114results about How to "Increase time complexity" patented technology

The invention discloses a video shot detection method based on histogram improvement and a clustering algorithm, and relates to image processing techniques. According to the method, the histogram improvement and the clustering algorithm are adopted to compute an intersection of histograms of two adjacent frames of images, and whether a shot change happens is judged according to histogram similarity; if the shot change happens, secondary detection on a shot boundary of the intersection of the histograms of the two adjacent frames of images is carried out by utilizing interframe gray scale / color difference values, pixel difference value computation on each block is carried out respectively by adopting non-uniform block weighting processing, pixel difference values and preset block frame differential threshold values are compared to obtain identification variables, the identification variable of each block is weighted and summarized, and the weighing and summarizing values and the preset block differential threshold values are compared to carry out shot detection. The video shot detection method improves shot detection accuracy and solves the problems of shot false detection, frame number discontinuity, and the like.

The invention belongs to the technical field of path or flight path planning of robots as well as low-altitude flight aircrafts, specifically relates to a path planning method of a passable area divided at unequal distance, and is used for solving the problem that existing planning algorithm has large time complexity in time and space complexity. The path planning method comprises the following steps of: calculating convex extreme points of each barrier curve; dividing the passable area by using each convex extreme point as a horizontal line; abstracting each small area obtained by dividing into a peak of a graph; forming an undirected graph by all peaks; finding out a peak serial number corresponding to the small area at which a starting point and a final point are located; finding out all paths for the undirected graph by breadth-first or depth-first scanning; finding out an actual to-be-travelled path of a moving object according to the situation on an actual map. The path planning method disclosed by the invention has the beneficial effect of overcoming the problems of algorithms of A* and the like on memory space and operation time, and overcoming a convergence problem of an ant colony algorithm at the same time. Besides, time complexity and space complexity are improved greatly in comparison with other algorithms.

The invention discloses a device and a method for estimating the health status and the state of charge of a battery pack on line. A monitoring unit measures the running status of the battery pack in real time; a storage unit stores data measured by the monitoring unit; an iterative computation unit computes an iteration parameter of a battery model; a voltage prediction unit calculates and outputs a voltage predicted value; an error computation unit computes an error between the voltage predicted value and a measured value; a parameter prediction unit calculates equivalent internal resistance and equivalent capacitance of the battery pack; an average temperature computation unit computes an environmental temperature average value within one hour; a temperature correction unit converts the parameter value of the battery pack into a corrected value at the normal temperature of 25DEG C; a health status estimation unit predicts the health status of batteries; a decoupling unit decouples the heath status and the state of charge of the batteries; a stage of charge estimation unit predicts the state of charge of the batteries. The device can estimate the health status and the state of charge of the batteries under real-time working condition.

The invention relates to the field of agricultural outlier detection, in particular to an outlier detection method based on agricultural big data. The method comprises the following steps: a data collection step of collecting agricultural production data, agricultural soil data and agrometeorological resource data, and integrating the data into a training data set; The step of constructing iTree tree is to select m sample points from the training dataset and continuously randomly select splitting attributes and splitting points until the termination condition is reached; the step of constructing iTree tree is to select m sample points from the training dataset. Constructing an isolated forest algorithm model, initializing the number t of iTree trees in the isolated forest and the set m ofsubsamples taken when constructing the iTree trees, entering the step of constructing the iTree trees in a loop, and constructing mutually independent iTree trees, wherein the set of all iTree trees constitutes the isolated forest algorithm model; An outlier judging step of calculating an outlier score s (x), and judging whether the test data x is an outlier by the outlier score s (x). The invention applies the isolated forest algorithm model to the outlier detection of the agricultural big data, and can effectively improve the detection effect of the outlier of the agricultural big data.

The invention discloses a method for calculating the author influence based on a citation propagation network. The method comprises the following steps of step 1, establishing a citation relationship network and a colleague relationship network: establishing an author citation relationship network and the colleague relationship network based on an essay public information database; step 2, establishing the citation propagation network, wherein the citation propagation network is an integration network of the author citation relationship network and the colleague relationship network; step 3, calculating the author influence based on the citation propagation network, wherein the calculation formula is as shown in the figure. The method for calculating the author influence based on the citation propagation network can obtain more reasonable author influence indexes, and is easy to be implemented and simple and convenient for calculation.

