106results about How to "Relieve computing pressure" patented technology

The invention relates to a method and system for screening face images. The method comprises the following steps: according to preset face integrity standards, face size threshold and face position threshold, obtaining a first screening result; according to a preset definition threshold, obtaining a second screening result; according to a preset best angle threshold, obtaining a third screening result; and according to the sum of similarities between each face image and other face images, the definition of each face image and the face deflection angle, obtaining the weighting sum of each face image in the third screening result, and acquiring a face image with the maximum weighting sum for face identification. The method can be used for solving the problem that the face is mistakenly identified due to the low quality of an input image, the data volume of the face images is effectively reduced, and the operation load during face identification is relieved; and moreover, the data volume of the face images is reduced, front face images with high quality are effectively screened to be identified, and the accuracy and real-time performance of face identification are improved.

The invention provides an image identification cooperative computation method and system based on a neural network. A self-adaptive distributed cooperative computation architecture scheme of a mobileuser equipment terminal, a network edge terminal and a cloud server terminal is achieved, according to real-time performance of the user equipment terminal, network edge terminal communication equipment and a cloud server, in combination with context and structure and parameter information of a trained neural network model, computation tasks of the cloud server are distributed to the user equipment terminal and network edge communication equipment nodes, resources such as idle computation power and storage space of mobile user equipment and network edge basic communication equipment can be sufficiently used, and the problems of high deployment cost, high difficulty and long time delay of the cloud server under the situation of high concurrency of a user request are solved; accordingly, onthe premise that the development of an augmented reality application / activity is guaranteed, the computation stress of the cloud server can be relieved, and the purpose of short time delay of the augmented reality application is achieved.

The invention relates to a text to speech (TTS) method and system. The method comprises: performing text analysis and linguistic analysis on an inputted text so as to generate front-end scripts comprising corresponding speech units; obtaining, verifying and correcting the front-end scripts; and obtaining the corrected front-end scripts for synthesizing correction speech. By adopting the provided TTS method and system, the errors of the front-end scripts such as word-segmentation errors and polyphone phonetic notation errors can be corrected so that the synthesized speech is more understandable and more user friendly for users, a conventional TTS's shortcomings of insufficient prediction accuracy of rhythm can be overcome, and the synthesized speech is more natural and more expressive.

The invention discloses a railway foreign matter clearance intrusion detection system and method. The railway foreign matter clearance intrusion detection system comprises at least one monitoring unit, an embedded foreign matter clearance intrusion detection unit connected with the monitoring units and a remote monitoring server. The embedded foreign matter clearance intrusion detection unit comprises an image acquisition chip for acquiring image data, shot by the monitoring units, of a railway area to be detected in real time, an FPGA used for judging whether a target exists in the image data or not through background subtraction frame by frame and extracting the features of the target to form feature vectors of the target if the target exists in the image data, a microprocessor for conducting foreign matter clearance intrusion judgment on the image data according to the feature vectors of the target, a remote monitoring server for confirming alarm information sent by embedded foreign matter clearance intrusion detection platforms and informing trains in relevant areas. By the adoption of the technical scheme, the railway foreign matter clearance intrusion detection system is high in recognition speed, high in detection efficiency and accurate in alarming and is suitable for an occasion needing long-distance along-railway foreign matter clearance intrusion detection.

The invention discloses a statistical analysis method for secret key leakage resisting encrypted data aggregation in a smart power grid. Each user area is responsible for one fog node; the fog node serves as a data aggregation gateway and a data repeater; user electric energy consumption data sent by an intelligent electric meter in a user area is aggregated for the first time by a fog node to obtain a fog-level aggregation ciphertext, the fog-level aggregation ciphertext is signed, and then the fog-level aggregation ciphertext and signature information are sent to a cloud server to be stored;the cloud server stores the aggregated ciphertext data and the signature information of different user areas in a database; data query, statistics and analysis services are provided for a control center of the intelligent power grid system; the cloud server can provide enough information for the control center on the premise of not destroying the privacy of any single user, so that the control center can calculate the sum, average value and variance of power consumption data of all users in a specified user area by utilizing the information.

