65results about How to "Reduce time and space complexity" patented technology

The invention provides an electronic image stabilization method based on characteristic coupling. The method comprises the following steps: (1) motion estimation: estimating motion parameters of a present frame and a previous frame to obtain a global motion vector and a local motion vector; (2) motion filtering: employing a Kalman filter to carry out filtering on the global motion vector, and obtaining a willful scan vector, wherein difference between the global motion vector and the willful scan vector is a vibration parameter needs to compensate; (3) motion compensation: utilizing a compensation parameter obtained by filtering carrying out frame by frame compensation; (4) video restoration: by utilizing space-time redundant information of video, carrying out restoration on the local motion vector with a motion restoration method and guiding pixel point filling. In order to solve motion carrier vibration influence on video shooting effect, the invention provides the electronic image stabilization method based on characteristic coupling. The method has the advantages of high accuracy, good anti-noise performance, strong tracking performance, good boundary consistency and the like.

The invention discloses a network hot event detection method based on text classification and clustering analysis. The method solves the problem that the efficiency and accuracy rate of the existing network hot event detection method based on clustering analysis need to be improved. The method comprises the steps that feature words are respectively selected for various classes of files through feature extraction and feature selection by utilizing a training corpus; each training text and test text are represented as vectors in all of the feature spaces by utilizing a vector space model method, and the weight of each dimension of the vectors is determined by utilizing a TF-IDF (term frequency-inverse document frequency) method, and then each test text is classified; the classified test texts in different classes are respectively subjected to clustering analysis, so the hot cluster of each class is obtained, the feature word representing the hot event is obtained through further analysis, and then the word property and other aspects of each feature word are analyzed; the description of each hot event is generated by utilizing relevant language knowledge and necessary linguistic organization. With the network hot event detection method based on text classification and clustering analysis, the detection efficiency and accuracy rate of hot events can be effectively improved.

The present invention provides a map pathfinding method, dividing plane map into one or more identical area, generating waypoint on area border; when finding path, processing path find based on the area to acquire area path; then generating final path based on the triangular facet Path finding according to the acquired area path. The invention also provides a mapping road finding system, adopting method and system of the invention not only saves space occupation, but also improves road finding efficiency, also suitable for process of all kinds of maps.

The invention discloses a fatigue driving state detecting system and method based on decision-making level data integration. The fatigue driving state detecting system firstly uses an acceleration sensor to acquire motion acceleration of a steering wheel, judges motion states of the steering wheel based on dynamic acceleration threshold values, and primarily judges fatigue driving states of a driver based on 4s immobility theory. Detecting error tolerance is improved by setting errors of the steering wheel turning from left to right. Pulse time domain values of the driver can be acquired via a pulse sensor, and physiological states of the driver can be detected based on dynamic pulse threshold values. Integrated detecting results can be obtained by integrating two kinds of detecting results in decision making level. By the fatigue driving state detecting method based on decision-making level data integration, a fatigue driving state detecting prototype system is constructed. Compared with existing methods, the fatigue driving state detecting method based on decision-making level data integration has certain advantages in detection accuracy and perfect performance in response time, time complexity, memory occupation and the like of algorithms.

The invention provides a sub-pixel level stereo matching method based on a scale space. The method comprises the following steps of: (1) establishing the scale space for a stereopair; (2) determining the size of a matching window for each point of a current level reference image in the scale space; (3) calculating a parallax searching range for each point of the level reference image according tothe parallax of the former level; (4) calculating an integer-grade parallax of the current level in the scale space; and (5) calculating a sub-pixel level parallax through a sub-pixel level matching method based on binary search. The method of the invention uses the stereopair generated under the condition of small height ratio to match; a dispersed scale space is used in the matching process to improve the matching efficiency; a bonding phenomenon generated in the matching process is reduced by using a mixed window selecting strategy; and the sub-pixel level parallax is obtained by the sub-pixel level matching method based on the binary search so as to ensure high sub-pixel level precision, and greatly improve matching efficiency of the algorithm.

The invention relates to a road matching method based on the fusion of a probe vehicle movement track and an electronic map. The method is characterized in that an urban road network is divided by adopting a mesh division technology, so that the road section search complexity can be greatly reduced; a map matching method fusing a three-section matching concept of reference historical data and prospective data and an improved weight model concept is adopted for overcoming the weaknesses by acquiring strong points, the matching degree calculated by adopting the weight model is adopted as the evidence, and the three-section matching concept is used for making the decision, so that the reliability of the decision evidence and the reasonability of a result decision can be improved, and the relatively high matching precision can be achieved by virtue of the minimum complexity.

