122results about How to "Guaranteed computing speed" patented technology

The invention belongs to the technical field of autonomous navigation of intelligent traffic vehicles, and particularly relates to a vehicle mileage calculation method based on double-eye vision. The vehicle mileage calculation method comprises the following steps of obtaining a video stream of a double-eye camera which is fixedly arranged at the top part of a vehicle, and transmitting to a vehicular processor; respectively extracting the features of a left image and a right image from the image of each frame of video stream of the double-eye camera, combining with the features of a left image and a right image of the previous frame, and searching a matched feature point set by a feature matching method; according to the matched feature points of the previous frame, using a three-dimensional vision method to calculate the space coordinates of a corresponding three-dimensional point; re-projecting the space coordinates of the three-dimensional point in the previous step to the coordinates of a two-dimensional image of the existing frame, and using a GN iteration algorithm to solve the minimum error of re-projection, so as to solve the motion transformation value of the vehicle of the adjacent frame; according to the motion transformation value of the vehicle, accumulating and updating the mileage information of the vehicle motion. The vehicle mileage calculation method has the advantages that by matching and combining the quick features and the high-precision features, the precision is improved, and the calculating speed is guaranteed.

The invention discloses a moving target detecting and tracking algorithm evaluation method based on data set compensation. The moving target detecting and tracking algorithm evaluation method automatically adds quantification factors of eight performance interference factors for data sets, and allows the user to select variation factors when evaluating an algorithm conveniently; the moving target detecting and tracking algorithm evaluation method can customize weight sizes of the variation factors according to degree of sensitivity of a specific evaluated algorithm to specific variation factors, so that comprehensive quantification factors corresponding to the data sets truly reflect the algorithm application environments, and finally the objectiveness of evaluation results is ensured; at last, since the comprehensive quantification factors are comprehensive quantification embodiments of the variation factors affecting algorithm performances in the corresponding data sets, the comprehensive quantification factors of the data sets are adopted for performing weighted compensation for threshold values used for carrying out success determination of the evaluation index precision OverlapP, can show different influences of the data sets with different variation factors on the algorithm performances in the evaluation result, can simulate results of the algorithm under some interference conditions, and provides certain guidance for the further improvement direction of the algorithm.

The invention discloses a neural network model training method. The method comprises a step of obtaining a first image feature and a second image feature, wherein the first image feature is an image feature outputted by a picture A through a trained model, and the second image feature is an image feature outputted by the image A through a model to be trained, a step of obtaining a classification probability which is a classification probability outputted by the second image feature through a classification layer of the model to be trained, and a step of updating the parameters of the model tobe trained until the second image feature and the first image feature are fitted according to the first image feature, the second image feature and the classification probability. The invention also discloses an electronic device and a storage medium. The invention provides the neural network model training method, the electronic device and the storage medium, the parameters of the model to be trained are updated according to the first image feature, the second image feature and the classification probability, a small model with the same precision of a trained large model is obtained through training, and the calculation speed and accuracy are ensured.

The invention discloses a frequency-domain aeroelectromagnetic method 2.5 dimension band landform inversion method. The method comprises the following steps of 1) defining a target function, setting iteration times i, fitting precision and the maximum iteration time and inputting an initial model and inversion data, wherein the i equals to 0; 2) carrying out forward calculation, calculating a forward equation KF=b so as to acquire secondary magnetic fields Hx and Hz; 3) calculating a fitting error, if the fitting error reaches setting precision or the maximum iteration time, quitting calculation, otherwise, continuously calculating; 4) using pseudo-forward modeling to calculate a Jacobian matrix and acquiring a model updating step; and 5) updating a model parameter and satisfying the following equation: mk+1=mk+deltam. In the invention, band landform inversion is considered; a landform influence is eliminated; and frequency-domain aeroelectromagnetic method 2.5 dimension inversion under a fluctuating land surface condition is realized. An effective calculating method is provided for mountain-area aeroelectromagnetic data processing and interpretation.

The invention discloses a temperature influence-considering optimal power flow algorithm of a power system. As the traditional optimal power flow is processed by taking power system network parameters as constants and the influence of the electric heating coupling relation of the actual power grid line is neglected, the calculation result of the traditional optimal power flow has great deviation to the scheduling operation result of the actual power grid. On the basis, the temperature influence-considering optimal power flow algorithm of the power system is characterized in that a temperature influence-considering optimal power flow model of the power system is established according to the electric heating coupling relation between line temperatures and line resistance. Besides, the problem of the temperature influence-considering optimal power flow algorithm of the power system is solved by use of a simplifying primal-dual interior point method, and generalized inequality constraints containing only upper limit constraints are formed by simplifying the inequality constraints in the optimization model, and the writing of a program can be greatly simplified. Finally, the simulation and analysis results of a plurality of examples indicate that the model and the algorithm are higher than the traditional optimal power flow in calculation accuracy and calculation efficiency and have significance to the scheduling operation of the power system.

