697results about How to "Proof of validity" patented technology

The invention discloses a graph-based vision SLAM (simultaneous localization and mapping) method. According to the method, the matching relation between an image and visual feature can be obtained based on the natural feature probability vector representation of the image, and the relative pose between two interframes can be calculated by utilizing the space geometry relation of images. Data association of visual odometry is obtained by utilizing the corresponding relation of continuous images, so that all constraints in an image sequence can be obtained. The camera relative pose is taken as a node in a map, the space constrained relation of image interframes is taken as an edge, so that an estimated track map based on the camera relative pose is constructed. Finally, a maximum likelihood method is employed for optimizing the map, and optimized pose estimation is obtained through a random gradient descent method. Related experiments are performed in the laboratory environment based on the provided method, also the moving track of a robot is displayed, and the validity of the algorithm is confirmed.

The invention discloses a big data based wireless real-time position positioning method. The big data based wireless real-time position positioning method comprises the following steps of gridding a positioned detection area, enabling a plurality of wireless routers which are arranged inside the detection area to transmit wireless signals, enabling a wireless signal receiver to collect wireless signal information of the wireless routers for a plurality of times in every grid of the area after gridding and forming collected wireless signal information into a big database wireless signal fingerprint database; performing signal preprocessing; performing parameter training; training DNN(Deep Neural Networks); performing signal character extraction and character classification based on the trained DNN; performing position estimation based on a HMM (Hidden Markov Model). The big data based wireless real-time position positioning method has the advantages of improving the accuracy of a positioning result under the condition that the real-time positioning speed is not influenced, successfully integrating positioning problems into the big data background and improving the performance of the real-time positioning system by effectively using advantages of the big data.

The invention provides an AUV intelligent collision prevention device and a collision prevention method. The state information of the AUV collected by the sensor is passed to a dynamic control machine by a serial port. The dynamic control machine transmits the state information to a mission management machine by a network; a multi-beam front looking-sonar collects barrier information which is converted to a digital signal describing local environment; the mission management machine receives the digital signal of the local environment; an AUV barrier decision-making system determines the position of the barrier, establishes a local environment model which is passed on to a collision prevention planning system; the collision prevention system figures out course, speed, and depth of the AUV by collision prevention algorithm and passes on the three instructions to the dynamic controller by the network; the dynamic control machine controls resolving by movement, uses a control voltage to drive an executing agency, and adjusts course, speed and depth of the AUV according to the steering instructions, thus realizing AUV collision prevention. The invention has the advantages that collision prevention is realized in real time without needing prior knowledge and reliability and validity of the proposal of the invention are proved by a lake testing.

The present invention proposes an unsupervised hash image retrieval system and method based on the convolution neural network. The system and method based on the data enhancement technology, proposes an efficient unsupervised hash model used for the field of fast image retrieval by using the existing hash algorithm structure. Through the data enhancement method, a triad training sample is constructed for the unlabeled data; and the triad loss function, the minimum quantization error loss function and the maximum entropy loss function drive the network to make full use of the information of each picture, so that a series of parameters with more expressive ability is learned to improve the accuracy of fast picture retrieval. The method provided by the present invention is a hash fast image retrieval method capable of using the unlabeled data learning network, the data enhancement is used to construct the triad training sample with more powerful expression ability to train the network, and the accuracy of fast picture retrieval is significantly improved.

The invention presents a design method of a four-rotor fault-tolerant controller based on a nonlinear observer. The method includes the following steps: establishing a dynamic model of a four-rotor unmanned aerial vehicle (UAV), and dividing the UAV system into an attitude subsystem and a position subsystem; for the attitude subsystem, establishing a fault model, designing a nonlinear fault observer, carrying out real-time detection and online estimation on an unknown actuator fault, and designing a fault-tolerant controller based on a fast non-singular terminal sliding mode method by using the obtained fault estimation information; and for the position subsystem, inversely solving the control rate and the desired attitude angle of the position subsystem by combining a backstepping methodand a sliding mode method and designing an intermediate virtual quantity. A four-rotor UAV is enabled to track the desired position and yaw angle even when the actuator fails, so as to ensure the stability of the pitch angle and roll angle.