The invention discloses a starting control system and control method of an automobile engine. The system comprises a micro computer module, an engine starting module, an engine closing module, an engine state detecting module, an engine temperature detecting module, a starting circuit fault detecting module, a control system detecting module, a control system maintaining module, an alarm module and a display module; the micro computer module controls the engine starting module and the engine closing module, controls the engine state detecting module to detect the state of the engine, controls the starting circuit fault detecting module and the engine temperature detecting module, controls the control system detecting module and the control system maintaining module and controls the alarm module and the display module. Whether an engine circuit fails or not can be effectively detected, an alarm can be given out in time, and meanwhile, various detected data are displayed.

The invention provides a deep reinforcement learning interactive recommendation system and method based on knowledge enhancement, and relates to the technical field of recommendation. The system comprises a data acquisition and cleaning module, an environment simulator construction module, a knowledge graph construction module, a graph convolution module, a user state representation module, a strategy network module and a value network module. According to the method, rich semantic information in a knowledge graph is combined, a graph convolutional network structure is utilized, embedded representation of adjacent entities is propagated recursively along high-order connectivity, a graph attention network thought is adopted, item representation is enhanced by utilizing the rich semantic information in the knowledge graph, and meanwhile, a user-item bipartite graph is fused, so that the method is more efficient and efficient. The potential relationship is fully mined from collective user behaviors, so that the dynamic preference of the user is accurately captured, and the optimal recommendation strategy is autonomously learned by using deep reinforcement learning, so that the recommendation accuracy is improved.

The invention provides an identification method of a woven fabric tissue chart, wherein the identification method relates to image analysis. The identification method comprises the steps of a first step, identifying the type of the woven fabric tissue structure based on a yarn boundary characteristic, namely performing woven fabric image pre-treatment, performing inclination correction and dividing for obtaining each tissue point image, determining the latitude and longitude attributes of the tissue point image by means of the yarn boundary characteristic of the tissue point image, obtaining the number of circulating yarns of the woven fabric and classifying the tissue structure; and a second step, performing woven fabric tissue chart identification based on an improved Gabor characteristic and the tissue structure type, namely extracting the characteristic of a non-twill tissue point image based on the improved Gabor kernel transformation, performing PCA dimension reduction, classifying the non-twill tissue point image by means of a support vector machine, performing classified correction on a woven fabric tissue chart matrix, and outputting a correct woven fabric tissue chart. The identification method of the woven fabric tissue chart can realize a high-robustness identification effect on woven fabric three-elementary tissues which are obtained through weaving yarns with different dimensions and colors and the woven fabric tissue chart with a simple changed tissue structure.

The invention discloses a polarization SAR terrain classification method based on deep learning and distance metric learning. The polarization SAR terrain classification method comprises the realization steps that (1) images are inputted; (2) filtering is performed; (3) features are extracted; (4) training samples and test samples are selected; (5) a stacked sparse auto-encoder is trained so that the deep features of a training sample set and the deep features of a test sample set are obtained; (6) a distance metric learning classifier is trained so that a classification result is obtained; (7) the classification result is colored; and (8) the colored classification result graph is outputted. The images are classified by using the polarization SAR terrain classification method based on deep learning and distance metric learning so that feature extraction is relatively comprehensive and reasonable, the classification result is more consistent with real terrains, time complexity is reduced and classification precision is enhanced.

A secret key isolation signcryption method includes the steps that a credible assisting device is arranged between a computer of a message sender and a computer of a message receiver, the assisting device assists the message sender and the message receiver in generating initial private keys and initial public keys, the private keys are updated at the start point of each period, when the message sender sends messages, signcryption is conducted on the messages through the private keys of the message sender and the public keys of the message receiver, unsigncryption is conducted on the messages through the private keys of the message receiver and the public keys of the message sender after the message receiver receives ciphertexts, and the messages are recovered. According to the method, digital signature and public key encryption are combined into one step, a secret key isolation mechanism is introduced to a signcryption system, so that the calculated amount and communication cost can be saved, and the requirements for confidentiality, non-repudiation, forward security and public verifiability are met.

The invention relates to a cancer recurrence prediction system based on a multi-dimensional Gaussian distribution Bayesian classification, which comprises a preprocessing module, a training module anda Bayesian classifier; the pre-processing module performs data cleaning on the training set and generates a class vector data set; the training module first calculates the first probability of two class attributes, and then divides the data attribute into a class data attribute set which is closely related to the class attribute and a class II data attribute set which is sparse with the class attribute association degree by using the pearson correlation coefficient, two types of data attribute sets are respectively used for calculating a corresponding probability by using a multi-dimensionalGaussian distribution and a one-dimensional Gaussian distribution; the Bayesian classifier combines both the probability of the two and the first probability of the class together as the probability of the data belonging to each class, and the classification test result of the cancer is judged accordingly. The cancer recurrence prediction system based on multi-dimensional Gaussian distribution Bayesian classification improves the predictive accuracy of the recurrence of the cancer.