The invention discloses a remote sensing image classification method of a double-branch fusion multi-scale attention neural network. The method comprises the following steps: reading a multispectral image from a data set; after an image matrix is obtained, preprocessing image data by using superpixels; performing normalization operation on the data, and taking a block from each pixel in the normalized image matrix to form an image block-based feature matrix; selecting a training set and a test set; constructing a classification model of the convolutional neural network based on two-channel sparse feature fusion; training the classification model by using the training data set; and utilizing the trained classification model to classify the test data set. According to the method, consideringfrom the characteristics of the image, the characteristics can be extracted by self-adapting to the size of the target region object in the image, and a new central pixel offset strategy is adopted for the boundary pixels, so that the classification accuracy of the boundary pixels is improved, and the operation speed of the whole training process is also improved.

The invention provides a method for automatically sorting and rating metallographic structures of different materials, comprising the following steps: firstly creating a database of crystal grain series of three different materials (20CrMnTi, CF steel and 55 steel) at different multiples; reading in one metallographic structure image F with the size of M*N, wherein M and N are positive integers; recognizing metallographic materials in the read-in metallographic structure image F, and recognizing kinds of the materials by adopting a convolutional neural network; and then carrying out rough sorting and fine sorting on the database, and rating grain size scales of the recognized materials by adopting a data enhancement and transfer learning method. The method provided by the invention firstlyapplies a convolutional neural network algorithm and a transfer learning algorithm to recognition and rating of metallographic images and can realize automatic recognition and rating of the metallographic structures of different materials, and accuracy and efficiency of the method provided by the invention are also greatly improved.

The invention discloses a massive blackout recovery technique which is realized through massive blackout recognition, effective isolation of a transfer region, analysis on a transfer load, optimized reconstruction containing distributed power supplies, dynamic adjustment operation and optimization and recovery on a running mode, power can be recovered and supplied within a short time after generation of massive blackout in a distribution grid, by means of convenience of distribution automation, a larger number of switching operation can be completed within the short time, during the process, the transfer region is effectively isolated to guarantee power supply safety, during optimized reconstruction of the distributed power supplies, a reasonable starting principle and a calculation mode are adopted to ensure the feasibility and rationality of a reconstitution scheme, during dynamic operation adjustment, small-range network reconstruction is adopted to the greatest extent during secondary network reconstruction so as to ensure time requirement of calculation speed and power recovery, and after fault removal and bus decompression relief, the original running mode can be timely recovered so as to guarantee optimized running of the running mode.

The invention discloses a deep Q network driven power system operation mode automatic optimization adjustment method. The method comprises steps of determining a load fluctuation range by taking a typical operation mode as an adjustment reference mode, and generating a large amount of target mode sample data for training and testing in combination with a Latin hypercube sampling method; determining all feasible single control actions in the power grid model, numbering the control actions, and setting the control actions as action spaces; initializing a power grid model, judging whether an untrained sample exists or not, if yes, assigning load data in the sample to the power grid model, performing convergence optimization processing on the output data of the generator in the current operation mode, and if not, terminating training and the like. According to the method, the problem that the optimal power flow method is difficult to converge when solving the multi-target optimal power flow is solved while the calculation speed is ensured, various indexes of the mode obtained after adjustment have no overlarge deviation, and method reference is provided for application of deep reinforcement learning to power grid optimization and control problems.

The invention belongs to the technical field of electric vehicle charging, particularly relates to an electric vehicle charging system and device based on mobile charging pile scheduling, and aims tosolve the problems that an existing mobile charging pile scheduling method is small in scheduling range, a charging vehicle cannot be found, and the electric vehicle is inconvenient to charge due to inconvenient recovery and maintenance. The system disclosed in the present invention comprises: an information interaction unit for acquiring charging request/reservation/charging completion information; a data acquisition unit used for acquiring a current charging demand, current mobile and fixed charging pile distribution diagrams and residual electric quantity information of each mobile chargingpile; a charging demand prediction unit used for predicting the charging demand of the next time period; a charging pile scheduling unit used for generating a charging demand prediction distributiondiagram and performing charging pile scheduling in combination with the data of the data acquisition unit; a charging unit used for selecting the optimal charging pile to charge the electric vehicle.According to the system, the mobile charging pile is scheduled in advance on the basis of predicting the charging demand, the charging efficiency of the electric vehicle is high, the accuracy is high,and the real-time performance is good.

