47results about How to "Reduce the calculation dimension" patented technology

The invention discloses a deep-learning-model-based steganography image detection method and system. The method comprises the following steps that: an initial deep learning model for fitting and analyzing a digital image steganography rich model feature is preset; a supervised pre training algorithm is selected to carry out pre training on the initial deep learning model to obtain a pre-training deep learning model; the pre-training deep learning model and a deep classification network are integrated to generate an integrated deep learning model, the integrated deep learning model is trained according to a training data set obtained in advance and thus a final deep learning model is generated, a steganography image is detected based on the final deep learning model, and a detection result is outputted. Therefore, a steganography image and a carrier image can be distinguished accurately. Moreover, the data calculation dimension is low; the computing load is low; and the steganography image detection speed increases.

The invention discloses an old people abnormal behavior identification method, which relates to image information processing of image feature extraction, and is based on an improved and trained stack-type convolution ISA model. The old people abnormal behavior identification method comprises the steps of: establishing a video sample database of old people behavior patterns and partitioning a video sample; preprocessing video sample data; extracting temporal and spatial characteristics from the video sample data by utilizing the improved and trained stack-type convolution ISA model; clustering the temporal and spatial characteristics to obtain a visual word list and a video visual word frequency histogram; training an X<2> kernel support vector machine SVM classifier model; and identifying old people abnormal behaviors. The old people abnormal behavior identification method does not need to carry out human body segmentation and background modeling, directly extracts the temporal and spatial characteristics from the video data through establishing an old people behavior database and adopting an unsupervised learning method, realizes the identification of the old people abnormal behaviors, and overcomes various defects in the prior art.

The invention provides a micro-expression recognition method based on the main direction of optical flow, which comprises the following steps: interpolating a micro-expression sequence formed by framing a micro-expression video; positioning the eyes and nose of each frame, and calculating the position of the middle point of the eyes as the origin of the coordinate axis; calculating the distance dem and dmn from the origin of the coordinate axis of each frame to the center of each eye and the nose; constructing normalized mesh according to dem and dmn; calculating the optical flow direction ofeach pixel region in all adjacent frame grids and clustering to obtain the optical flow main direction; a rectangular characteristic region being formed in the main direction of the optical flow, andthe pixel region in the main direction of the optical flow being assigned with weight coefficients to obtain the optical flow field optimized by adjacent frame grids, and the curves formed by connecting the related data drawing points being smoothed to obtain the global optical flow field; the pixel region in the main direction of the optical flow being assigned with weight coefficients. A training classifier inputs each optical flow characteristic region and compares the sum of similarities between the traversed-level microexpression sequence and the trained microexpression sequence with S toobtain a classification result.

The invention provides an island microgrid distributed coordination control method considering a communication problem. Based on distributed information and in consideration of the actual communication problem encountered in the information transmission process, the stability control problem of the frequency and the voltage of the AC microgrid under off-network autonomy is effectively solved, andthe system instability caused by excessive fluctuation of centralized communication pressure and frequency voltage and communication interruption is avoided. A droop controller is designed, and the dynamic performance of the frequency and the voltage is optimized. In secondary control, a centralized event trigger consistency protocol is introduced, communication resources are saved, irregular switching of a network topology structure is considered, a data compensation mechanism is designed under a communication interruption fault, and a scheme of combining neighbor data compensation and centralized data compensation is adopted; and on the aspect of data prediction, an improved exponential smoothing model based on a time sequence is provided, and finally secondary adjustment of the frequency and the voltage is achieved.

The invention belongs to the field of information technology, the invention relates to a computer network authentication system and a computer program. The IP terminal of the computer network authentication system comprises a login unit and a registration unit. The registration unit comprises a basic information interface and an authentication module. The authentication module comprises a face recognition interface, an iris upload interface and a fingerprint input interface. The login unit comprises an identification module, wherein the identification module comprises a dynamic two-dimensionalcode interface, an authentication code verification interface and an iris verification interface; the dynamic two-dimensional code interface information is scanned to enter the verification code interface, and the verification code interface information enters the iris verification interface. The invention integrates real name authentication, face recognition authentication, eye iris authentication and fingerprint authentication to avoid leakage of personal identity and privacy information. Scanning two-dimensional code to improve the convenience of system login.