The invention discloses a method for constructing a three-dimensional (3D) map by a mobile robot in an unknown environment, and proposes a multi-exploration strategy method combining information gainguided local exploration strategy and global boundary exploration strategy for a problem of constructing a map of an indoor unknown 3D environment. The method comprises the following steps: firstly, constructing an autonomous exploration and 3D map construction system framework of the mobile robot, which comprises a map construction module and an exploration decision planning control module; secondly, establishing an information gain guided decision model based on a Shannon information theory and designing an information gain objective function calculation applied to multi-step exploration action evaluation; and finally, constructing a local exploration strategy by using an offline local motion trajectory, realizing a multi-strategy exploration method by combining with the information gainguided global boundary exploration strategy, and dynamically switching the two strategies during exploration according to real-time construction situations of the 3D map. Experimental results show that the method provided by the invention ensures integrity of map construction while rapidly constructing the 3D map.

The invention relates to a combined judging strategy-based visual target tracking method. The method comprises the following steps of: (1) aiming at a target size change problem, establishing a self-adaptive scale so as to automatically adjust the size of a tracking frame; (2) aiming at a problem that a position of a target cannot be continuously determined after the target is sheltered, predicting a current observed quantity by adoption of Kalman filtering, and after the target reappears, continuously tracking the target by using a KCF algorithm; and (3) aiming at a problem of tracking failure caused by rapid movement of the target. According to the method, a movement speed of the target is calculated through detecting a position movement distance of the target in adjacent frames, and a detection area size extension coefficient is self-adaptively adjusted, so that relatively high precision can be obtained under different environments, the robustness of the whole tracking system is effectively improved, a relatively high calculation speed is kept, and a high engineering practical value is provided.

The present invention belongs to the field of computer science and technology, in particular, it relates to a new characteristic vector weight method for text classification. Said invention provides a new weighting method (TF*IWF*DBV). It is characterized by that in TF*IWF method n-th root of DBV and TF can be introduced, the tests show that the adoption of said new method can raise FI measure by 11.8%, and can fully show its effectiveness.

The invention relates to an RBF neural network adaptive dynamic surface control method based on a flexible joint of a robot arm, belonging to the technical field of artificial intelligence and intelligent control. The invention aims at modeling the flexible joint existing in the robot arm, and designs a controller by combining the RBF neural network and the dynamic surface technology with an adaptive control method. The RBF neural network is used to compensate the uncertainties of system parameters, an adaptive law is used to adjust the weights of the neural network to improve the approximation ability of the RBF neural network to nonlinear functions and eliminate the need for precise dynamic model of the robot arm. A position tracking control algorithm suitable for the light robot arm with flexible joints is studied. At last, the design controller is verified by a simulation example, and the invention can ensure that the joint can effectively track a given signal under the condition that the output torque of the robot arm is limited, the tracking error is unconstrained in a certain range, and all signals are semi-globally bound.

The invention discloses an output optimal operation method of a thermal power unit in the environment of ensuring electricity price bidding marketization, and the method comprises the following steps: establishing a manual input cost parameter table and a unit production real-time system database; establishing a generating fuel cost curve of the unit; calculating the start-stop cost and the fixed cost of the unit; calculating the environment cost of the unit, internalizing the environment cost and calculating the environment cost directly according to harmful gas emission and a discount standard; analyzing bidding risk; establishing an objective function of a mathematical model of economic operation of a single unit based on profit maximization; guaranteeing constraint conditions of optimal operation of the unit; and establishing the best output model table of the unit at various periods. The method considers the factors of environmental protection cost and bidding risk comprehensively, thus reducing harmful gas emission of the power plant when a generation company does not add environmental protection equipment, and moreover, reasonable analysis of the bidding risk can realize price bidding of low risk and high profit.