The invention relates to a network flow measurement method and system based on an approximate zero error probability measurement data structure Sketch. The method comprises the steps that a CM-Sketchdata structure used for storing small flow information and a data structure used for storing labels are set; when the table entry is updated in the network flow measurement process, the CM-Sketch is firstly updated, and then the data structure of the storage label is updated based on the observation of the CM-Sketch. When information of a new stream is inserted, label information and a timestamp of the new stream are stored in a designed data structure used for storing labels; when the flow information needs to be restored, the equation set can be solved according to the label stored in the data structure, and the accurate flow information is obtained. According to the method, on the premise that the algorithm time complexity is not obviously improved, the defect that Elastic Sketch is insufficient in small flow information query accuracy is overcome, and instantaneous query can be supported.

The invention discloses an XML (Extensible Markup Language) document structure based on an extended adjacent matrix and a semantic similarity calculation method, belonging to the technical field of data excavation. The method concretely comprises the following steps of: encoding an XML document tree; as for two encoded documents, generating a schema document node list and a data source document node list firstly and then generating a schema extended adjacent matrix and a data source extended adjacent matrix (P1, P2); and calculating the similarity of XML documents through cos (P1, P2). In the method, different contributions of nodes with different levels to the documents are fully considered, and the highest time complexity of the method is O (n2) under the condition that the amount of the XML document node is n and is prior to that of an edit distance algorithm.

The invention relates to a scheduling method of hardware accelerators in a C-RAN. The method comprises the following steps of 1) establishing a task queue work_queue in a virtual machine controller Dom0, and simultaneously, aiming at each hardware accelerator, establishing a task queue write_queue and a task queue read_queue; 2) after the virtual machine controller Dom0 receives a request of a virtual machine DomU, adding the request in the task queue work_queue according to a first-come-first-served policy; 3) successively taking out tasks in the task queue work_queue, selecting the hardware accelerator with a current minimum load, and putting the task which is taken out into the corresponding task queue to complete a scheduling design of the plurality of hardware accelerators so that a virtual machine shares the accelerators. A signal processing speed is increased, hardware accelerator loads are balanced and a system throughput is increased.

The invention discloses an IRA (Irregular Repeat-Accumulate) codes open set blind recognition method, which realizes the recognition of the code length, the codeword starting point and the code rate, the sparse reconstruction of a check matrix and restoration of an interleaving permutation relation in the absence of any prior knowledge, and aims to solve the problems that the prior art cannot accurately recognize the IRA codes with arbitrary code length and bit rate under a bit error condition, and the recognition speed is slow. The method comprises the following steps: firstly, constructing an analysis matrix according to the estimated code length and the codeword starting point, and solving a dual vector by using a Gauss column elimination method and a statistical decision criterion and other methods to realize the recognition of the code length; secondly, using the dual vector to eliminate the bit errors, iterating the above operations and obtaining multiple groups of dual vectors and further analyzing the codeword starting point; thirdly, realizing the sparse reconstruction of the check matrix of the IRA codes by an obtained effective check vector; and finally, according to the coding structure features of the IRA codes, analyzing the interleaving permutation relation, and finishing the overall recognition of the IRA codes. Through theoretical analysis and simulation verification, the IRA codes open set blind recognition method provided by the invention has lower computation complexity and excellent fault-tolerance performance.

There is provided a method and an apparatus for updating a background model used for background subtraction of an image. The method comprises: receiving an image (220), and classifying a region (226) in the image as foreground by performing background subtraction using a background model (240). The background model comprises a collection of background samples (248, 248b, 248c, 248d) for each pixel (228) in the image. The collection of background samples is arranged in a list of background images (242a, 242b, 242c, 242d). The method further comprises: replacing image contents in the region (226) of the image which is classified as foreground by image contents of a corresponding region (246) in a background image in the list, and adding the image to the list of background images by rearranging a collection of pointers (244a, 244b, 244c, 244d) which each points to one of the background images in the list, such that one of the pointers instead points to the image.

The invention discloses a violent sorting behavior recognition method and device. The method comprises the steps of obtaining a to-be-predicted video; graying the video to be predicted; and inputtingthe grayed to-be-predicted video into a predetermined three-dimensional convolutional neural network model to perform violent sorting behavior identification, wherein the predetermined three-dimensional convolutional neural network model has a behavior feature extraction capability. According to the method, a three-dimensional convolutional neural network model is constructed by utilizing a modelwith behavior feature extraction capability, and learning and violent sorting behavior recognition are carried out by utilizing a grayscale video. Therefore, the violent sorting behavior recognition period is shortened, and the judgment accuracy is improved.