The invention discloses a matrix decomposition recommendation method based on joint clustering. The matrix decomposition recommendation method includes the steps that a user-item rating matrix is constructed; the user-item rating matrix is divided into multiple categories through joint clustering; for the categories obtained after clustering, a probability matrix decomposition method is used for parallely predicting unknown scores of all the categories, and recommendation is carried out according to the predicting scores. Close correlation in the clustering and high precision of the probability matrix decomposition algorithm can be used fully, and for the big data processing problem in the information overload times, recommendation can be conducted at high speed while high precision is ensured.

The invention discloses a power load clustering analysis method, device and equipment. The method comprises: calculating the shortest bending path of the daily load curve data of the power consumer through a dynamic time bending path algorithm; the minimum average distance is calculated, the data representation of the power consumer is determined, effective integration of historical load data of the power consumer is completed, redundant calculation of repeated information is reduced, calculation expenditure is effectively saved, and the problem that the performance of a traditional clusteringalgorithm is reduced due to time fluctuation and uncertainty of the load is solved to a certain extent. According to a preset fuzzy margin, compression dimension reduction processing is carried out on the data representation set, so that the calculation dimension is greatly reduced, and the calculation efficiency is improved. Therefore, the method provided by the invention solves the technical problems of high calculation dimension by using the original daily load data, information redundancy, low calculation efficiency and large calculation result error caused by time uncertainty and volatility in the existing power load clustering analysis method.

The invention relates to a power distribution network operation management technology, and specifically relates to a power distribution network multi-dimensional typical scene generation method basedon load clustering and network equivalence. The method comprises the following steps: 1, dividing load nodes into n types according to power fluctuation features, and dividing distributed power generation into photovoltaic power generation and wind power generation; 2, performing equivalent processing on the grid structure of the power distribution network, and simplifying the grid structure inton + 2 single-node power distribution networks; 3, respectively generating n + 1-dimensional and 1-dimensional typical scenes for voltage drops of equivalent lines of load/photovoltaic power generationand wind power generation; and 4, restoring the typical line voltage drop scene to each load/distributed power generation node power typical scene of the power distribution network. According to themethod, the problem of difficulty in voltage distribution evaluation in the power distribution network is solved, so that access position planning of distributed power supplies, loads, power electronic equipment and the like has a better effect. According to the multi-dimensional typical scene generation method, the calculation capability is not influenced by the number of nodes, and the multi-dimensional typical scene generation of the large-scale power distribution network operation state can be realized.

The invention discloses a small disturbance stability margin probability assessment method considering wind-speed random fluctuation. The method considers the problem of uncertainty of small disturbance stability margin caused by randomness and uncertainty of wind power generation power, and is characterized by modeling and simplifying the system through a generalized short circuit ratio method, simulating fan output through a multiple-wind-farm random output model considering wind speed correlation between wind farms, providing a fan grid-connected oscillation stability margin probability assessment model, and then, solving oscillation stability margin and risk evaluation indexes through a Monte Carlo method. Simulation results show that the small disturbance stability margin probabilityassessment method considering wind-speed random fluctuation can effectively calculate the probability of system instability under different wind speed correlation and different stability margins, andhas a certain guiding significance for multiple-wind-farm grid-connected small disturbance stability margin probability assessment.

The invention discloses an electric vehicle charging station distributed electric energy management method. The method is implemented through a processing unit and comprises the following steps: (A) obtaining at least one to-be-planned pane of each electric vehicle; (B) for each electric vehicle, obtaining charging electric power or discharging electric power of the electric vehicle in each to-be-planned pane according to the electric vehicle information corresponding to the electric vehicle, the at least one selling price, the at least one buying price, the at least one bidding price and thedegradation cost; (C) obtaining the total consumed electric power of the charging station in each time pane; (D) judging whether at least one overload pane exists or not according to the total consumed electric power of each time pane; and (E) when it is determined that the at least one overload pane exists, adjusting the buying price of each overload pane, and repeating the steps (B) to (D) untilit is determined that the at least one overload pane does not exist.

The invention provides a path rapid generation method in intelligent driving. The path rapid generation method includes the steps that 01, the positions of the starting point and the ending point of ato-be-generated path are determined; 02, the heading angles of the starting point and the ending point of the to-be-generated path are determined; and 03, a vehicle passable path connecting the starting point and the ending point is generated, and the vehicle passable path meets a vehicle kinematics model. Higher system real-time performance can be achieved in intelligent driving, the algorithmiccalculation difficult is lowered, algorithmic real-time performance is further improved, the learning rate is further introduced, and algorithmic convergence performance is improved.