The invention provides an expressway vehicle detection and multi-attribute feature extraction method based on a local image, and relates to the technical field of intelligent transportation. A video acquisition terminal reads expressway monitoring video in real time and transmits the expressway monitoring video to an edge end, and the edge end analyzes the real-time video data by adopting a background difference method to select a key frame; a cloud end uses a VOC2007 data set and vehicle pictures collected by an expressway to train a YOLO-v3-tiny detection model, the edge end loads the trained YOLO-v3-tiny detection model to predict the position of a vehicle bounding box in the selected key frame, and then a local image of a vehicle is obtained and transmitted to the cloud end; a ResNet-50 residual neural network model is trained by the cloud end by utilizing the training set data with the multi-label type, the edge end loads the trained ResNet-50 residual neural network model, and the acquired local image of the vehicle is input into the neural network model to realize the extraction of multi-attribute features of the vehicle; and the extracted multi-attribute features of the vehicle are made into a label, and the label is uploaded to the cloud end.

The invention discloses an active content caching method based on federated learning, and belongs to the technical field of wireless communication. The method comprises steps that firstly, in each round of communication, a user downloads a global model, training is conducted locally through a stacked automatic encoder, and a local model and implicit features of the user and a file are obtained; secondly, in each round of communication, the user updates the model and sends the model to a server, and all the local models are aggregated to generate a global model; thirdly, after training is finished, the user sends the implicit features of the user and the file to the server, the server firstly calculates the user similarity and the file similarity, then a certain user is randomly selected, and a decoder of the stacked automatic encoder is used for recovering pseudo score matrixes of the user and the file; and finally, scores of the group of users on all the files are calculated by usingcollaborative filtering, and the file with the highest average score is selected for caching. On the premise of ensuring the cache hit rate, a problem of data sharing between neighbor users is effectively avoided, so private data of the user is safer.

The invention relates to the technical field of intelligent driving, in particular to a multi-vehicle collaborative mapping method applied to automatic driving. The multi-vehicle collaborative mappingmethod applied to automatic driving comprises the following steps of step 1, data collection and perception; step 2, preprocessing of point cloud data; step 3, a local map and a global map; step 4, acommunication module; step 5, matching between two vehicles; step 6, after matching succeeds, sending a matching result to the opposite side by each vehicle; and step 7, after each vehicle receives aglobal edge map and trajectory data of the opposite side which are continuously sent by the opposite side, converting data through a matrix obtained by calculation in step 5 and then transferring a result to mileage calculation and mapping part to perform real-time collaborative mapping.

The invention discloses a wireless body area network resource allocation and task unloading algorithm with maximum system revenue. The algorithm is based on a classic bargaining game model in the gametheory. An edge server for providing computing resources is used as a resource seller in the model; and the wireless body area network distributed in the service range of the server is used as a resource buyer, the two parties perform bargaining by taking maximization of own benefits as a target according to a selfish behavior mode in allusion to CPU resources and wireless channel resources, andfinally, an optimal resource allocation scheme and an optimal task unloading scheme of the whole system are obtained. The timeliness and reliability of wireless body area network data processing can be improved.

The present invention discloses a hybrid test method based on a vector finite element and an FPGA. The method comprises the following steps: dividing a to-be-analyzed structure into a numerical sub-structure and an experimental sub-structure; establishing a numerical model in an FPGA by using a parallel computing technology and adopting a vector finite element; writing an external compensation controller in the FPGA by using a control theory of feed-forward plus feedback; designing I / O in the FPGA to implement conversion and transfer between a numerical computation amount and an experimental simulation amount; running an FPGA hardware program regularly by using a LabVIEW real-time module; transferring data between FPGA hardware and a host computer by using FIFO; setting up a displacement loading device of the experimental sub-structure and establishing a connection with an I / O module of the FPGA hardware; and in the host computer, establishing a data communication interface between the host computer and an FPGA hardware terminal, so as to realize visualization of experimental data. The method provided by the present invention can effectively solve the problem of excessive computing pressure caused by numerical sub-structure calculation in a hybrid experiment, and is particularly suitable for a hybrid experiment in which a numerical sub-structure has a large number of computing degrees of freedom and requires real-time operation.

