104results about How to "Improve accuracy and effectiveness" patented technology

A system for producing corrective movement in a subject's teeth includes a dental aligner device having one or more through-holes and one or more connectors fixed to the subject's teeth. The connectors can snap into the through-holes when the dental aligner is worn on the subject's teeth.

A method for 3D reconstruction of the positions of a catheter as it is moved within a region of the human body, comprising: (a) ascertaining the 3D position of a point on a catheter for insertion in the body region; (b) acquiring a fixed angle, single-plane fluoroscopic image of the body region and of the catheter; (c) transferring the image data and catheter-point position to a computer; (d) determining the 2D image coordinates of the point on the catheter; (e) changing the length of catheter insertion by a measured amount; (f) thereafter acquiring a further single-plane fluoroscopic image of the body region and the catheter from the same angle, transferring the length change and additional image data to the computer, and determining the 2D image coordinates of the point on the catheter; (g) computing the 3D position of the catheter point; and (h) repeating steps e-g plural times. A 3D model is constructed by assembling the plural 3D positions of the point within the body region into a 3D model of the region.

A method for automatically determining the 3D position and orientation of a radio-opaque medical object in a living body using single-plane fluoroscopy comprising capturing a stream of digitized 2D images from a single-plane fluoroscope (10), detecting the image of the medical object in a subset of the digital 2D images, applying pixel-level geometric calculations to measure the medical-object image, applying conical projection and radial elongation corrections (31) to the image measurements, and calculating the 3D position and orientation of the medical object from the corrected 2D image measurements.

The invention discloses a method for data transmission. The method is used for an object carousel system consisting of a front end and a terminal and comprises the following steps of: requesting to acquire a directory tree structure of a webpage from the front end by using the terminal and establishing the directory tree structure of the webpage according to the data acquired by the request; determining the hierarchy which the page requested to be opened by a browser at present belongs to; requesting to acquire file data corresponding to the page requested to be opened at present from the front end; displaying the page and caching the data according to the data acquired by the request; requesting to acquire file data corresponding to a next-stage page of the page requested to be opened atpresent; caching the file data which the next-stage page corresponds to and is acquired by the request; and then waiting for the next request of the browser. The invention also discloses a system fordata transmission. The method and the system for data transmission has the advantages of reducing occupied storage space when the front end stores the webpage data, preventing the terminal from receiving repeating data repeatedly when requesting data, and improving the effective downloading speed of the terminal.

The invention discloses a longitudinal partition ultrasonic endoscope image real-time auxiliary diagnosis system and method based on deep learning. The system comprises an image acquisition module, acommunication module and a focus recognition module. The image acquisition module is used for acquiring a picture and a video which are obtained by performing longitudinal partition inspection by using ultrasonic endoscope equipment and decoding the video into a picture; the focus recognition module is used for receiving the picture of the image acquisition module, judging features corresponding to the image in time and obtaining a disease recognition and classification result; and the image acquisition module and the focus recognition module are connected and communicated through the communication module. According to the invention, an image acquisition module is firstly used for acquiring longitudinal ultrasonic endoscope pictures and videos; taking the mediastinal ultrasound endoscopy picture as a parameter, and calling a convolutional neural network model to identify and classify the disease; and finally, receiving a classification result of the convolutional neural network model,and assisting doctors in disease diagnosis. According to the method, the detection accuracy and effectiveness are improved, the misdiagnosis missed diagnosis occurrence probability is reduced, and thelabor cost is saved.

The invention discloses an intelligent control close-loop artificial pancreas system, comprising a continuous glucose monitoring device, an intelligent controller, and an insulin infusion device. The continuous glucose monitoring device is used to continuously monitor blood sugar concentration of a human body according to set interval. The intelligent controller is used to receive human body blood sugar monitoring data of the continuous glucose monitoring device in real time, and according to the real-time human body blood sugar monitoring data, by using a built-in fuzzy self-adaption proportion calculus control algorithm, calculates real-time insulin infusion device dosage. The insulin infusion device is used to receive the real-time insulin infusion device dosage of the intelligent controller, and perform real-time insulin on human bodies according to the real-time insulin infusion device dosage. The system can greatly reduce fluctuation range of human body blood sugar, and control blood glucose level in a set target interval, and improves accuracy and effectiveness of an insulin injection therapy, and makes blood glucose level of diabetics close to or reach standard of normal people, thereby greatly reducing various complications and improving living quality of the patients.

