16results about How to "Guaranteed estimation accuracy" patented technology

The application discloses a mixed expert network-based optical flow estimation method and system, belonging to the field of computer vision and artificial intelligence. First, the input two frames of images are preprocessed, low-resolution features are extracted through a mixed expert feature extractor (MoEE) containing a sparse activation mechanism to reduce redundant calculation; then, dot product operation is performed on the low-resolution feature maps to construct a 4D correlation volume to capture pixel motion matching relationship; subsequently, a mixed expert updater (MoEU) is used to iteratively update the hidden state and regress the residual flow increment through a dynamic expert selection mechanism to optimize the optical flow accuracy; finally, a multi-scale upsampling module is used to reconstruct the high-resolution optical flow field to output the high-precision optical flow result. The algorithm realizes dynamic resource allocation through the MoE architecture, significantly reduces the calculation cost while ensuring the accuracy of optical flow estimation, and can adapt to resource-constrained scenarios such as automatic driving and unmanned aerial vehicles, and can balance efficient inference and flexible deployment.

This application relates to a joint estimation method for wireless communication delay and frequency offset based on hill-climbing fine search. The method includes: setting a time offset search range, a frequency offset search range, a coarse search step size, and a fine search step size based on prior information; generating a two-dimensional grid point set within the search range according to the coarse search step size; calculating the cost value of each grid point based on a cost function; taking the time offset and frequency offset corresponding to the maximum value as the coarse estimate; using the coarse estimate as the initial point, performing fine search iterations using hill-climbing; calculating the cost value of multiple candidate points in the neighborhood in each iteration, comparing and determining the update direction, and updating the current time offset and / or frequency offset; repeating the iteration until a preset exit condition is met, and outputting the final estimate. This application effectively solves the parameter coupling problem of traditional step-by-step estimation in NTN scenarios, while balancing search range and computational complexity, facilitating engineering implementation, and being applicable to various NTN systems and terminal forms.

The application provides a DOA estimation method based on a hybrid analog-digital array architecture design, which comprises the following steps: a hybrid analog-digital array is arranged at a receiving end architecture; the hybrid analog-digital array adopts a low-complexity HAD array architecture; the hybrid analog-digital array architecture is used to receive an incident signal and model; an initial value of an analog phase shifter network is designed; a difference virtual array of the low-complexity HAD array architecture is calculated; an initial angle estimation value is obtained from the difference virtual array receiving signal by using a classical direction finding algorithm; after the initial angle estimation value is obtained, the phase shifter network weight is redesigned, and then a difference virtual array output signal is obtained; and a DOA estimation result is obtained from the classical direction finding algorithm; the method can effectively avoid the problems of phase ambiguity and beam scanning, can reduce the cost and complexity, and can effectively approach the estimation performance of a full-digital array architecture.

The application discloses a cross-modal gaze target estimation method and system based on hierarchical supervision and a medium, relates to the technical field of computer vision, and inputs a target person in an acquired RGB image into a gaze target model after adding face frame coordinate labeling to the target person, so as to generate a corresponding gaze target estimation heat map; the training process of the gaze target model is as follows: a training set is constructed, original visual features and original text features are obtained through feature extraction, a face mask and reinforced visual features are generated at the same time, the original text features and the reinforced visual features are encoded and spliced, the spliced feature sequence obtained through splicing is input into a cross-modal Transformer, the reinforced visual features are connected with visual feature residuals output by the cross-modal Transformer, and a gaze target estimation heat map is generated; hierarchical supervision loss is set to optimize model parameters; and the method improves target estimation precision.

The application provides a mobile robot self-learning robust control method based on a hybrid Kalman filter and belongs to the technical field of robots.The technical scheme comprises the following steps: step 1, a dynamic model of an uncertainty factor;step 2, real-time estimation of state parameters and model parameters of the mobile robot to optimize the original dynamic model;step 3, introduction of an adaptive forgetting factor recursive least square method to dynamically adjust filter parameters;step 4, design of a non-singular fast terminal sliding mode controller based on a saturation function as a system main controller;step 5, establishment of an error correction RBF neural network and introduction of an adaptive error learning strategy;step 6, completion of stability analysis and proof of the control system by means of Lyapunov stability theory and simulation experiment of the optimized model.The application effectively suppresses the chattering of the sliding mode control and significantly improves the control precision and robust performance of the four-wheel robot.

The application belongs to the field of computer vision three-dimensional point cloud registration, and particularly relates to a point cloud registration method based on registration error zero norm minimization. * ; using l2 norm to quantify the overall fitting error of the fitting error of each relative point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K r ; using l2 norm to quantify the overall fitting error of the fitting error of each relative point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K * ; using l2 norm to quantify the overall fitting error of the fitting error of each relative point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K * ; using l2 norm to quantify the overall fitting error of the fitting error of each relative point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K t ; using l2 norm to quantify the overall fitting error of the fitting error of each relative point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K * ; using l2 norm to quantify the overall fitting error of the fitting error of each relative * point pair in three spatial dimensions, and according to the matching point pair corresponding to the minimum K The application can realize efficient point cloud registration in the presence of a large number of outliers.

