical relation, a connection relationship between a port of a lower-level device and a port and an upper-level device is calculated and a network topology is established and updated based on the connection relationship. In addition, the embodiment of the invention also discloses an automatic network topology discovery apparatus.