The present invention discloses an offline model improvement and selection method for junk short message classification. The method comprises the following steps: (1) carrying out feature selection and expansion, selecting features by using a feature selection method, constructing a feature word vector, and using a feature word vector model to express an original short message text; (2) carrying out optimization training and testing on the offline classification algorithm and improvement, carrying out improvement for junk short message classification on the offline classification algorithms, according to each offline classification algorithm and improvement, carrying out data preparation on the training set and the testing set obtained in the step (1), and using the training set to carry out optimization training and testing on each offline classification algorithm and improvement; and (3) based on the offline classification algorithm selection of the evaluation criteria, putting forward an evaluation index for the junk short message classification, analyzing the test result obtained in the step (2) by using the evaluation index, and selecting an optimal offline classification algorithm.

The invention discloses a method and device for detecting network attacks. The method comprises the following steps: receiving a network request from a sender; when determining that the type of the network request is an attack request by adopting an attack detection rule set, forwarding the network request to a receiver, and acquiring a network response corresponding to the network request; and detecting the network response by adopting the attack detection rule set, and selecting a processing mode of the network response according to a detection result. By adopting the method and device disclosed by the invention, the technical problem that a one-way network attack detection method adopted in related technologies is relatively low in accuracy can be solved.

The embodiment of the present invention discloses a matching degree calculation method for improving the accuracy of matching degree among users in a social network. The method comprises the following steps: outputting test information to the user under test, the test information corresponding to at least one social dimension for social matching, and the test information includes at least two items representing the attribute of the social dimension; obtaining the first option result of the test information fed back by the user to be tested; calculating the first social information of the user to be tested according to the first option result; determining the user matching degree between the user to be tested and the target user according to the first social information and the second social information of a target user, wherein the second social information is obtained by calculating the second option result of the test information fed back by the target user. The embodiments of the present invention also provide an device and user equipment capable of improving the accuracy of matching between users in order to realize mutual recommendation among users who may be truly matched.

The invention provides a method and a device for monitoring a wind turbine set. The method comprises the following steps: acquiring first data from practical sensors mounted on the wind turbine set; inputting the first data into a machine learning model; acquiring second data from virtual sensors of the wind turbine set; monitoring the wind turbine set according to the first data and the second data. As mapping between the practical sensors and the virtual sensors can be realized due to the machine learning model, omnidirectional state monitoring of the wind turbine set can be realized via the mounted practical sensors. Therefore, the accuracy and the effectiveness of monitoring can be improved.

The invention relates to a voice technology, provides a voice recognition method and solves the problem that misoperation is caused by easy misrecognition in the present voice recognition. The technical scheme can be summarized in that an activate word is added to ensure that voice equipment does not process when not receiving the activate word. The voice recognition method has the benefits that misoperation is effectively avoided and the voice recognition method is suitable for voice control equipment.

The embodiment of the invention discloses an arrival reminding method, device and terminal as well as a storage medium and belongs to the field of artificial intelligence. The method comprises the following steps: when in a vehicle, acquiring an ambient sound by virtue of a microphone; identifying the ambient sound; when a fact that the ambient sound contains a target alarm bell sound is identified, plusing one to number of travelled stops; and when the number of the travelled stops reaches a target number of stops, performing arrival reminding. By adopting the method provided by the embodiment of the invention, the ambient sound is acquired in real time, and whether the current ambient sound contains the target alarm bell sound or not is identified, so that the number of the travelled stops is updated when the fact that the ambient sound contains the target alarm bell sound is identified, and the arrival reminding is performed when the number of the travelled stops reaches the targetnumber of stops; and the alarm bell sound is used for giving out a caution to passengers, sound features are relatively obvious and can be easily identified, so that arrival reminding accuracy and effectiveness can be improved when the arrival reminding is performed on the basis of the alarm bell sound in the ambient sound.

The invention provides a vehicle behavior similarity calculation method based on bayonet data, Firstly, the data information of the moving object passing through each bayonet is collected and a database is established, and then bayonet data of vehicle A and B is retrieved to form the original trajectory of the two vehicle, the original trajectory is cleaned and cut to obtain a set about sub- trajectories of vehicle A and B, and then the sub-trajectories in the two sets are compared in pairs according to the LCSS algorithm, by calculating the longest common sub-sequence between any two sub-trajectories and the time difference between two vehicles passing through the adjacent bayonets in the longest common sub-sequence, the similarity between the two sub-trajectories is obtained. Finally, the multi-trajectory similarity is calculated by using all the similarity data, and the result is regarded as the behavioral similarity between the two workshops. The invention introduces a time influence factor in the calculation, the obtained result is more accurate and has wider application range, and the trajectory similarity result is also used for judging the similarity degree of the vehicle behavior.