The invention relates to a based method for improving seismic data resolution based on a seismographic record varitron wave, which is characterized by comprising the steps of: 1, constructing a molecule-Gabor window, adaptively dividing an instable seismographic record into a plurality of stable segments, wherein each segment has an approximately unchangeable equivalent wavelet which is easy to extract from the segment; 2, converting the instable seismographic record into a molecule-Gabor domain by using a molecule-Gabor frame generated by the molecule window obtained by the construction in the step 1; 3, carrying out frequency extension and energy compensation processing on molecule-Gabor coefficients corresponding to the seismographic record segments in each molecule-Gabor window; and 4, inversely transforming the processed molecule-Gabor coefficients into a time domain to obtain the seismographic record after the resolution ratio is increased. In the method, on the basis of the modern quasidifferential operator principle, an adaptive time-frequency analyzing method is used as a tool, and the processed seismic data has high resolution rate and relatively keeps amplitude characteristics.

The invention discloses a dynamic economic dispatching method for a power system considering photovoltaic power generation. The method comprises the following steps: historical meteorological data are acquired for statistical analysis, and solar irradiance and temperature prediction curves in the next 24 hours are obtained; statistics on historical load data is carried out, and a power system load curve in the next 24 hours is predicted and obtained; a photovoltaic power station output model is built, and a photovoltaic output prediction curve in the next 24 hours is obtained; according to a photovoltaic output probability distribution model, the photovoltaic output probability distribution in the next 24 hours and photovoltaic output values in different scenes are obtained; a dynamic economic dispatching optimization model for the power system with the photovoltaic power station is built, and a risk reserve curve is obtained; the risk reserve capacity is calculated, and a risk capacity curve is obtained; a basic UC optimization problem is solved, and a unit commitment scheme is corrected; and a dispatching scheme with the minimum target function found out during an operating cost expected value process is solved as the optimal economic dispatching scheme. The operation risk controllability and the operation economy of the power system are ensured.

The invention provides a method of reconstructing a power distribution network containing a distributed power supply. A network topology adjusting module considering various load modes is established according to actual operation conditions of the power distribution network and has good practical value. Meanwhile, topological adjusting is performed by means of a binary particle swarm optimization algorithm according to features of network topology adjustment; a dynamic inertial weight adjusting mode is provided; better coordination with global convergence speed and local convergence precision is provided; optimizing effect is better; the topological adjustment results and analysis of an IEEE-33 node typical distribution system show that the method is effective and practical.

The invention discloses an acoustic model based on a depth bidirectional long short-term memory (DBLSTM) recurrent neural network (RNN). The DBLSTM network is mainly divided into three parts: in a full-connection part of the DBLSTM, Maxout nerve cells are used for replacing original Sigmoid nerve cells to solve the problems of gradient disappearance and explosion commonly appearing in RNN; simultaneously, Dropout regularization training algorithm is used for preventing the neural network from causing overfitting in the training process. In the multi-layer BLSTM part, in order to adapt to the bidirectional dependency of DBLSTM on each time step length, a Context-sensitive-chunk Back-propagation through time (CSC-BPTT) algorithm is provided for training the network. A selection link layer is adopted on the back of the multi-layer BLSTM part, and is used for carrying out conversion on the output of the DBLSTM to obtain the input of the full-connection part. According to the acoustic model, higher voice recognition rate can be obtained.

The invention discloses an electronic chromatic dispersion compensation equalizer for optical communication and a method for adjusting a tap thereof. The electronic chromatic dispersion compensation equalizer comprises two symmetrical linear equalizers arranged at a transmitting terminal and a receiving terminal respectively. The structure of the chromatic electronic dispersion compensation equalizer is simpler than a decision feedback equalizer, and the performance of the chromatic electronic dispersion compensation equalizer is better than the linear equalizer. The invention adopts linear fragmentation algorithm to arrange symmetric linear equalizers at the transmitting and the receiving terminals respectively so as to acquire better anti-interference effect than normal linear equalizers and reduce complexity of the system. The invention further uses adjustment step factors to adjust the tap coefficient of the equalizer, and the method is simple in algorithm and good in convergence effect.