The invention relates to a speaker confirmation method and a speaker confirmation device used in a short voice condition. The method comprises the steps of extracting a linear prediction cepstral coefficient, a Mel frequency cepstral coefficient and a Delta characteristic from an objective voice signal; combining the linear prediction cepstral coefficient, the Mel frequency cepstral coefficient and the Delta characteristic, thereby obtaining a plurality of effective characteristic vectors; reducing the dimension number of the effective characteristic vectors by means of a partial fuzzy PCA method; and modeling by means of the dimension-reduced effective characteristic vectors according to a Gaussian mixture model, and identifying the speaker of the objective voice signal. Compared with the prior art, the speaker confirmation method and the speaker confirmation device are advantageous in that a single characteristic is replaced by a characteristic combination, thereby improving number of dimensions of the effective characteristics and compensating defects of a characteristic sample, and furthermore the local fuzzy PCA is used for performing effective dimension reduction on the combined characteristic, thereby reducing time-space complexity of the system under a precondition of low effect on identification rate.

The invention relates to an application layer DDOS (distributed denial of service) attack and defense method, which is easy to realize, is low in time complexity and high in accuracy, is transparent to users, and has no influence on the user visit. The method provided by the invention is divided to two stages of a training stage and a work stage. In the training stage, real and legal visit flow is adopted as training data to generate a standard array used for real-time detection and protection of the work stage. According to the invention, application layer DDOS equipment needs to be arranged in front of an application server in series, so that the visit flow can be filtered through a defense device before entering the server.

The invention discloses a three-dimensional packing overall optimization method and system for putting multiple goods and materials into multi-specification packets and belongs to a method for intelligentized processing of goods and material packing. A packing overall optimization calculation module is started, the packing overall optimization calculation of one task is finished by combining a quantum genetic algorithm with a heuristic three-dimensional packing algorithm, and packing schemes are output. Reasonable schemes to be presented are stored in a database after judgment. Through organic combination of the heuristic three-dimensional packing algorithm and the quantum genetic algorithm, the overall packing optimization calculation can be performed for the multiple goods and materials which are to be placed into multi-specification packet containers. Compared with existing optimization calculation for processing a single packet container or one type of packet containers, the three-dimensional packing overall optimization method and system has the obvious overall advantages.

The invention discloses a method and a device for retrieving the similarity of pictures and belongs to the field of the graphic images. The method comprises the steps of obtaining the to-be-retrieved picture characteristics of to-be-retrieved pictures, hashing the to-be-retrieved picture characteristics by use of an LSH (Locality Sensitive Hashing) algorithm to generate hash values of the to-be-retrieved pictures, finding database picture hash values of which are matched and similar to the hash values of the to-be-retrieved pictures om a hash table corresponding to every hash values of the to-be-retrieved pictures, finding database pictures according to the database picture hash values, and according to Euclidean distances of the characteristics of the database pictures and the characteristics of the to-be-retrieved pictures, selecting a preset number of database pictures from the database pictures having relatively smaller Euclidean distances. The method and the device have the characteristics that the problem that a user cannot retrieve the most similar pictures from a plurality of results obtained by the LSH algorithm at the present is solved, and the advantages of LSH in reducing the time-space complexity in picture similarity retrieval and supporting high-dimensional data retrieval are better developed by virtue of the combination of the LSH algorithm and a linear retrieval algorithm.

The invention discloses an audio-video fingerprint generation method based on key frames. The audio-video fingerprint generation method comprises the following steps: firstly extracting a plurality of the key frames from an I-frame of video data based on a minimum vertex covering principle so as to reduce fingerprint expenditure; then by virtue of characteristics of DCT (discrete cosine transform) coefficients of the I-frame, selecting a direct current (DC) coefficient x<0> and N low-frequency alternating current (AC) coefficients x<1>, x<2>...and x<N> with minimum frequency of the key frames, generating a 0 / 1 sequence with N+1 digits by a quantization coding mode, converting timestamps of the key frames into a binary sequence, and placing the binary sequence in front of the 0 / 1 sequence with the N+1 digits so as to form a video frame fingerprint with M-1 digits for the key frames; and by virtue of characteristics of exclusive OR (XOR) operation, and uniting audio and video fingerprints into one, an audiovisual fingerprint is generated through the matching and synchronous authentication of the timestamps. The audio-video fingerprint generation method has the advantages of achieving the purpose of authenticating a plurality of media data at the same time through one fingerprint, greatly reducing the fingerprint expenditure and realizing certain error location, error detection and repair functions on problems such as incorrect playback and error playback of audio-video programs, insertion of illegal programs, program interpolation and the like.