The invention discloses a real-time auxiliary diagnosis system and method for endoscopic retrograde pancreatic duct radiography. The system comprises a client, a server and a database, An ERCP image is collected through a client side, the obtained ERCP image is input into a trained deep learning model to be recognized and classified, and corresponding image information (including duodenal papillaand papilla opening recognition and disease diagnosis) is obtained and fed back to the client side in time. A doctor can be helped to determine the duodenal papilla and the papilla opening position, and the intubation success rate is increased; In addition, doctors can be assisted to diagnose diseases, misdiagnosis and missed diagnosis are prevented, and the diagnosis accuracy and the working efficiency are improved.

The invention discloses a litchi picking flexible combing-shearing mechanism and a rolling combing-shearing device. The flexible combing-shearing mechanism comprises a tooth-shaped plate, branch separating components, a sorting component and cutting components, wherein the branching components comprise a plurality of tapered rubber blocks, all the tapered rubber blocks are arranged at the tail ends of all teeth on the tooth-shaped plate respectively, a sorting component is arranged in the middle of a tooth gap between any two adjacent teeth on the tooth-shaped plate, and a cutting component is arranged at a tooth root position of any two adjacent teeth on the tooth-shaped plate. The rolling combing-shearing device is provided with the litchi picking flexible combing-shearing mechanism. According to the litchi picking flexible combing-shearing mechanism and the rolling combing-shearing device, braches with fruits and branches without fruits of litchis can be separated, and effective shear can only be carried out on the branches with the fruits.

The invention relates to the technical field of semiconductors and discloses a testing structure for the matching degree of an electron beam flaw scanner. The testing structure comprises a first gray scale testing module, a second gray scale testing module and a third gray scale testing module which respectively comprise a metal connection through hole with a first gray scale, a metal connection through hole with a second gray scale and a metal connection through hole with a third gray scale; metal connection through hole structures with different scanning gray scales in a conventional semiconductor device are basically covered; and detection conditions of various through hole structures in wafers to be detected by the electron beam flaw scanner can be reflected sufficiently. The invention further provides a testing method for the matching degree of the electron beam flaw scanner based on the testing structure; different electron beam flaw scanners adopt the same picture snapping forms and the different wafers are selected; and first gray scale information, second gray scale information and third gray scale information of scanning regions are respectively collected and compared so as to realize the monitoring of the matching degree of each electron beam flaw scanner on a processing line.

The invention discloses a weight factor-introduced large-scale hydraulic machine intelligent diagnosis method, and relates to the technical field of hydraulic machine fault diagnosis. The method comprises the following steps of firstly, carrying out feature conversion on the original vibration signals by utilizing deep forest multi-granularity scanning, and then substituting the converted featuresinto a cascade forest of the depth forest which introduces weight factors for supervised step-by-step training; finally, obtaining a final classification result by using an extreme gradient lifting method (XGBoost) for result fusion, and completing intelligent diagnosis of a pump station unit. The deep forest model is utilized, so that the complex theoretical analysis and a complex parameter adjusting process of the deep neural network applied in the fault diagnosis field are avoided; and the weight factor is used for correcting the class vectors, so that accuracy and effectiveness of fault diagnosis are improved; by virtue of the extreme gradient lifting method (XGBoost) for result fusion, the intelligentization and the accuracy of the diagnosis are further improved, so that end-to-end intelligent fault diagnosis from the original monitoring data is realized.

The invention relates to an account security technology, and discloses an account transaction clustering method and system thereof. The method comprises the steps that a characteristic factor, which represents an account transaction characteristic, of each account in to-be-clustered transaction is extracted, wherein the characteristic factor comprises a first class of characteristic factor which represents the independent information of each account and a second class of characteristic factor which represents association information of accounts; for the first class of characteristic factor and the second class of characteristic factor, a measurement scale independent of each characteristic factor is used to carry out normalization; after normalization, Euclidean distance is used to calculate the similarity of clustering centers of various types of transactions; and according to the similarity, account transaction clustering is carried out. For the first class of characteristic factor and the second class of characteristic factor, the independent measurement scale is used to carry out normalization, and then the similarity is calculated for clustering. The problem of different characteristic factor variable dimensions is overcome. Small characteristic factors are prevented from being exaggerated.