The invention discloses a method for profiling a spherical involute spiral bevel gear. The method is characterized by comprising the steps that S1, a mathematical model of a spherical involute spiralbevel gear pair is established; S2, a tooth surface equation of a concave surface and a convex surface of a pair of meshing gears is derived; S3, a tangent plane of the position, at the center point,of the tooth surface is established, and a plane coordinate system is established on the tangent plane; S4, a point on the tooth surface is projected to a the tangent plane, and the coordinate of theprojected point in the tangent plane coordinate system is obtained; S5, the coordinate of the projected point in the S4 is substitute to a shape correction equation to obtain the correction amount ofeach point on the tooth surface; S6, the point on the tooth surface in the S2 is shifted towards the interior of the gear teeth along the tooth surface normal vector at the point; S7, the point cloudobtained after shifting in S6 undergoes high order fitting to form a smooth surface; and S8, a shape correction three-dimensional model of the spiral bevel gear is established on the basis of the profiling tooth surface. The method has the beneficial effects that a very smooth profiling tooth surface can be obtained, full controllable shape correction is achieved, and the transmission quality of the gear pair is effectively guaranteed.

The present invention relates to a convolution neural network training method and an ultrasonic image recognition and location method and system. The method comprise the following steps of establishing a convolution neural network model; taking the convolution neural network model trained on ImageNet as the starting point, reading the B-ultrasonic images in the training data set by different network nodes / layers, and then fine turning the B-ultrasonic images to carry out the migration learning of the neural network model, wherein the images in the training data set come from B-ultrasound examination images of a single center and multiple machines, and are marked with pathological examination results; determining the object position and classification by regression. The invention provides an artificial intelligent identification and positioning system for realizing ultrasonic images, such as ultrasonic images of thyroid nodules, by using migration learning and depth convolution neural network, which greatly improves the accuracy of a model and realizes the shaping of a benign and malignant thyroid nodules auxiliary evaluation and positioning system thereof.

Provided is an analytical method based permanent magnet motor field analysis and torque calculation method. The method comprises the following steps of: step 1: establishing a magnetic vector potential mathematical model of a gap zone and a permanent magnet region; step 2: establishing a magnetic vector potential mathematical model of a slot area; step 3: calculating a cogging torque; and step 4: analyzing and comparing an analytical calculating result. The invention provides an analytical method based permanent magnet motor field analysis and torque calculation method. According to the structural features of a surface mounted radially magnetized permanent magnet motor model, the analytical model is divided into three sub-regions, namely a permanent magnet region, a gap zone and a slot area; a magnetic vector Laplace equation or a Poisson equation is established in each region; sub-region equations are connected through junction conditions, so as to obtain the analytical expression of a load air gap field and a slot field by using a direct analytical method.

The invention discloses a fine-grained image classification method based on a depth convolution neural network, comprising the following steps: 1) preparing a fine-grained image classification data set, and dividing the training data into a training data set and a verification data set; 2) building a depth convolution neural network model for fine-grained image classification, training the model with a training data set, and saving the network model parameters when the trained model reaches the set precision on the verification data set; 3) calculating the tags of test image classification orclassification accuracy of test data sets by using the trained model. The fine-grained image classification framework provided by the invention is independent of and can be applied to any DCNN structure, and has good portability.

The invention discloses a large K-waveband conformal antenna array plane and aims to provide a conformal phased array antenna array plane which does not generate a frequency deviation or a relatively high derotation component, and has a good broadband wide-angle scanning gain characteristics and good broadband wide-angle scanning circular polarization characteristics. The large K-waveband conformal antenna array plane is realized through the following technical scheme: conductor patches are divided into four groups and are arranged on an upper surface of a lower dielectric substrate in a form of a quadrant sequence rotation array, and projections of feeding points of a micro-strip patch antenna on a plane plate are arranged triangularly; the conductor patches are sandwiched between an upper dielectric substrate and the lower dielectric substrate; a radio frequency coaxial connector sequentially passes through the plane plate, a metal cone and the lower dielectric substrate to jointly form the micro-strip patch antenna; the radio frequency coaxial connector carries out coaxial probe feeding on the micro-strip patch antenna; and a pair of orthogonal polarization degenerate modes which have equal amplitudes and a phase difference of 90 degrees are excited on the conductor patches to form the conformal phased array antenna array plane for radiating right-handed circular polarization electromagnetic waves.