The invention provides a virtual node-based power grid single line diagram automatic layout method, which comprises the following steps of: reading a power grid topological structure, acquiring all node and line sets, performing first-stage layout on all node lines by applying a first differential evolution algorithm to obtain an initial layout diagram, applying a second differential evolution algorithm to carry out optimized layout of a main region in the second-stage layout, determining the range of a virtual region based on virtual node coordinates, applying a third differential evolution algorithm to carry out layout of the virtual region in the second-stage layout, and carrying out layout in the range of the virtual region corresponding to each virtual node, obtaining the coordinate of each node, and completing the automatic layout of the power grid single line diagram. According to the method, a differential evolution algorithm is applied, a comprehensive layout and routing mode is adopted, a penalty term is introduced into a target function, the problem of automatic layout of the single line diagram of the annular belt radiation power grid is effectively solved, and automatic layout of the single line diagram can be completed more accurately and quickly by introducing virtual nodes.

The invention provides a low-voltage power distribution network fault positioning method based on an adjacent matrix shortest path. A traditional intelligent positioning algorithm is improved, a graphtheory principle is combined, and a fault current monitoring unit installed at the head end of each feeder line section is used for generating an adjacent matrix of a low-voltage power distribution network; a path of the estimated fault current is generated according to the estimated fault current connectivity in the fault state of the estimated section, and a shortest path estimation method is proposed to generate an estimated fault monitoring information matrix; and the iterative search is performed by using a particle swarm algorithm, and estimated fault information most similar to the actual fault information is found as a fault positioning result. According to the method, the accuracy of fault positioning of the low-voltage active power distribution network is effectively ensured, the interference of non-fault area monitoring information on fault positioning is suppressed, the reliability of the low-voltage power distribution network is improved, and the method has important theoretical and practical application values.

The embodiment of the invention discloses a risk early warning control method for an automatic driving vehicle and an intelligent network connection system. The method comprises the steps: a plurality of automatic driving vehicles carry out real-time interaction on the position and speed of each automatic driving vehicle; any automatic driving vehicle calculates forward collision time between the vehicle and the front vehicle and backward collision time between the vehicle and the rear vehicle in real time according to the positions and speeds of the vehicle, the front vehicle and the rear vehicle; if the forward collision time and/or the backward collision time are/is larger than or equal to a preset collision time threshold value, the vehicle sends an early warning prompt to the front vehicle and/or the rear vehicle; after the front vehicle and/or the rear vehicle receives the early warning prompt, the speed is adjusted according to the safety state of the front vehicle and/or the rear vehicle; and the vehicle adjusts the speed of the vehicle according to the adjustment result of the front vehicle and/or the rear vehicle. According to the embodiment of the invention, road network risk control is realized through mutual communication between the autonomous vehicles.

The invention belongs to the technical field of power grid systems, and particularly relates to a medium and long term unit combination accurate modeling technical method comprehensively considering complex large power grid operation constraint conditions, which comprises the following steps: establishing a medium and long term bus load prediction tool considering renewable energy output, including: restoring renewable energy associated bus load sample data; performing medium and long term bus load prediction considering renewable energy data correction; performing medium and long term bus load prediction considering equipment maintenance plan correction; and establishing a medium and long term unit combination and power generation plan compilation tool for promoting renewable energy consumption, wherein the medium and long term unit combination and power generation plan compilation tool comprises a medium and long term unit combination and a power generation plan optimization calculation function module. According to the method, the calculation dimension of the problem model is reduced by considering the aspects of model description, calculation time period, effective integer variable pre-recognition, effective security constraint pre-recognition and the like through the physical characteristics of an actual system, the solving efficiency is greatly improved on the premise of not losing the solving precision, and the engineering application of a medium-and-long-term unit commitment algorithm is realized.

The invention relates to a power grid online risk calculation method based on regression learning and feature mining. The power grid online risk calculation method comprises the following steps: step 1, data collection based on offline modeling; 2, selecting characteristic variables based on a multi-target particle swarm; 3, selecting an optimal parameter of the SVM model based on an improved ant colony algorithm; 4, establishing a regression vector machine model in combination with the key parameters of the training model, and inputting the obtained data into a regression model by using the model after feature screening to carry out one-time offline training; and step 5, performing feature screening on the system operation state according to the result of the step 2, and inputting the feature into the multiplexing network to obtain the node power failure risk in the corresponding state. And performing weighted summation according to the node risk and the state probability to obtain an annual expected risk index. The operation state covered by the method is more comprehensive, the uncertain information of the system can be more fully reflected, and the evaluation result is more systematic globality.