The invention discloses an Asian female skin value scoring model method based on a depth convolution network, which comprises the following four steps: a sample data acquisition step; constructing a convolution network model; a method for constructing a stochastic forest model comprises the following steps of: marking the key point position of a human face based on a face recognition database technology, extracting the coordinates of the key point position of the human face, traversing a cycle to generate a group of calculation points composed of a part of the key points of the human face, andconstructing and optimizing the stochastic forest model; based on booster trees algorithm, the convolutional network model and stochastic forest model are fused to complete the final color score of the fusion model stage steps. At the same time, it combines the key information of Asian women's face, facial features, skin color and so on, which have strong discriminating ability, and collects theadvantages of convolution network and stochastic forest model, so as to improve the accuracy of skin value scoring prediction, the effect of the model is significant, and the mean value is closer to the real value.

The invention discloses an alarm event handling method. The alarm event handling method is applied to a cluster management system, data in an alarm database are analyzed with an association rule, and an alarm association rule base is formed; the method further comprises steps as follows: acquiring fault alarm information; processing the fault alarm information to acquire index information; comparing the index information with data in the alarm association rule base to determine a handling scheme corresponding to the index information. With the adoption of the method, the alarm information can be effectively processed, meanwhile, waste of manpower and material resources caused by trials of different handling schemes is reduced, and the efficiency for solving the cluster alarm problem is improved. The invention discloses an alarm event handling system.

The invention provides a man-in-the-loop satellite control system and a control method. The man-in-the-loop satellite control system comprises a satellite camera device, a ground receiving / transmitting device and a computer device, wherein the satellite camera device is carried on a satellite and used for obtaining a video image of a target ground object in real time; the ground receiving / transmitting device is arranged on the ground and used for receiving the video image in real time and transmitting the video image to the computer device; the computer device is arranged on the ground and comprises a control device; and the control device is used for controlling the ground receiving / transmitting device to transmit an operation instruction to the satellite according to the video image. According to the man-in-the-loop satellite control system provided by the invention, the satellite is adjusted according to the real-time video image obtained by the satellite so that the man-in-the-loop of the satellite is controlled; therefore, a user can flexibly control the satellite to rapidly change an observation area so that targets in a sensitive area is observed according to needs.

Embodiments of the present invention provide a method and apparatus for searching for an online advertisement resource. The method includes: setting at least one label for each of online advertisement resources; categorizing the at least one label according to a categorizing rule; categorizing a keyword inputted by a user when searching for an online advertisement resource according to the categorizing rule; and if a category to which the keyword belongs has a label, transmitting the online advertisement resource corresponding to the label to the user. The embodiments of the present invention increase success rate and accuracy for finding the online advertisement resource, and lower the requirement for searching conditions. Compared with the prior art, the problem that it is difficult to find a matching result and that a potential online advertisement resource may be missed is solved.

The invention discloses a household energy optimization scheduling management system and method. The system includes a data collection module, a data pre-processing module and an energy management module, wherein the data collection module is used for collecting power generation / power consumption data and environment data of a user, the data pre-processing module is used for detecting and correcting the data and performing power generation prediction of a photovoltaic system to obtain real-time state information data, and the energy management module is used for performing calculation in combination with the power consumption data of the user and the real-time state information data and performing power scheduling control of electrical appliances of the user based on the calculation result. The system is advantaged in that considering peak load control of the user, for scheduling management, the load is optimally scheduled based on the user's power consumption demand, the impact of improper use of electricity on the grid by the user is reduced, power consumption cost is reduced, negative impact on the user's electrical comfort is avoided, and user's economic efficiency and enthusiasm for participating in demand response are effectively improved.

The invention is suitable for the technical field of medical image processing, and provides a vascular tree extraction method, device and equipment and a storage medium. The method comprises the steps that a discontinuous vascular ridge line acts as an initialized vascular curve; in the iterative process of the vascular curve, the tentative points of a target curve are acquired by using the endpoints of the vascular curve after iteration, and the candidate points of the target curve are determined from the tentative points according to an inclined angle cosine threshold; the fusion points of the target curve are determined in the candidate points by calculating the minimum value of the inclined angle of a connection vector and the tangent vector of each candidate point; and fusion connection of the endpoints of the current vascular curve and one side curve of the fusion points of the target curve is performed so as to form a new vascular curve. The effectiveness and the accuracy of vascular tree extraction can be enhanced so as to have important clinical significance for the cardiovascular interventional operation.