The invention provides an attribute weighting method based on information gain ratios. Firstly, the information gain ratio of each attribute is calculated and then by utilizing the information gain ratios, the weight of each attribute is respectively calculated. The invention also provides a polynomial naive Bayesian text classification method relying on the attribute weighting method based on information gain ratios, a complementary set naive Bayesian text classification method relying on the attribute weighting method based on information gain ratios and a polynomial and complementary set integrated naive Bayesian text classification method relying on the attribute weighting method based on information gain ratios. The invention improves the classification precision of the original naive Bayesian text classifier, and also keeps the simplicity and time complexity of the original naive Bayesian algorithm.

The present invention discloses an anti-malicious attack proxy secret key mixing-based electronic voting system and method. The system comprises a proxy voting program module and a voting information verification module. The proxy voting program module comprises a pre-processed parameter generation unit, a signature private key encryption unit, a mixing program unit, and a voting information output unit. The information verification module comprises a voting result receiving unit, a decoding unit, and a voting information verification unit. The system has the characteristics of low complexity, high security, and simplicity in verification. The system and method are applicable to an electronic voting server that is likely to be attacked maliciously. Even if the voting server suffers malicious attack, whether invalid votes exist can be found (each voter signs on his own vote, and a notary can find out ballot rigging), and a hacker cannot acquire signature private key information of the voter from the attacked electronic voting server.

The invention discloses a functional block intelligent wiring method applied to the modeling of a complex industrial control system, which relates to intelligent wiring technology. In the invention, according to the characteristics of complex control relations, a plurality of functional blocks and a plurality of control loops, and the modeling requirements of high efficiency and high flexibility in the modeling process of the industrial control system, the intelligent wiring method is designed. In the method, through judgment on the number of break points of connecting lines, the trend selection of the connection lines, connection index determination, the resolution of coordinates of the break points of the connection lines and a transfer process of a state space in a wiring territory area, and by utilizing mapping and inverse mapping methods, the problems of the complex structure, much equipment, complex configuration relations and disordered configuration interface of an industrial control network are solved, the problem on interconnection among the different functional blocks is solved, and flexibility and applicability are relatively higher. The method has the advantages of high wiring speed, high connecting line quality, specifically remarkable superiority, certain promotion value, and high time complexity and space complexity when the functional blocks are in a certain scale.

Aiming at the problem of camera movement when moving object detection in natural environment, the invention provides a moving object detection scheme adapted to camera movement. Firstly, the background model is initialized by a Gaussian mixture model. Then the pyramid optical flow method is used to detect the characteristic points of the moving optical flow of two adjacent frames, The RANSAC algorithm is used to filter out the correct matching point pairs, and the offset of the matching point pairs is used to estimate the camera offset, then the image matching homography matrix is obtained, and the background model matched with the current frame is restored. The model is used as the final background model for moving object detection in the current frame. In the process of background modeling, a Gaussian mixture background model based on the concepts of spatial element and temporal element is proposed. The spatial element improves the detection accuracy, and the temporal element improves the adaptability of background model.

The invention discloses a fast optimization algorithm of D2D power distribution in case of single-channel cellular users. A communication rate optimization problem of D2D users based on a cellular system is a complex nonconvex nonlinear optimization problem. The cellular users that the fast optimization algorithm aims at utilize a single channel, while the D2D users utilize multiple channels; the algorithm can be used for quickly optimizing and solving the transmitting power pi of the cellular users and the transmitting power qi of the D2D users in a plurality of cases such as cellular users, power limitation and independent power limitation, and D2D users, power limitation and independent power limitation, and the like. The transmitting power obtained by use of the optimization algorithm is capable of ensuring the communication rate requirements of the cellular users and maximizes the sum of the communication rate of all the D2D users on all the cellular frequency bands. The algorithm has the advantages of high convergence rate, small calculation amount, easy implementation, high result accuracy and the like.

The invention is a packet exchanging method and the device in omputer network. It consists of three parts of weight development sequence scanner, weight matrix supervisor and queue supervisor. The output value of the weight development sequence scanner is used to determine a defined row of weight matrix in weight matrix supervisor by the queue supervisor and to scan the row from top to bottom. When nonzero term of weight matrix in the weight matrix supervisor, the queue supervisor proceeds services for relative flow by utilizing contained packet selector. After the row in weight matrix being scanned, the quece supervisor fetches next value from weight development sequence in the weight development sequence scanner, then proceeds scanning for relative row in weight matrix supervisor, and so on, it will proceeds scanning from the top of sequence when the bottom of sequence of scanned by weight development sequence scanner.

