55results about How to "Optimize output" patented technology

The invention relates to a cascaded neural network-based face key point detection method. The method includes the following steps that: a) a training-used face image set is established, and a key point position requiring detection is marked; b) a first-layer depth neural network is constructed and is used to train a face region estimation model; c) a second-layer depth neural network is constructed and is used to perform face key point preliminary detection; d) local region division is continued to be performed on an inner face region; e) a third-layer depth neural network is constructed for each local region respectively; f) the rotation angle of each local region is estimated; g) correction is performed according to the estimated rotation angles; h) a fourth-layer depth neural network is constructed for the correction data set of each local region; and i) any face image is given, and the above four-layer depth neural network model is adopted to perform key point detection, such that final face key point detection results can be obtained. With the cascaded neural network-based face key point detection method of the invention adopted, face key point detection can be improved, and especially the accuracy and real-time property of dense face key point detection.

A shift register and a shift register unit for diminishing clock coupling effect are introduced herein. Each stage shift register unit includes at least one pull-up driving module, a pull-up module, at least one pull-down module and a pull-down driving module. Before a waveform of either a first clock signal or a second clock signal employed by the pull-up module transits into a rising edge, the pull-down driving module employs a first periodic signal to turn on the pull-down module in advance for a specific period, and / or before the waveform of the first or second clock signal employed by the pull-up module transits into a falling edge, the pull-down driving module employs a second periodic signal to turn off the pull-down module in advance for a specific period. Accordingly, the pull-down module can gain a sufficient capability against the clock coupling effect so as to optimize the waveform outputted from the shift register unit.

A system and method for predicting future disturbance in MPC applications by segregating a transient part and a steady state value associated with the disturbance. A dynamic state space model that includes a variable disturbance prediction module can be utilized for analyzing a dynamic behavior of a physical process associated with a process model. The process model represents a dynamic behavior of the physical process being controlled and the dynamic state space model represents current deviations from the process model and future deviations over a predetermined prediction horizon. A predicted trajectory can be calculated as a response to the initial conditions estimated by a Kalman Filter for the process model extended by a disturbance model. The output of the dynamic state space model utilized for the disturbance prediction can be further provided as an estimated input to a MPC.

The invention discloses a medical knowledge base construction method based on a question-and-answer system, and relates to a medical knowledge base. The medical knowledge base construction technology based on the question-answering system mainly includes data processing, data analysis and an expert question-and-answer platform. Firstly, entity and correlation relations are extracted from clinical data by a hierarchical partitioning method at a data processing stage. Secondly, extracted results are analyzed by an association rule algorithm, entity correlation is mined, the mined entity correlation is matched with a medical dictionary, and correctly matched results are directly stored into the knowledge base. Finally, questions are automatically generated from entity correlation which cannot be confirmed by crowd-sourcing technology, the questions are submitted to an expert question-and-answer platform for answering the questions, and final results are selected from answers of experts by a majority voting algorithm and stored into the medical knowledge base.

The invention relates to a convolutional neural network road scene classification and road segmentation method. A dual-task joint structure model based on the convolutional neural network is established.and composed of an encoder and two decoders. Feature information sharing is achieved through end-to-end training.Classification of various road scenes such as urban roads, rural roads and highwaysis completed, road drivable areas in the scenes are segmented, and the dual goals of road scene classification and road area segmentation are achieved. Based on a convolutional neural network, scene classification and road extraction in road scene perception are combined. Real-time output is obtained through an end-to-end training mode, the functions of an intelligent driving assistance system areimproved, and an effective and reliable driving assistance function is provided. A mode of combining a full convolutional network and a conditional random field is adopted for a part extracted from adrivable area in the model, a high-resolution image optimization output result can be generated, and a relatively accurate segmentation effect is achieved.