The invention discloses a shielding robust face key point positioning method, and belongs to the field of face key point positioning. The method comprises the following steps of S1, collecting face images marked by key points as a training data set and a test data set; S2, for the training data set and the test data set in the step S1, firstly carrying out face detection and five-point key point positioning by using a face detection algorithm MTCNN, and then converting a face by using Procrustes analysis transformation T to obtain a corrected face sample so as to form a training sample and a test sample; taking the corrected face sample as the input of the S3 stage; S3, learning the training sample in the step S2 through a convolutional network; and S4, after the convolutional network completes the training in the step S3, inputting the test sample in the step S2 into the convolutional network, and obtaining the positions of the face key points in the image. The method has robustness for face key point positioning under the shielding condition, and the test result and the qualitative positioning result of the corresponding picture also prove the effectiveness of the method.

The invention relates to a multi-label image classification technology oriented to the field of multimedia content understanding and analysis so as to construct a new classification model and realizea correlation-level-based multi-label classification algorithm. A ranking-learning-based multi-label zero sample classification method provided by the invention comprises the following steps: at a feature extraction stage, carrying out feature description of different modes by using an existing feature extractor to obtain a training data set; at a multi-mode feature transformation stage, giving atraining sample set pair including an original image and a corresponding label between which determined marking information is provided so as to train the model; and at a classification marking stage,giving an original image of a testing sample and a possible label to carry out testing. At the classification marking stage, the correspondence relationship between the original image and the label is not determined. The ranking-learning-based multi-label zero sample classification method is mainly applied to the multi-label image classification occasions.

The invention discloses a local-oscillation leakage automatic removing device, comprising a coupling capacitance used to test the local-oscillation leakage signal, an amplifier used to amplify the detected local-oscillation leakage signal, a high-frequency peak detector used for demodulation of the amplified signal and output DC signal, a programmable lag comparator used to compare the DC signal and reference signal after demodulation. According to the result of comparison, the control signal is exported to control the programmable current source to export compensation signals to remove local-oscillation leakage, and at the same time, control the digital signal processing module of the programmable lag comparator to change the size of he lag window. The invention is capable to remove the local-oscillation leakage in a CMOS wireless radio-frequency transceiver, and improve the effective power of signals.

In order to overcome the defects of the background technology, the invention provides an end-to-end two-stage neural network model to fuse low-light and infrared information to enhance an image so asto achieve the purpose of enhancing a low-light picture. Firstly, we find that local information loss is caused by low light of a low-light picture, and lost information cannot be supplemented by simply enhancing the low-light picture. Inspired by the problem, information in two states is mutually supplemented by considering and combining an infrared image under low light. In the whole model, in the first stage, brightness and contrast improvement is carried out on an area where image pixels are not zero. A foreground attention module is applied to the second part of the network to reconstructan image of the missing area, so that a complete bright image is finally obtained. The beneficial technical effects are that compared with other algorithms, the algorithm introduces the infrared information, and more details are recovered under the condition of local low light, and the enhanced image is more natural as a whole, and the algorithm can process a darker image.

The invention discloses a vibration damping and energy absorption method for an automobile sub-frame, which comprises the steps of (1) performing finite-element analysis to an sub-frame to obtain the quality, the first-order bending mode and the corresponding maximum displacement site of the sub-frame; (2) arranging a mass block with an elastic damping system to damp vibration and absorb energy; and (3) performing modal analysis and experimental verification on the improved sub-frame to correct the elastic damping system and the mass block parameters. The inventive method does not change the current structure of the sub-frame, and arranges the mass block with the elastic damping system in a specific position on the sub-frame to reduce the vibration amplitude by utilizing the principle of resonance energy absorption, so as to reduce the transmission of vibration toward automobile body. The method can also adjust the frequency of vibration damping and energy absorption devices by adjusting the mass of the mass block and the vertical rigidity of the elastic damping system. The inventive method has the advantages of simple structure and not remarkably increased weight and cost, and can maximally damp transmission of the vibration toward the automobile body.