The invention aims to disclose a method for quickly searching a mass digital certificate backlist, which is based on a search method of preprocessing, multilevel classification and a dichotomy, wherein the complexity of the backlist research time is improved from O(N) to O(LogN), so the research performance is greatly improved, in the practical test, the method is used to research backlist items which have a scope of 1 million, the average time is controlled within about 1microsecond, so the purpose of the invention is realized.

The invention discloses a service cluster constructing method based on a semantic Web. The service cluster constructing method comprises the following steps of S101 constructing a service space based on semantics, S102 building a mapping relation of a Web service and the service space, S103 constructing service clusters and a dynamic library of the service clusters and S104 describing the organization structures of the service clusters by a logic Petri net. According to the service clusters generated by the service cluster constructing method, only n concepts need to be quantified, the number of inquiring a body tree is n, and the time complexity of generating of the service clusters is greatly improved. According to the service cluster constructing method, the type of service ports and service quality are clarified through semantic concept blending, and the construction of the service clusters is reasonable. The service clusters are generated in the service space, the service clusters orienting the requirement of a user are matched, calculated and converted into coordinate inquiring, the more of the constructing basis of the service clusters is, the higher of the constructing precision of the service clusters is, the time complexity of the system for matching the service clusters according to the user need is small, and the limitation of a common service clustering method is solved.

The invention discloses a template-oriented Word2vec-based log anomaly detection method and device, and the method comprises the following steps: carrying out the preprocessing of an original log, obtaining a log template, and carrying out the segmentation of the log template, so as to obtain a log sequence; solving a feature vector of the log template based on Word2vec, wherein the ID serial number of the log template is used as the input of the Word2vec; solving a feature vector of the log sequence according to the feature vector of the log template; and performing machine learning on the feature vector of the log sequence to obtain an anomaly detection model, and performing detection according to the anomaly detection model. Starting from a Word2vec processing object as a template, thescale of training data can be reduced. Moreover, the original log is preprocessed, and the time consumed by log anomaly detection is reduced through preprocessing so as to avoid affecting the final anomaly detection result.

The invention discloses a subway tunnel disease detection method based on deep learning, and mainly solves the problems that the subway tunnel detection amount is large, and the existing detection algorithm is low in efficiency and large in missing detection amount. The method specifically comprises the following steps: making a to-be-detected subway tunnel image database, including a subway tunnel normal region picture database and a subway tunnel disease region database; based on an improved Faster R-CNN target detection algorithm, carrying out target positioning and accurate identification and classification on the subway tunnel diseases; detecting a disease state in the to-be-detected area by adopting a deep learning-based method; and finally, inputting a to-be-tested image into the detection model to obtain a tunnel disease detection result. The solution thought of the traditional algorithm is image-preprocessing-artificial feature extraction-classification. According to the method, the problems existing in current subway tunnel disease detection are solved by utilizing the thought of deep learning, namely image-feature extraction network-classification and regression.

The invention discloses an expression recognition method based on reverse synergetic salient region features. The whole recognition process is mainly divided into five parts, namely preprocessing, detection of an expression salient region, feature extraction on the salient region, weight assignment and recognition and classification. The method comprises the specific steps that (1) a face region is divided in a training test sample; (2) a reverse synergetic salient detection algorithm is utilized to extract the expression salient region from the divided part; (3) an LBP operator and an HOG operator are utilized to perform feature extraction on the salient region; (4) a support vector machine is utilized to perform preliminary classification on all salient local features, and weight assignment is performed; and (5) a multi-classification decision-making mechanism is used for recognition and classification. The method is combined with the relevancy between expressions, the relevancy is utilized to extract a local region containing rich expression information, and therefore the calculated amount is greatly reduced; and meanwhile, the multi-classification decision-making mechanism is used for classification, so that the recognition rate is increased accordingly.

The invention provides a D2D power distribution fast optimization algorithm in presence of multichannel cellular subscribers. The communication speed optimization problem of cellular network based D2D subscribers is a complex non-convex and non-linearity optimization problem. Aiming at the multichannel cellular subscribers and multichannel D2D subscribers, the transmission power pi of the cellular subscribers and the transmission power qi of the D2D subscribers can be optimized and solved rapidly under the conditions of independent power limitation of the cellular subscribers and independent power limitation of the D2D subscribers by applying the algorithm. The transmission power obtained by adoption of the optimization algorithm can ensure requirements of the cellular subscribers on the communication speeds on all the cellular frequency bands and can maximize the sum of the communication speeds of all the D2D subscribers on all the cellular frequency bands. The algorithm has the advantages of a fast rate of convergence, low calculated amount, easy implementation, high result precision and the like.