The invention discloses a question answering method and device, computer equipment and a storage medium. The question answering method comprises the steps: obtaining input information of a user; recognizing named entities in the input information, and linking the named entities to candidate entities, corresponding to the named entities, in the Chinese knowledge graph, to form entity pairs, whereinthe entity pairs comprise the named entities and the candidate entities; matching candidate relations of candidate entities in the Chinese knowledge graph through a relation model; according to the entity pairs and the candidate relations, forming candidate triples, wherein the candidate triples comprise named entities, candidate entities and candidate relations; obtaining a sorting result corresponding to each candidate triple based on the learning sorting model; and querying the Chinese knowledge graph according to a sorting result to obtain an answer to the input information. According tothe question answering method, external resources can be effectively utilized, and a large amount of context information can be provided through text mining, and based on the learning sorting model, good answers can be obtained when question and answer corpus data are few.

A liquid ejection control apparatus which controls a liquid ejection mechanism having a plurality of nozzle groups made up of a plurality of nozzles for ejecting liquid, comprising: a usage ratio determination unit which determines a usage ratio between the plurality of nozzle groups when performing liquid ejection; a correction data generation unit which generates correction data corresponding to the determined usage ratio, by using respective correction data for correcting density of liquid ejection results corresponding respectively to a plurality of usage ratios between the nozzle groups; an image data correction unit which corrects prescribed image data on the basis of the generated correction data; and an ejection execution control unit which executes liquid ejection by driving the respective nozzle groups on the basis of the corrected image data and the determined usage ratio.

The invention discloses a distributed public verifiable random number generation method. A random number is generated through the cooperation of a participant, and the ciphertext is made public in the process of generating the random number so as to be verified. Unpredictability requires that the random number is private before being issued, but the public verifiability requires to discloses the information for verification, one of the main contributions of the patent is to solve the contradiction between two characters. Furthermore, the second contribution of the patent is that the trusted third party can be eliminated in the verifiable random number generation process; on one hand, the method can prevent single participant from knowing the random number result in advance, the security of the protocol is improved; on the other hand, the method is conductive to solving sing point of failure problem, and the robustness of the protocol is improved.

Disclosed is a road traffic state estimation method based on microwave radar data. The method includes the steps that firstly, the maximum traffic capacity of road segments is acquired through statistics, and saturability is estimated; secondly, the mean value and standard difference of cross section speeds of a traffic flow lasting five minutes on each road segment are calculated through statistics; thirdly, an individualized subordinating degree function is automatically generated for each road segment; fourthly, a smoothness constant, a slow moving constant and a congestion constant are given respectively based on fuzzy rules, the strength of each rule is calculated for each set of input data (speed and saturability), three output maximum rule strengths are sequentially recorded as w1, w2 and w3, and then an output value Z is worked out through the weighted average method; fifthly, it is judged that traffic is smooth if the final value Z is the most approximate to the smoothness constant, it is judged that cars need to run slowly if the final value Z is the most approximate to the slow moving constant, and it is judged that the traffic is congested if the final value Z is the most approximate to the congestion constant. A model of the method has good continuity, high adaptive ability and good practicability.

The invention provides a suction type SF6 gas leakage monitoring device and method. The device comprises an air outlet (1), an air inlet (2), an infrared light supply (3), a window piece I (4), a window piece II (5), a reference spectral filter (6), a measuring spectral filter (7), a dual channel infrared detector (8), a temperature probe(9), a signal conditioning module (10), an A / D (Analog / Digital) conversion module (11), a data processing module (12), a light source driving module (13), a device body (14), an end cover I (15) and an end cover II (16). The invention provides the suction type SF6 gas leakage monitoring device and method have the advantages that the suction type SF6 gas leakage monitoring device has a digital output and thermal compensation function and is simple and reliable, and the sensitivity is high. The method and the device can carry out ration monitoring on SF6 leaking so as to externally output a high-precision SF6 leaking concentration value, and the method and the device have the advantages that the monitoring accuracy is high, the method and the device are stable and reliable, the structure is compact, the volume is small and exquisite and the like.