The invention discloses a cloud storage key generation method, a cloud data storage method and related cloud storage data auditing methods. A group user generates a current private key according to a group identity key and a group current time update key; only the group user registering at the current time in the group can obtain the group current time update key, update the current private key and generate labels for cloud data; the user cancelled from the group cannot generate the private key in a current time period, the previous private key becomes invalid, and the labels cannot be generated for the cloud data. According to the cloud storage key generation method, the cloud data storage method and the related cloud storage data auditing methods provided by the invention, for the cancelled user, a cloud no longer updates a signature previously generated by the cancelled user and the computing pressure and economic burden of the cloud are greatly mitigated.

The invention relates to a base frame operation and maintenance technology, and discloses a micro-service architecture monitoring method and device, computer equipment and a readable storage medium. The micro-service architecture monitoring method comprises the steps of: acquiring the operation state of each service node under a micro-service architecture; comparing a preset standard operation state with the operation state of each service node in sequence, and judging whether each service node meets the standard operation state or not; if not, setting the service nodes which do not conform tothe standard operation state as hysteresis nodes; acquiring a call link which has the hysteresis nodes and does not belong to a core link, and setting the call link as a degradation link; and intercepting a request message sent to the degradation link and returning intercepted information. The micro-service architecture monitoring method further relates to a blockchain technology, and informationcan be stored in blockchain nodes. According to the micro-service architecture monitoring method, the overall reliability and stability of the micro-service architecture are ensured, the influence ofthe interception request message on the micro-service architecture is reliably controlled, and the controllability of the operation and maintenance of the micro-service architecture is effectively improved.

The invention relates to a mobile unloading migration algorithm based on an improved HRRN (High Resolution Response Next) algorithm and a multi-attribute decision. The method comprises the following steps: step 1, a single wireless body area network generates a plurality of tasks with different priorities at a certain moment, if the tasks can be locally processed, the tasks are directly locally processed, otherwise, the tasks are uploaded to an edge server; 2, establishing a benefit function of an edge server, and taking the edge server with a high benefit value as an initial edge server of the current task; 3, judging whether the current edge server is suitable or not; 4, judging whether the edge server meets two criteria or not, wherein one of the criteria can be used as a candidate edgeserver; and step 5, according to the benefit function of the edge server established in the step 2, calculating the benefit values of all candidate edge servers, and selecting the edge server with the highest benefit value to migrate.

According to a multi-agent electric vehicle battery replacement scheduling method based on the Internet of Vehicles, a vehicle-road cooperative service is deployed on an MEC platform, and man-vehicle-road cooperative interaction is realized by means of a Uu interface or a PC5 interface and communication modes such as VANET, 4G or 5G; according to a map around an electric vehicle with a battery replacement service requirement, a roadside unit divides a battery replacement station cluster with a high potential cooperation matching degree into a whole and gathers the battery replacement station cluster into a battery replacement area, and the battery replacement area with the maximum service capability probability shares a plurality of electric vehicles with the battery replacement service requirement at the same time; the service rate of each battery swap station is taken as an assessment target to mainly assess the self-service capability, self-service quality and sitting information ofeach battery swap station node and the current self-state of the electric vehicle with the battery swap requirement; and the optimal joint action of global electric vehicles is provided, so that theoverall service balance of each battery swap station is maintained, and the long-term performance of the Internet of Vehicles is improved. According to the invention, the electric vehicle can exchangepower as soon as possible, and each power exchange station can maintain service balance.