The invention provides a method for detecting the parameters of a double-fed wind generating set. The method includes the steps that first, the double-fed wind generating set is connected into a power grid for generating electricity, and a voltage signal and a current signal of the double-fed wind generating set are collected through wave recording equipment; second, a master controller of the wind generating set is used for collecting the wind direction, the wind speed, the pitch angle, the generator speed, the generator torque and a master control low voltage ride-through signal; a converter controller is used for collecting a converter low voltage ride-through signal, a converter Chopper actuating signal and a converter Crowbar actuating signal; third, the equivalent parameter and the low voltage ride-through response delay time of a double-fed wind generator are acquired according to the current signal and the voltage signal. Compared with the prior art, by means of the method for detecting the parameters of the double-fed wind generating set, errors caused by an experimental platform which is composed of parts of devices are avoided, the reciprocal influence of the double-fed wind generating set and the power grid can be reflected, and the detection method is high in operability, simple and practical.

The invention provides a commodity association degree determination method and device, and relates to the technical field of electronic commerce. The method comprises the following steps: according tothe category purchase data and the user non-purchase behavior data, obtaining an association relationship corresponding to a category purchase behavior of the purchase target category and a user non-purchase behavior and a category purchase behavior of other commodity purchase categories; and based on the association relationship and the weights corresponding to the category purchasing behaviorsand the user non-purchasing behaviors, determining the association degree between the commodity purchasing categories, and selecting the associated categories according to the association degree. According to the method and the device, user behavior data of browsing, shopping cart adding, attention, purchasing and the like are comprehensively analyzed. The product association degree is calculatedthrough the quantifiable association coefficient, the storage space can be saved, the calculation speed can be ensured, the method can be applied to different commercial marketing scenes, accurate product positioning analysis can be met, marketing activities with high recall rate can be met, and the use experience of users can be improved.

A shielding door obstacle detection method is characterized by firstly judging whether that scene is changed significantly after the image is taken; if the scene is changed significantly, the detection area in the taken image is compared with the detection area in the train sample image to determine whether an obstacle exists or not; if the scene is not changed significantly, the obstacle recognition model is used to identify whether an obstacle exists or not. The invention also relates to a shielding door obstacle detection system, comprising a camera arranged above a gap between the shielding door and a vehicle door, a data processing unit electrically connected with the camera, a data storage unit electrically connected with the data processing unit, and a background computer communicatively connected with the data storage unit through a network communication interface. The screening door obstacle detection method and the system adopt different recognition schemes according to the scene change situation, thereby effectively avoiding the influence of the scene change on the foreign object recognition and improving the accuracy of the obstacle recognition.

The invention provides a lazy interaction mode-based video tracking method. The invention aims to eliminate the defect of low computational speed or insufficient robustness of an existing video tracking algorithm. According to the method, when a target is partially obscured and illumination changes obviously, the abnormal conditions of blocks are judged; the judgment results of the blocks can be divided into three situations, wherein the first one is the departure of the blocks from a target, the second one is proportional imbalance of positive samples and negative samples, and the third one is the excessively low confidence value of the trackablility of the blocks; with the above conditions adopted, abnormal blocks can be screened out, and re-modeling and tracking are performed; and since a plurality of blocks are required to be tracked, the KCF algorithm is introduced to be adopted as a basic tracking operator to ensure the efficiency of the tracking. With the method of the invention adopted, the robustness of resampled samples can be ensured to the greatest extent, and therefore, the continuity and correctness of a tracking effect can be achieved, and at the same time, a higher calculation speed can be guaranteed.

The invention discloses a power system frequency response analysis method based on data-model fusion driving, and the method comprises the steps of carrying out the preliminary prediction on the system frequency response through employing a model driving method, and carrying out the error correction on a preliminary measurement result through employing a data driving method; and further providing a power system frequency response analysis method based on data-model fusion driving. The method specifically comprises a data processing link, an offline training link and an online application link. According to the method, the calculation precision is greatly improved on the premise of ensuring the calculation speed, meanwhile, the dependence degree of the model on the sample data is reduced, and the interpretability of a result is improved; the method can be applied to situation awareness and real-time evaluation of frequency security and stability, and provides a rapid and effective decision basis for the subsequent scheduling and control so as to prevent frequency collapse events and improve the stability of a power system.