The invention discloses an audio detection method for road tunnel traffic events. Audio data collected in a road tunnel is divided into frames, second-level long-term and short-term memory network models are utilized to improve the identification rate and robustness, specifically, 1), the first long-term and short-term memory network model A is utilized to optimize extracted audio features, and the attention mechanism is further introduced to improve robustness of the features; and 2), the second long-term and short-term memory network model B is utilized to weight the optimized and refined features, the output result is further optimized, so road tunnel traffic event detection has high timeliness and accuracy, and the method has good application prospects.

The invention provides a solder paste printing performance influence factor analysis method based on SMT big data, and solves the problems of incomplete analysis and low precision in solder paste printing performance influence factor analysis. The method comprises the following implementation steps: collecting solder paste printing parameters and performance indexes to construct a solder paste printing data set; Processing the data by using a Mahalanobis distance and a null value; Calculating correlation coefficients among the features, and filtering redundant features; Dividing training and testing sample sets; Randomly extracting a part of features and constructing a random forest model; Setting a model termination condition; Estimating a characteristic importance degree score accordingto the model mean square error increment, and performing sorting; and determining a subset of key impact factors. According to the method, the key influence factors of the SMT solder paste printing performance are mined through random forest feature selection in combination with the big data processing technology, the correlation between the performance indexes and the printing parameters is determined, the solder paste printing performance is optimized, and the circuit board printing quality is improved. The method is used for process optimization and solder paste printing performance improvement in the surface mounting technology solder paste printing process.

A liquid ejection control apparatus which controls a liquid ejection mechanism having a plurality of nozzle groups made up of a plurality of nozzles, comprising: a correction data acquisition unit which acquires correction data for each nozzle group; an image data separation unit which inputs image data comprising a plurality of pixels and separates the image data into respective split image data, each comprising pixels which are to be formed for liquid ejection by the respective nozzle groups; a split image data correction unit which corrects the respective split image data, on the basis of the correction data for the nozzle groups to which the split image data respectively correspond; and an ejection execution control unit which executes liquid ejection by driving the respective corresponding nozzle groups, on the basis of the corrected split image data.

The invention discloses a CO2 gas concentration monitoring device and method with the automatic temperature and air pressure compensation. The CO2 gas concentration can be accurately detected in a relatively wide temperature range and a relatively wide air pressure range, and the device and the method are applied to monitoring the CO2 gas concentration in a low-temperature low-pressure environment. The CO2 gas concentration is monitored by adopting a dual-wavelength infrared monitoring principle and a suction structure, and the device has the advantages of adaptation to a low-temperature low-pressure environment, high precision, high reliability and the like. The interference caused by factors such as light source instability is eliminated by adopting a dual-channel infrared detector, the absorption length of infrared light is increased by adopting a reflector, the environment suitability of the device is enhanced by adopting a heating element and a heat preservation material, the receiving light intensity is enhanced and the automatic temperature and air pressure compensation and correction is performed by adopting a conical optical collector, and the CO2 gas monitoring precision is effectively improved.

There are provided a display apparatus, a display system and a display method. The display apparatus includes: a first input unit through which a first image signal having a first type is inputted; an image processing unit which processes the first image signal for display; a converting unit which converts the type of the first image signal into a second type different from the first type; and an output unit which outputs the first image signal converted into the second type to a next connected display apparatus.

The invention discloses a mid-point potential balance control system for a T-type three-level inverter based on neuron PI adaptive control. A voltage outer loop employs a neuron PI adaptive controller, and a current inner loop employs a traditional PI controller. The input of the voltage outer loop at the direct current side is a deviation value of two capacitor voltages to take as input of the PIadaptive controller, output of the voltage outer loop through the neuron PI adaptive controller is taken as input of the current inner loop, the difference of the input of the current inner loop andthe output of the voltage outer loop is subjected to the neuron PI adaptive controller to compare with a triangular wave to obtain mid-point potential balance control system for the T-type three-levelinverter with a balance circuit of complementary two-path PWM signal control. The voltage outer loop employs adaptive controller based on the neuron PI parameters to perform regulation of the parameters of the controller on line in real time. Compared to a traditional PI controller, the adaptive ability is high, the response speed is fast, and the output result of the voltage outer loop controller is better.