The embodiment of the invention discloses an intelligent terminal control method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining and storing a to-be-executed linkage rule; generating a control instruction for controlling the intelligent terminal according to the linkage rule; and sending the control instruction to the intelligent terminal to enable the intelligent terminal to execute an instruction task represented by the control instruction. According to the method, the intelligent gateway is constructed locally, the local area network between the intelligent gateway and the intelligent terminal is constructed, the intelligent gateway can download the linkage strategy used for controlling the intelligent terminal through the server side,and a control instruction is issued in the constructed local area network through the linkage rule to control the intelligent terminal. The linkage rule is placed in the local intelligent gateway, sothat the current situation that the intelligent terminal needs to depend on an external network to be unblocked under control is avoided.

The invention relates to an edge computing task unloading method for the industrial internet of things. The method comprises the following steps: constructing a three-layer edge computing node networkmodel; planning a point-to-point optimal path in the network through an SDN (Software Defined Network) controller; defining a time delay model from an equipment unloading task to each layer of edge computing nodes, and selecting N better edge computing nodes as candidate nodes for task unloading through a primary selection rule in the edge computing nodes of which the total time delay of the computing equipment task is less than the maximum tolerable time delay of the task; and taking the length and priority of the equipment task and the ratio of the transmission delay of task unloading to the maximum tolerable delay of the task as fuzzy logic input variables, and selecting an optimal node in the candidate nodes by adopting a fuzzy logic algorithm to unload the task. By means of the taskunloading method, the difference requirements of the equipment for data real-time performance and computing resources are guaranteed, and the transmission and processing time of equipment tasks is shortened.

The invention relates to a distributed uniting coordination control method of a large-scale electric automobile charging load, and belongs to the technical field of energy management. Client terminals determine and send the initial charging power to a coordination center according to basic information of an automobile. The coordination center calculates an optimized load curve and an optimized judging threshold value. A current load curve and a representation value of the current load curve are calculated according to the charging power of the client terminals. If the representation value is smaller than the judging threshold value, a stopping mutual order is sent, and the client terminals charge the electric automobile according to the charging power, and if not, a probability transfer matrix is calculated and is sent to the client terminals. The client terminals calculate a transfer matrix according to the probability transfer matrix, and the charging power is updated and sent to the coordination center. Interaction is executed repeatedly until the representation value is smaller than the judging threshold value. Staggering-peak charging can be effectively achieved, the calculating pressure of the coordination center is relieved, the charging demand privacy of a user is protected, the data communication resources between an upper layer and a lower layer are saved, and the distributed uniting coordination control method is suitable for orderly controlling large-scale electric automobiles.

The invention relates to a virtual point cloud three-dimensional target detection method based on supervised monocular depth estimation, and the method comprises the steps: 1, carrying out the measurement through employing a laser radar, and collecting the depth information of a scene; and 2, training a monocular depth estimation model by using the data set obtained in the step 1. The beneficial effects of the invention are that: the camera is directly used as a main sensing means, the application of expensive sensors such as a laser radar in a three-dimensional target detection system is avoided, and meanwhile, the problems of joint calibration and data synchronization in a multi-sensor sensing method are directly avoided, the sensor deployment cost is further reduced, the dependence on alaser radar is reduced, the cost of a three-dimensional target detection method is reduced, and the application of the technology in various fields is promoted. Besides, the algorithm model is deployed to the edge equipment through offline training and online prediction modes so that the intelligent level of the edge equipment can be enhanced while the equipment calculation pressure can be relieved.

The embodiment of the invention provides a target tracking method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining an ROI (regionof interest) in an Nth frame of image, wherein the ROI in the Nth frame of image comprises an image of a target, and N is a positive integer; and performing target tracking on the (N + 1) th frame ofimage by using the Nth frame of image and the ROI in the Nth frame of image. According to the embodiment of the invention, a region of interest ROI in an Nth image is acquired and the target trackingmode is utilized to perform target tracking on the ROI in the (N + 1) th frame of image, and since the target tracking mode is that a certain association exists between two adjacent frames of images,the stability and continuity of the ROI tracking result are ensured.