The invention relates to a video-based automatic identification system for abnormal conditions of a fused magnesium furnace. The method includes the steps of acquiring fused magnesia production fieldcondition information through an image information acquisition module and transmitting the information to a sample generation module; generating a training sample by the sample generation module, performing label processing and manual classification label processing on images through label-image software, and transmitting the processed video and image information to a detection and classificationmodule; performing, by the detection and classification module, feature extraction and processing on the video and image information through a corresponding algorithm to obtain an intelligent detection and classification model of the condition identification system; and displaying, by a display unit, the novel uses the test video and image information in a visualized manner by using the intelligent detection and classification model processed by the detection and classification module. The video-based automatic identification system of the invention has obviously improved identification effects as compared to a system using non-visual information alone such as the current and voltage, is reasonable in hardware device structure, convenient in operation, low in cost and high in the accuracyof discrimination and can replace manual routing inspection.

The invention provides a method for extracting an attribution and a comment word with a template based on internet, which comprises the steps of: memorizing an attribute template aiming at a concept, the classification of which is manually labeled, by a machine readable dictionary; checking the evaluation index of a classification algorithm, selecting the maximum entropy, and ensuring a PMI value and an attribution word when training a classification model; extracting an original attribution set by collocating a template based on the internet, filtering the attribution based on a classification rule, expanding an attribution set with a connecting phrase template based on the assumption of Resnik, filtering again to form a process of cyclic iteration, evaluating the performance of the attribution extracting method according to the precision, the recall ratio and the comprehensive index F to obtain a relationship curve graph of the precision and the recall ratio; and extracting an individal comment word to the product attribution based on the product attribution extracted by the internet to form an effective attribution-evaluation pair, judging good evaluation and bad evaluation according to the word characteristic of an evaluation adjective in evaluation information, and generating a market feedback comprehensive value.

The invention discloses a forward-looking radar imaging sea-surface target key point detection and recognition method. The method includes the steps that radar echo data are quantized into a gray scale image; a region of interest is extracted from the gray scale image, partition is performed, and a target region partition image is obtained; the radar echo data and the region of interest of the image are used, and a target region peak point information matrix is obtained; information merging is conducted on the target region partition image and the target region peak point information matrix, the number K of peak points in a target region in a merging result is obtained through counting, the previous K peak points are selected and serve as target effective peak points and form a target effective peak point image in a binaryzation mode, target axial features are extracted, and a target position is determined; the center of gravity of target energy is calculated and serves as the target key point. According to forward-looking radar target features, multiple mode identification methods are used comprehensively, the inherent features of a target can be reserved, meanwhile interference factors such as artifacts and side lobes can be restrained, and the recognition accuracy and the positioning precision of radar imaging sea surface target key points are improved.

The invention discloses an observer-based adaptive command filter control method for output consistency of a multi-single-arm manipulator. Modeling of dynamic characteristics of a multi-single-arm manipulator is performed; and an adaptive controller is designed by combining a backstep method with a command filtering technique. During the controller design process, an unknown item of the system approximates by using a fuzzy logic system, so that the need of the precise dynamic model of the manipulator is eliminated. A state observer estimates an unmeasurable state in the system. With the command filtering technique, a complicated signal derivation calculation explosion problem is solved. Since the manipulator system output is restricted, an output restriction problem is solved by using an obstacle Lyapunov function. For a mechanical connection part of the manipulator, the driver always has a non-smooth input dead zone problem; and the influence on the system by the input dead zone is compensated by using the slope information of the input dead zone. The manipulator system output can track the reference signal effectively; and all signals are semi-globally bounded ones.

The invention discloses a gesture output compensation method of a single-axial rotary inertial navigation system. The method comprises the following steps: deducting the rotational angular velocity of the earth and gyroscopic drift from angular velocity output by a top z and integrating to obtain an angle; making a difference between the angle of the integral of the top z and a rotating angle output by a coded disc and fitting the difference value to configure compensation quantity of the course angle; deducting gravity acceleration projection from output of x and y accelerometers and angle of pitch and roll angle obtained by initial alignment so as to configure compensation quantity of the angle of pitch and the roll angle; and compensating the gesture output of the single-axial rotary inertial navigation system according to a gesture output compensation algorithm. The gesture output compensation algorithm is easy to implement, and not only can satisfy the real-time requirement of project application, but also can greatly enhance the gesture output precision of the single-axial rotary inertial navigation system, so that the method is of important meaning to improvement of navigation performance of the single-axial rotary inertial navigation system.