The invention relates to a knowledge graph optimization method based on a large language model. The method is suitable for the field of knowledge graph technology and natural language processing technology. According to the technical scheme, the knowledge graph optimization method based on the large language model is characterized by comprising the steps that entities in a knowledge graph to be optimized are obtained, clustering is carried out based on semantic feature vectors of the entities, and subdivided entity classifications of all the entities are obtained; any two entities belonging to different subdivided entity classifications are subjected to relation description through a large language model, and relation language description between different classification entities is generated; updating the knowledge graph to be optimized on the basis of the generated relational language description among the different classification entities; performing deep reasoning on the updated knowledge graph to further mine and discover implicit relationships among entities in the knowledge graph; and updating the knowledge graph again based on the implicit relationship among the entities to obtain an optimized knowledge graph.

The invention provides a training method of a weak light image enhancement model and a weak light image enhancement method, and the training method of the weak light image enhancement model comprises the steps: inputting a normal light image and a weak light image into an initial decomposition module, and obtaining a normal light reflection image, a normal light illumination image, a weak light reflection image and a weak light illumination image; inputting the weak light reflection image and the weak light illumination image into an initial color recovery module to obtain a color recovery image; determining a first loss function value, a second loss function value and a third loss function value; training an initial decomposition module according to the first loss function value and the second loss function value, and training an initial color recovery module according to the third loss function value. According to the invention, the trained decomposition module and color recovery module can obtain a better output result, the detail features and color information of the image are gradually recovered, and the weak light enhanced image generated by the weak light image enhancement model obtained through training can obtain a better result in image detail and color recovery.

The invention belongs to the technical field of sonar image processing, and relates to a target identification method based on sonar image segmentation and a computing device.The target identification method based on sonar image segmentation comprises the following steps that 1) a data set of existing sonar images is acquired; 2) building a target identification model composed of an encoder and a decoder; 3) inputting the data set into the target recognition model from the encoder, and performing initial training on the target recognition model; 4) performing sample augmentation on categories with a small number of samples in the data set to obtain an augmented feature domain data set; 5) using the augmented feature domain data set to carry out fine tuning training on the target identification model; and 6) after the target identification model is finely adjusted and trained, inputting a newly obtained sonar image into the target identification model to obtain a segmentation image of the target. The method can achieve the accurate recognition of a target.

The invention discloses a CO2 gas concentration monitoring device and method with the automatic temperature and air pressure compensation. The CO2 gas concentration can be accurately detected in a relatively wide temperature range and a relatively wide air pressure range, and the device and the method are applied to monitoring the CO2 gas concentration in a low-temperature low-pressure environment. The CO2 gas concentration is monitored by adopting a dual-wavelength infrared monitoring principle and a suction structure, and the device has the advantages of adaptation to a low-temperature low-pressure environment, high precision, high reliability and the like. The interference caused by factors such as light source instability is eliminated by adopting a dual-channel infrared detector, the absorption length of infrared light is increased by adopting a reflector, the environment suitability of the device is enhanced by adopting a heating element and a heat preservation material, the receiving light intensity is enhanced and the automatic temperature and air pressure compensation and correction is performed by adopting a conical optical collector, and the CO2 gas monitoring precision is effectively improved.

The invention provides an underwater image restoration method based on wavelength compensation. The method comprises the following steps: performing nine-digit-based subdivision hierarchical search onan original image to estimate atmospheric light spots; and then calculating a transmission image by using a Haze-lines method in different Jerlov water types, compensating atmospheric light by usingdifferent attenuation coefficients, and calculating a distance image and a depth image to obtain a restored image. In order to further improve the contrast of the underwater image, a contrast-limitedadaptive histogram is used to enhance the restoration result. A final output image is determined according to a selection rule based on gray world hypothesis and information entropy. According to theinvention, the atmospheric light is obtained by using a nine-digit method, so that the influence of scenery and light spots on estimation of the atmospheric light is effectively avoided, and the atmospheric light is accurately estimated. The distance map and the depth map are estimated by using different water types, the underwater image degradation problem is effectively solved, and finally, thecontrast and the brightness of the image can be effectively enhanced by using the contrast-limited adaptive histogram for enhancement.