The invention discloses a dynamic path generation method directed at indoor navigation. The dynamic path generation method can provide real-time dynamic navigation path for indoor navigation service. A door state trigger defined in the invention comprises a fingerprint sensor and a smog sensor; and a variety of dynamic semantic information capable of influencing indoor topological structure can be captured through the door state trigger, so dynamic updating of an indoor topological network is realized. A corridor dividing method provided by the invention divides a corridor into a group of spatial units, and virtual doors are additional arranged among the spatial units for connection of the spatial, which is favorable for providing more accurate indoor navigation path. According to a network extension planning method utilized in the invention, a search tree generated in previous iteration is updated through real-time positional information and the dynamic topological network during path calculation, and eventually, navigation path is directly obtained through the search tree; so recalculation during each path planning is avoided, and temporal and spatial complexity during dynamic path planning is reduced.

The invention discloses a rapid defogging method for a road image. The rapid defogging method for the road image includes: step 1, obtaining an atmosphere light value and a propagation diagram according to a saturation component of the initial fog weather road image; step 2, obtaining an enhancement region segmentation diagram according to a visibility expression, and restoring the initial fog weather road image; step 3, performing contrast stretching treatment on the restored image so as to confirm the road image after being defogged. The rapid defogging method for the road image is established based on road scene image characteristics, different from a unified enhancement mode for the whole image in most existing defogging methods due to the fact that the rapid defogging method for the road image uses weak enhancement to process an area of the image, close to the ground, and simultaneously uses a mode of focusing on enhancing a far area where a driver is interested, and thereby achieves efficient and rapid defogging treatment for the fog weather road image, and can be widely used in fields of a safety driving assist system, autonomous car driving and the like. Additionally, the rapid defogging method for the road image is low in space-time complexity and high in working speed.

The invention discloses a retrieval system oriented to meta keywords of a source document. The retrieval system comprises a source document receiving module, a definition module, an indexing module, a retrieving module, a retrieving result processing module and a source document and meta keywords management module. The retrieving result processing module comprehensively considers the distribution character of the semantic weight of the meta keywords when sequencing the source document in a retrieving result set. The invention has the positive effects of introducing the meta keywords as a middle layer to ensure the source document to form loose coupling with the index thereof, introducing the semantic weight of the meta keywords and processing the distribution character of the semantic weight of the meta keywords, thus completing the compensation of the semantic weight of the meta keywords during sequencing computation and increasing the reasonability of the sequencing of the retrieving result. The invention has lower requirement on computing resources.

The invention discloses a fabric weft inclination rapid-detection method based on machine vision. The method comprises the following steps that a CCD video camera is adopted to collect texture images of fabrics, the best segmentation threshold value T is automatically found by using the Otsu method, image target and background segmentation is conducted on the texture images of the fabrics according to the best segmentation threshold value T, and binary texture images of the fabrics are obtained; the stripe trend in the binary texture images of the fabrics is tracked in a window pixel searching method, isolated point noise is removed, a stripe tracking starting point and end point are determined, and finally the stripe weft inclination angle is obtained; a plurality of stripes are tracked by adopting a stripe space domain retraining method, a set of stripe weft inclination angle values are obtained, selective averaging is conducted on the stripe weft inclination angle values, and final fabric stripe image weft inclination angle values are obtained. The method effectively avoids influence on weft inclination angle detection of environment light, is high in fabric weft inclination angle detection accuracy, and can rapidly detect the fabric weft inclination angle in real time.

The invention discloses a field observation data transmission method and device. The device comprises data transmission devices of a sending terminal and a receiving terminal, and the data transmission devices of the sending terminal and the receiving terminal separately comprise two parts, namely a protocol board and a data radio station board card. The transmission method comprises the followingsteps: improving a response mechanism, a timeout mechanism and a window mechanism of a selection-retransmission ARQ; extracting observation data, and sending the observation data to a data receivinghost; and finally integrating a power management function and a working time management function on the protocol board. The field observation data transmission method and device disclosed by the invention have the advantages that after the realization processes of various reliability protocols are analyzed, the low-power-consumption design is incorporated into the protocol improvement, and under the assistance of the power management on the respect of the hardware system, the transmission power consumption can be effectively reduced during field data transmission, the volume is small and the cost is low.