The invention discloses a method and apparatus for generating an eye image. The method comprises: an eye model for generating an eye image is obtained, wherein the eye model includes an eyeball portion for expressing the eyeball in the eye; picture pasting is carried out on the eyeball surface of the eyeball portion in the eye model, wherein the pasted picture is used for simulating the uneven texture of the retina surface in the eye; and rendering is carried out on the eye model after picture testing to obtain an eye image. Therefore, a technical problem that the eyes of the role are not vivid in an online game can be solved.

The invention provides an edge cloud resource scheduling method managed and controlled by an edge autonomous center. The method comprises the following steps: receiving a service request from a terminal; obtaining a state space of the edge cluster, the state space being used for representing a resource state of the edge cluster; inputting the state space of the edge cluster into a service request assignment model to obtain a state transition probability for assigning a service request action, wherein the service request assignment model comprises a deep reinforcement learning network; and determining a target cluster used for responding to the service request according to the state transition probability, the target cluster comprising the edge cluster or the cloud cluster, and the edge cluster comprising edge nodes. The invention further provides an edge cloud resource scheduling device controlled by the edge autonomous center and an edge cloud resource scheduling system controlled by the edge autonomous center.

The invention discloses a transformer substation monitoring system based on edge computing. The system comprises an external monitoring device, an internal monitoring device, an acquisition device, aprocessing device, a judgment device, a storage device, a first alarm device and a supervision platform, and is characterized in that the external monitoring device is arranged outside a transformer substation monitoring point; the internal monitoring device is arranged in a transformer substation monitoring point; the acquisition device, the processing device, the judgment device, the storage device and the first alarm device are connected with one another and correspondingly arranged near a substation monitoring point; the processing device is used for preprocessing, matching and identifyingthe image data acquired by the external monitoring device; and the storage device is in wireless connection with the supervision platform, and the supervision platform comprises a database, a secondalarm device and a display terminal. High-quality image resources are obtained, detection and recognition of various different parts and defects are achieved, various factors of the transformer substation are comprehensively monitored, an alarm is given in time and backed up to a supervision platform, and the computing pressure of a cloud processing center is relieved.

The invention discloses a neural network model construction method based on pruning and a related device. The method comprises the following steps: sequentially arranging the weight values of the preset neural network model; pruning the preset neural network model according to the number of the total branches of the neural network model and a pruning threshold determined by the weight sequence toobtain a first pruned neural network model, calculating an error function after retraining of the first pruned neural network model to perform Taylor expansion, and calculating to obtain an error change value; and performing cyclic pruning operation on the neural network model by judging the size relationship between the error change value and the error threshold to obtain an optimized target neural network model. The technical problems that an existing compression optimization method of the neural network model is difficult to design and limited in applicability, and the accuracy loss of thecompressed model is large are solved.

The invention provides an edge-cloud collaborative micro-service scheduling method and device and equipment, and aims at optimizing the service completion time and the resource utilization rate of an edge server by decoupling an application service program into a plurality of micro-services based on an improved particle swarm optimization algorithm. A micro-service scheduling strategy for realizing multi-objective optimization is finally obtained through continuous iteration. On one hand, service execution is close to a data source as far as possible through service scheduling, and the low-delay requirement of a user is met; and on the other hand, micro-services of part of non-delay sensitive applications are scheduled to a cloud server, so that the calculation pressure of an edge server is well relieved, and the problem that the edge server is overloaded and the service quality is influenced is avoided.

The invention discloses a federated learning method based on network unloading, which is used for improving the performance of the federated learning method while ensuring (improving) the privacy of terminal data. The method comprises the following steps: in a wireless communication system with a plurality of terminals and an edge server, jointly training an artificial intelligence network by using the terminals and the edge server; compared with a traditional federal learning working mode, carrying out the training of partial networks by unloading a terminal to an edge server; in the training process, selecting a terminal which is good in wireless communication network environment and small in calculation and transmission load to participate in training of the network model. According to the invention, the calculation load of the communication terminal is effectively reduced, and the communication load caused by a federated learning method is reduced.