The invention discloses a lung lobe segmentation optimization method and system based on lung segmentation, and relates to the field of medical images. The method comprises the steps that 1, a first output result of a first lung automatic segmentation algorithm model and a second output result of a second lung lobe automatic segmentation algorithm model are obtained, and the precision of the first lung automatic segmentation algorithm model is larger than that of the second lung lobe automatic segmentation algorithm model; the method also includes: step 2, carrying out point multiplication on the first output result and the second output result to obtain an independent mask of each lung lobe; step 3, respectively acquiring the maximum connected domain matrix of the independent mask of each lung lobe; step 4, performing superposition processing on each maximum connected domain matrix to obtain first optimization data; and step 5, performing calculation and elimination processing on the first optimization data to obtain a final optimization result. The invention can achieve the effect of improving the prediction precision.

The invention discloses an air-ground collaborative edge calculation method and system. The air-ground collaborative edge calculation method specifically comprises the following steps: setting initialization information; in response to the completion of the initialization parameters, performing distribution of calculation tasks; in response to completion of allocation of the calculation task, performing allocation of communication resources; in response to completion of communication resource allocation, performing allocation of computing resources; determining the trajectory of the unmanned aerial vehicle in response to the completion of the allocation of the computing resources; in response to completion of trajectory determination of the unmanned aerial vehicle, determining a duration value; in response to the determination of the completion duration value, updating the calculation efficiency; judging whether the calculation efficiency meets preset convergence precision or not; and if yes, outputting the allocation result of the communication resources, the allocation result of the computing resources, the track of the unmanned aerial vehicle and the duration value as optimal results. According to the invention, the purposes of system resource allocation including communication resources and computing resources and combined unmanned aerial vehicle flight path design are achieved.

The invention provides an aspect-level sentiment analysis method and device based on a graph convolutional network and a medium, and the method comprises the steps: inputting a to-be-processed text into a trained aspect-level sentiment analysis model for analysis, and obtaining a sentiment polarity classification result corresponding to the to-be-processed text; the aspect-level sentiment analysis model comprises an input module, an attention module, a graph convolution module for masking and aggregating features output by a previous level, a self-adaptive weight module for weighting the aggregated features, and an output module; compared with the prior art, the method has the advantages that the graph convolution module is arranged in the aspect-level sentiment analysis model, and information of context nodes is aggregated after mask processing is performed on extracted features, so that more attention can be paid to key parts contributing to sentiment analysis; and the adaptive weight module is set, and the output of the graph convolution module is subjected to attention weight distribution, so that the network is better endowed with adaptive capability, the output result is optimized, and the sentiment classification accuracy is improved.

The invention provides a suction type SF6 gas leakage monitoring device and method. The device comprises an air outlet (1), an air inlet (2), an infrared light supply (3), a window piece I (4), a window piece II (5), a reference spectral filter (6), a measuring spectral filter (7), a dual channel infrared detector (8), a temperature probe(9), a signal conditioning module (10), an A / D (Analog / Digital) conversion module (11), a data processing module (12), a light source driving module (13), a device body (14), an end cover I (15) and an end cover II (16). The invention provides the suction type SF6 gas leakage monitoring device and method have the advantages that the suction type SF6 gas leakage monitoring device has a digital output and thermal compensation function and is simple and reliable, and the sensitivity is high. The method and the device can carry out ration monitoring on SF6 leaking so as to externally output a high-precision SF6 leaking concentration value, and the method and the device have the advantages that the monitoring accuracy is high, the method and the device are stable and reliable, the structure is compact, the volume is small and exquisite and the like.