The invention discloses a method of determining insulation characteristics of transformer oil. The method particularly comprises the following steps: 1) a transformer oil sample containing copper and iron metal particles and silica nonmetallic particles is prepared; 2) an insulation characteristic test is carried out on the transformer oil sample; 3) a fuzzy relation between particle pollutants in the transformer oil and the insulation characteristics of the transformer oil is built; 4) after the particle pollutants in the to-be-tested transformer oil are detected, the insulation characteristics of the transformer oil are analyzed and judged according to the fuzzy relation between the particle pollutants in the transformer oil and the insulation characteristics of the transformer oil. The fuzzy identification technology is used for determining the insulation characteristics of the transformer oil, determination can be realized without complicated test operation, human factor influences during the test process can be avoided, and the accuracy and the effectiveness of determining the insulation characteristics of the transformer oil are improved.

The invention discloses a method, a device and a system for managing service information, belonging to the technical field of electro numerical data processing. The invention aims at solving the problem that a service system fails to inquire record files in the prior art. The method for managing service information comprises the following steps of: obtaining record information of a record system in the presence of a client; storing the record information and service data associated with the record information; and inquiring the associated record information according to the service data. The invention ensures that service data processing personnel can hear a record corresponding to the service data while inquiring the service data record, and improves the accuracy and the validity for processing data by the service data processing personnel.

The embodiments of the present invention provide an information pushing method and a device, which relate to the technical field of computer applications. The information pushing method comprises: determining a comparing merchant for a target merchant; obtaining the first sales information and the second sales information with the former corresponding to the comparing merchant and latter corresponding to the target merchant; and pushing marketing suggestion information to the target merchant's device according to the first sales information and the second sales information. In the present invention, marketing suggestion information is proactively pushed to a target merchant to help it make or adjust marketing decisions, thereby eliminating the need for the merchants to collect relevant information necessary for making marketing decisions and making up the limitations of the access to information in prior art. Therefore, the marketing decision-making process is more streamlined and more intelligent. In addition, more targeted marketing suggestion information can be provided.

A system for recommending a service for use by a particular user comprises a processor and a computer-readable storage medium, wherein the computer-readable storage medium contains instructions for execution by the processor. The instructions cause the processor to perform the steps of selecting from a plurality of users similar users that are similar to the particular user, collecting services used by similar users, determining for collected services the level of satisfaction of similar users using the respective collected service, and recommending a service for use by the particular user from the collected services based on the determined levels of satisfaction of similar users.

The invention relates to a method and system for quickly positioning bus active imbalance based on multi-source data. The method comprises the following steps: (1) if the unbalanced bus has a voltage grade of 500 kV, the SCADA real-time active data of each branch connected by the bus are compared with the real-time active data of WAMS, if the error between the two kinds of data is larger than 5 MW, the line is determined to be a defective line, and the line is subjected to defect elimination processing; and (2) if the unbalanced bus has a voltage grade of 220 kV, SCADA measurement of each branch connected by the bus is compared with state estimation and the real-time active data of an opposite line, if the error between the two kinds of data is larger than 3 MW, the line is determined to be a defective line, and the line is subjected to defect elimination processing. The problems that the existing bus power imbalance processing method can not meet quick defect positioning and processing for bus active imbalance, and the safety, quality and economic operation of the power grid is influenced can be solved.

The invention discloses a drug molecule screening method and system. The method comprises the steps of: collecting drug molecule data related to a specific disease, carrying out the preprocessing of the data, and calculating an encoding vector and drug physicochemical properties; constructing and training an AI model based on a conditional variation auto-encoder, combining the encoding vector andthe molecular drug physicochemical property to serve as an input layer of the model, converting the encoding vector into a hidden layer encoding vector through an encoding layer of the model, generating a possible drug molecular structure through a decoding layer of the model, and in the model training process, minimizing a loss function of the model through a gradient descent algorithm, and continuously updating and iterating weight parameters of neural network structures of the encoding layer and the decoding layer; and generating potential drug molecules for curing the specific disease according to the trained model of the conditional variation auto-encoder. According to the drug molecule screening method and system, the drug physicochemical property data of the compound molecules are also utilized, the drug physicochemical property has great correlation with whether the compound can finally form a drug or not, and the druggability of the compound is improved.