The invention relates to the technical field of artificial intelligence, in particular to a model compression method and device, equipment and a storage medium. The model compression method comprises the following steps: acquiring a test image and a to-be-compressed model, wherein the to-be-compressed model comprises a plurality of cascaded feature extraction layers; inputting a test image into each feature extraction layer, and performing feature extraction on the test image through a filter to obtain a multi-channel feature map, wherein each feature channel in the multi-channel feature map corresponds to one sub-feature map; converting the sub-feature map into a visual feature map, and determining the importance of a feature channel corresponding to the sub-feature map based on a plurality of feature values in the visual feature map; determining a target pruning channel according to the importance degree of the feature channel; and performing pruning processing on the target pruning channel to obtain a compressed target model. According to the method, the model pruning accuracy can be effectively improved, accurate compression of the model is realized, the model calculation amount can be effectively reduced, and deployment is convenient.

The embodiment of the invention provides a binary-neural-network-based lightweight Web AR identification method and system. Thebinary-neural-network-based lightweight Web AR identification method comprises the steps: loading a target image through a mobile Web browser, preprocessing the image, and sending a target identification task request to an edge server; receiving a binary neural network model and a related executable script returned by the edge server, and performing binary neural network feedforward calculation, obtaining an image identification result, temporarily storing an output result of the sharing layer, judging whether the cross entropy of the image identification result meets a preset threshold; and if not, sending the output result of the sharing layer to the edge serverfor feedforward reasoning. According to the embodiment of the invention, the binary neural network is introduced to accelerate network reasoning, and reduce image recognition loading time delay and equipment energy consumption pressure; computing resources of the mobile terminal are fully utilized, thus effectively relieving computing pressure of the edge server, and a real-time solution for Web AR application is provided.

The invention discloses an electric power video image analysis system and method based on cloud edge collaboration, and the system comprises an electric power video image platform, a business processing system, an artificial intelligence cloud platform, and a plurality of edge analysis devices, and the edge analysis devices are used for collecting electric power video image data in an electric power production environment, and performing first-layer analysis processing on the electric power video image data, and the electric power video image platform is used for uploading the acquired electric power video image data to the cloud platform and reporting an analysis result to the service processing system; and the artificial intelligence cloud platform is used for carrying out second-layer analysis processing on the video image data. According to an existing cloud end centralized processing mode, computing power distribution is carried out on a cloud end and an edge end by calculating the cloud end analysis capability, the edge end analysis capability, the algorithm complexity and the algorithm calculation amount, and the problems of real-time performance of data processing, unreasonable occupation of analysis resources and the like are effectively solved.

The present disclosure relates to a method, an apparatus and a computer readable storage medium for processing an order request, and relates to the technical field of data processing. The method comprises the following steps of: obtaining an intelligent contract corresponding to a target item stored in a block chain; Obtaining a corresponding number of tokens according to the smart contract in response to the user booking the target item; In response to the user initiating an order request for the target item, determining whether the user has a sufficient number of tokens to place an order according to the smart contract; Validation of the user's order request, provided that the user has a sufficient number of tokens; All validated order requests are sent to the nodes in the block chain for validation to determine if the order was placed successfully. The technical scheme disclosed by the invention can avoid the waste of resources and improve the credibility of order processing.

The invention relates to the technical field of virtual power plant energy management and control optimization, in particular to a virtual power plant energy management and control optimization method and system based on end-side cloud architecture. Based on an operation mechanism of the end-side cloud architecture, distributed control architecture is adopted, the computing pressure of a cloud center is greatly relieved, the operation efficiency of the system is improved, meanwhile, the system further considers the regulation potential of a user-side flexible load, an energy optimization scheduling model of the system is further improved, and the utilization efficiency of energy is improved. According to the method and the system, a plurality of flexible resource clusters are constructed in different areas for different flexible resources, so that the scheduling control of a virtual power plant on each flexible resource can be enhanced.