The invention provides a method for realizing a CLI command interface on a single chip microcomputer. The method comprises the following steps: firstly, defining a global variable CLI _ SCAN and calling a performance function information recording structural body USRCMD _ PARA _ STRUCT; receiving login information of the user; identifying an interrupted reading character input from the keyboard and outputting processing information; reading a whole command output result character according to the output processing information; judging the type of the action executed in the next step accordingto the result symbol and returning; responding to the action, executing a performance function matched with the action, and outputting a result; the command is added to a historical command. Based onthe method, the invention further provides a system for realizing the CLI command interface on the single-chip microcomputer, and the auxiliary functions which can be realized by the method comprise user permission setting and TAB key or-key supporting. Completing the commands, checking historical commands and the like. A man-machine interaction interface of a command line interface is realized, so that the configuration, management and test of the embedded single-chip microcomputer are more efficient, and scripts are automated.

The invention discloses an optical satellite image pixel level change detection method. The method comprises the following steps: S1, data acquisition and preprocessing; s2, designing a convolutional neural network model phi: the convolutional neural network model phi comprises a multi-scale feature fusion strategy psi and a multi-scale output structure Omega, and the multi-scale feature fusion strategy psi extracts multi-scale features through four types of nonlinear activation operations to output a multi-scale feature map, the multi-scale output structure Omega classifies the feature map through the activation layer and outputs a detection result; s3, model training; s4, optimizing the model; and S5, model prediction: fixing the convolutional neural network phi, and performing change detection on each to-be-tested image to obtain a change detection result graph CBI. The method effectively resists the influence caused by the registration error between the dual-time-phase satellite remote sensing images and the projection difference of the remote sensing images, improves the precision of pixel-level change detection, and improves the processing efficiency of the pixel-level change detection.

The invention discloses a data open output method which is executed in a computing device, an M * N first sparse matrix is stored in the computing equipment, an embedded model is trained according tothe matrix, the model is suitable for decomposing the input sparse matrix into at least three sub-matrixes, and two sub-matrixes are reconstructed into an output matrix. The method comprises the stepsof obtaining L user identifiers of to-be-output data, extracting original data of N user features corresponding to each user identifier, and generating an L * N second sparse matrix; splicing the second sparse matrix into the first sparse matrix to obtain a third sparse matrix; updating the first sparse matrix into a third sparse matrix, and retraining the embedded model by adopting the third sparse matrix to obtain a new embedded model; and learning the second sparse matrix by adopting the new embedded model to obtain a corresponding output matrix, and carrying out open output on the outputmatrix. The invention further discloses a corresponding data open output device and computing equipment.

The invention discloses an ultra-precision lathe motion control method based on fuzzy PID self-tuning calculation, and relates to the technical field of ultra-precision cutting processing. The methodmainly comprises the following steps: acquiring preset position information and actual position information of a linear motion guide rail of a lathe as a controlled object; and calculating a control output quantity by adopting a fuzzy PID control algorithm according to the deviation value between the preset position and the actual position and the change rate of the deviation value, and inputtingthe control output quantity into a control system to control the linear motion guide rail to move. A discourse domain parameter self-tuning method is adopted in fuzzy PID control and introduced into the fuzzy PID control, autonomous learning can be achieved, the static and dynamic performance of the ultra-precision lathe guide rail control system is improved, and then the purpose of improving theultra-precision cutting machining precision is achieved.

The invention discloses a multi-parameter uncertainty analysis method based on a value set visualization technology. The multi-parameter uncertainty analysis method comprises the following steps: 1, constructing a value set function in a complex form based on a known analytical expression; 2, quickly drawing a corresponding value set graph and a parameter action arrow by utilizing a value set function; 3, selecting key parameters, and drawing a discrete value set and a parameter action arrow; and 4, completing multi-parameter uncertainty problem analysis according to the coverage range and themoving trend of the value set and the action arrows of the parameters. When solving the problem of multiple uncertain parameters, the multi-parameter uncertainty analysis method based on a value setvisualization technology, on the basis of a value set mapping mathematical theory and in combination with a value set rapid drawing skill, provides a visual expression image to clearly present the effect of each parameter on an output result, and corresponding analysis results can rapidly and visually reveal different influences of multiple parameters on the output result, and then optimal parameter configuration is found, and the output result is improved.