The invention discloses an optimistic concept-based incomplete decision information system rule extraction algorithm, provides an optimistic concept lattice based on a compatible relationship for thefirst time for the rule extraction algorithm of the incomplete decision information system, and provides a corresponding rule extraction algorithm of the incomplete decision information system on thebasis of the optimistic concept lattice. According to the algorithm, a granularity concept is introduced; the algorithm has the following steps: solving an optimistic concept of each layer in a granularity space from coarse to fine; and obtaining the simplest rule in the decision information system according to the relationship between the optimistic concept and the decision attribute in the decision information system, setting whether the discourse domain element of the extracted rule covers the whole discourse domain as an algorithm termination condition, and finally realizing the rule extraction process of the decision information system. Theorem proves and instance analysis can explain correctness and effectiveness of the new algorithm, and effectiveness and rapidity of the algorithm are verified through experiments.

The invention relates to a computer text classification system and a text classification method thereof. The computer text classification system comprises a text preprocessing module, a text formalizing module, a text weight calculation module, a model training module and a noise reduction module. The computer text classification system has the advantages that the time for computer text classification and space complexity can be effectively reduced, and accordingly computer text classification is quick, effective and accurate.

The invention discloses an order-preserving sequence pattern mining method and relates to the technical field of electric digital data processing. According to the method, a candidate mode is generated by using a mode fusion method; the number of candidate modes is reduced, and the mode support degree of the candidate modes is calculated through a series of conversion and verification steps. The defects that in the prior art, for mining frequent patterns from a time sequence, solving accuracy, generality and completeness are difficult to consider at the same time, it is difficult to ensure that important information is not lost when the time sequence is processed, and it is difficult to analyze key trends by mining the frequent patterns are overcome.

The invention discloses an application layer DDoS attack identification method based on an access rhythm matrix, which is based on a data structure of the access rhythm matrix, uses a variation anomaly degree to carry out attack detection, and uses an outlier point to identify an attack host IP. The time complexity and space complexity of the method are low, and the actual network condition will not be affected. The detection system integrated with the algorithm can be deployed between the nearest routing and the host. The inbound traffic packets can be obtained by port mirroring technology, and the access rhythm matrix can be constructed to detect DDoS attacks at the application layer and identify the attacking host IP.

The method includes following steps: after decoding, acquires norm NDCT of DCT coefficient matrix of current macro block / sub block; calculates the transformed norm Nq of quantify matrix set by the current macro block / sub block; calculates relevant quantify step size of DCT coefficient matrix being relative to transformation of quantify matrix DCT, a=Nq / NDCT; compares a with experience-threshold k; if a<k, then ends the process. If a>k, then makes smoothing process for current macro block / sub block and their neighboring pixels. The unit used in video encode and decode system consists of DCT coefficient matrix norm calculating device, quantify matrix DCT transformation calculation device, relevant quantify step size calculation device, comparator and smooth-processing device.

The invention discloses an Internet resource sharing method based on the HBase. The method includes setting an HBase model of Internet resource sharing; performing table division on the model, namely dividing the model into four tables: a resource table, a resource sharing information table, a user transmission sharing table and a user receiving sharing table; implementing resource sharing through the Internet, namely, adding sharing records, deleting the sharing records, transmitting the shared records acquired by the user and receiving the shared records acquired by the user. Compared with the prior art, the method has the advantages that the mass Internet resource sharing problem is solved, the speed of a user acquiring the shared records is guaranteed, and the practicality is high.

The invention provides an autonomous ship route planning method based on a D*Lite optimization algorithm. The method comprises the steps that S1, in the navigation process of an autonomous ship, obtaining environment information according to a preset time interval, and according to the currently obtained environment information and the environment information obtained last time, if the environmentinformation in a current target map area is changed, judging whether a target route point of the autonomous ship is changed or not; and S2, if the target path point is changed, updating the target route point, taking the current route point as a navigation starting point, determining a termination point according to the updated target route point, and planning a route for the autonomous ship to continue to sail by adopting the D*Lite algorithm according to the environmental information, the navigation starting point and the termination point in the current target map area, wherein the targetmap area is a map area influencing route planning between the current route point and the target route point. When the position of the target route point is changed in the navigation process of the autonomous ship, the route can be dynamically replanned in real time.