The invention belongs to the field of electric logging value simulation, and discloses a method for simulating fracture-vug type reservoir body electric logging response. The method includes establishing an equivalent fracture-vug type reservoir body stratum model according to regional geological data, determining basic parameters of the equivalent fracture-vug type reservoir body stratum model; establishing finite element functions according to the DC logging response of the equivalent fracture-vug type reservoir body stratum model; carrying out grid discretion and resistivity assignment on the equivalent fracture-vug type reservoir body stratum model based on finite element theory; and assembling and solving finite element conductance array and obtaining depth lateral logging value. The method solves the problem of trans-scale grid dissection of a fracture-vug type reservoir body, effectively carves the boundary of the fracture-vug type reservoir body, scientifically simulates the values of fracture-vug type reservoir body electric logging, and provides theoretical basis for identifying and evaluating electric logging of carbonatite and volcanic fracture-vug type reservoir body according to electric logging data.

The invention discloses an image distortion method based on matrix inverse operation in a virtual reality (VR) mobile end. The method comprises the following steps: 1, reading a current frame and an equipment state; 2, converting coordinates under a screen coordinate system at current-frame time into coordinates under a standard equipment system at the current-frame time; 3, according to the coordinates under the standard equipment coordinate system at the current-frame time, obtaining coordinates in a world coordinate system; 4, according to the coordinates in the world coordinate system, obtaining coordinates under the standard equipment coordinate system at corresponding next-frame time; 5,performing linear transformation on the coordinates under the standard equipment coordinate system at the next-frame time so as to finally obtain coordinates under the screen coordinate system through the transformation; and 6, endowing the coordinates under the screen coordinate system at the corresponding next-frame time with pixel RGB values of the coordinates under the screen coordinate system at each current-frame time so as to obtain a final distortion image. The method is a method for generating intermediate frames in VR, can effectively reduce jittering in a VR game, and improves user experience.

The invention discloses an unmanned aerial vehicle obstacle avoidance flight method based on digital spatial slices. The method comprises the following steps of: step 1, acquiring a picture shot by anunmanned aerial vehicle; 2, adjusting the resolution of the picture and performing processing to obtain a grey-scale image; 3, for the single grayscale image, extracting a plurality of slices from different regions of the picture to form a set I by using a digital spatial slicing algorithm; 4, simultaneously inputting the slice picture sets in the set I into a neural network, and outputting corresponding obstacle calibration information to form a set R; 5, estimating the position and direction of an obstacle according to the set R; 6, planning a flight mode of the unmanned aerial vehicle, andsetting a yaw angle and a speed; and step 7, enabling the flight control system of the unmanned aerial vehicle to perform motion control according to the acquired speed and yaw angle information. Themethod has a very good effect on tree detection in an under-forest environment, and hardware adopted by the method is light in weight and low in energy consumption, and therefore, the method is suitable for being used on a small unmanned aerial vehicle.

The invention relates to a visual inertial odometer method and system for online calibration of equipment parameters, which adopts a Kalman filter to online calibrate external parameters of a camera and bias of an IMU (Inertial Measurement Unit), and adopts nonlinear optimization and a sliding window to online calibrate internal parameters of the camera. Image data and IMU data are acquired by using the camera and the inertial measurement unit carried by the equipment, parameters are calibrated without depending on the equipment and a calibration plate, the calculation speed is ensured by using a Kalman filter to calibrate external parameters of the camera and bias of the IMU, and the precision requirement is met by using nonlinear optimization and a sliding window to calibrate internal parameters of the camera. And the Kalman filter and the nonlinear optimization sliding window are in a tight coupling relationship, so that the method is more robust. The invention provides a set of perfect algorithm process and system of the real-time Kalman filter and the nonlinear optimization sliding window visual inertial odometer, the performance of the visual inertial odometer in different environments and different devices is remarkably improved, and the algorithm has the advantages of universal process, strong pertinence, simple and easy online operation process and the like.

A tetrahedral mesh virtual cutting method based on route separation comprises the steps that 1, data of a tetrahedral mesh is read, and a surface triangular patch set of the tetrahedral mesh is obtained; 2, a collision detecting thread is started, and a thread from the surface triangular patch set to a screen is displayed; 3, after collision detection occurs, track discretization of a tool is started, feature point processing is carried out, and meanwhile selection of an optimal route is started; 4, a tetrahedron shared by two first points on the optimal route is separated; 5, the two points in the step 4 are deleted. In the cutting process, each colliding patch is evaluated through an evaluation function, the optimal point is selected to serve as a route of a cutting track, and then tetrahedron separation is carried out; due to the fact that only topological separation exists in the cutting process, and no pathological tetrahedrons are generated, the stability of finite element computing is improved, the number of the tetrahedrons is not increased, and the computing speed is kept.