The invention discloses a method for obtaining and verifying kinematic parameters of industrial robot based on a space vector method, which is used for solving the technical problem of poor practicality of the obtaining method of the kinematic parameters of the existing industrial robot. The technical scheme is as follows: key point coordinates are extracted from the 3D model of an industrial robot; DH parameters of the industrial robot are soved by the space vector method; and then the obtained DH parameters are verified. Since the robot 3D model provided by the industrial robot manufactureris used to solve the kinematic parameters of the robot, any hardware equipment and related costs are not required; meanwhile the operator is not required to master the professional kinematic parametercalibration method; the 3D modeling software can be used for solving the robot kinematic parameters, so the space-time complexity of the traditional calibration method is reduced; and the hardware has low cost and good practicability.

The invention discloses a method for selecting knowledge of a computer-assisted decision making system based on an undistinguishable relation. The method is characterized in that a computer comprises a central processing unit and a memorizer, the computer is connected with data collecting equipment, data collected by the data collecting equipment are a set of attribute values and target categories for monitoring targets, and the attribute values are in a continuous type or a discrete type; a target database, a knowledge set base, a basic knowledge base, a session module and a question processing module are stored in the memorizer; the central processing unit controls the memorizer and executes the first step of constructing the target database, the second step of determining decision making questions, the third step of constructing the knowledge set base, the fourth step of selecting the knowledge and the fifth step of updating the basic knowledge base.

The invention belongs to the technical field of the artificial intelligent communication data monitoring, and discloses an artificial intelligent communication data monitoring system and monitoring method. The artificial intelligent communication data monitoring system comprises a communication data collection module, a master control module, an analysis module, a feature extraction module, an error correction module, a compression module, a data storage module, and a monitoring display module; the error correction on the received communication data is realized through the error correction module, and the communication efficiency is improved; and meanwhile, a tolerance parameter value of the next compression period is dynamically adjusted according to the step length change of the adjacentcompression period through the compression module, the record quantity of the stable data with small fluctuation can be reduced, and the data compression rate is improved; an operation of importing the complex mathematic computation is avoided, the algorithm space-time complexity is relatively low, and the system is suitable for compressing mass monitoring data.

The invention provides a monitoring data compression method for improving a spinning door transformation SDT compression method. The improvement is as follows: using loose standard with large tolerance parameter DeltaE for non-sensitive data in original data, and using strict standard with small tolerance parameter DeltaE for the sensitive data in the original data, namely, the tolerance parameterof the strict standard is less than the tolerance parameter of the loose standard; automatically the tolerance parameter DeltaE according to the data fluctuation size in the different loose or strictstandard ranges; furthermore, representing the change fluctuation tendency of the current data by using the step length of the compression period, wherein the tolerance parameter DeltaE is increasedwhen the step length of the compression period is enlarged, and the tolerance parameter DeltaE is reduced when the step length of the compression period is reduced. The storage amount of the monitoring data can be effectively reduced, and the low compression error can be guaranteed at the same time.

The invention relates to an action detection method and device, electronic equipment and a storage medium. The action detection method comprises the following steps: acquiring a video stream of a scene area; detecting an action direction of a target object in the scene area based on a plurality of image frames in the video stream; detecting skeleton key points of the target object in the image frame; and determining action information of the target object according to the detected geometrical relationship between target key points corresponding to the action direction in the skeleton key points.

The invention discloses a large-scale image retrieval method and system based on a deep convolutional neural network (DHN), large-scale image retrieval is carried out by constructing the deep convolutional neural network (DHN), and the DHN comprises ResNet50-based feature extraction, channel and space attention (CSA)-based feature refinement, a classification layer, a hash layer and a weight layer. The DHN realizes pixel saliency attention from bottom to top through the CSA, and realizes semantic constraint from top to bottom through classification label supervision; the DHN adopts an adaptive weighted learning algorithm to generate a weight for each bit of Hash code, and then directly generates a short Hash code from a long Hash code according to the importance of the bit represented by the weight. The method provided by the invention has higher Hash code generation precision and speed, thereby being suitable for large-scale image retrieval tasks.