This invention relates to a collaborative underwater acoustic signal delay difference estimation method, belonging to the field of underwater acoustic positioning technology. This invention solves the problem that existing methods cannot achieve both low computational complexity and high delay difference estimation accuracy, and also cannot meet real-time requirements. Specifically, the invention involves: obtaining the arrival time of the received signal at each hydrophone and the corresponding copy correlation envelope peak value for each hydrophone based on the received signal received by hydrophones installed on an unmanned platform; calculating the delay difference between adjacent hydrophones based on the copy correlation method; further calculating the delay difference corresponding to the maximum and second-largest peaks of the generalized cross-correlation of adjacent hydrophones, as well as the standard cross-correlation envelope value; calculating the effective amplitude of the received signal at each hydrophone based on the calculated copy correlation envelope peak value and the standard cross-correlation envelope values ​​corresponding to the maximum and second-largest peaks of the generalized cross-correlation; and calculating the final delay difference for each group of adjacent hydrophone channels based on the effective amplitude of the received signal. This invention can be applied to the field of underwater acoustic positioning technology.

This invention presents a novel health status estimation method for lithium-ion batteries based on novel health features. First, it obtains the charging voltage curve under constant current and constant voltage charging conditions, using the voltage difference at equal starting voltages and time intervals as the novel health feature. Then, since the starting voltage and time interval are two parameters affecting the novel health feature, they are optimized, resulting in an optimized starting voltage and multiple optimized time intervals from the same charging voltage curve. Combining the optimized starting voltage with each optimized time interval yields multiple novel health features extracted from the same charging voltage curve. Finally, a lithium battery health status estimation model is constructed based on a multiple linear regression model. The model is trained using the extracted novel health features and corresponding health status values, and the trained model is then used for lithium battery health status estimation. This method has low computational complexity and significantly improves estimation speed while maintaining accuracy, making it applicable even when current battery management systems have insufficient computing power.

This invention relates to a time-domain synchronous averaging method based on correlation theory, comprising the following steps: calculating the average rotational speed of the gear shaft using the maximum spectral energy method; obtaining the most stable periodic gear meshing vibration harmonics through bandpass filtering; performing iterative correlation analysis on the most stable periodic gear meshing vibration harmonics to estimate the starting sequence; and using the starting sequence to segment, spline fit interpolate, and synchronously average the vibration signal to obtain the time-domain synchronous average signal. This invention introduces correlation theory twice to ensure the accuracy of the final time-domain synchronous averaging result. The bandpass filtering step selects a filter bandwidth including first-order sidebands, and since gears are usually subject to frequency modulation when faulty, this method is more robust to fault data analysis and has good application and promotion value.

The invention provides a multi-judge scoring deviation correction method and system based on hierarchical estimation, and relates to the technical field of evaluation, and the method comprises the steps: obtaining scoring data; estimating initial relative severity parameters of the judges on the basis of pairwise comparison among the judges in the scoring data; the initial relative severity parameter is used as a fixed estimation value, and an item reaction theoretical model matched with the scoring data type is adopted to estimate and obtain an initial distinction degree parameter of each judge and an initial ability parameter of each candidate; taking the initial relative severity parameter, the initial distinction degree parameter and the initial capability parameter as initial values, and based on a pre-constructed joint optimization objective function, performing joint optimization by minimizing the joint optimization objective function to obtain a final estimation value; and generating a corrected scoring result based on the final estimation value. By applying the technical scheme of the invention, the ability parameters of the testee can be efficiently determined in a multi-judge subjective scoring scene, and the more fair correction of the scoring data is realized.

The invention provides a communication method, a communication node, a storage medium and a program product. The method comprises the steps of determining a first time domain reference position, determining a first time domain length corresponding to N transmission units according to the first time domain reference position, N being a positive integer, determining a time domain position of a demodulation reference signal according to the first time domain length, the N transmission units being capable of sharing the demodulation reference signal, effectively reducing the load of the demodulation reference signal, and improving the user experience. In the embodiment of the invention, the more reasonable time domain position of the demodulation reference signal is obtained according to the time domain length corresponding to the N transmission units, even if N is equal to 1, DMRS time domain position alignment of a plurality of transmission units or a plurality of user equipment can be realized through the first time domain reference position, interference management of the demodulation reference signal is facilitated, and demodulation performance is ensured.

The invention relates to the technical field of underwater acoustics and inertial navigation integrated navigation, in particular to a USBL delay compensation method and system based on selection state augmented information filtering, and the method comprises the steps: predicting an information matrix and an information vector at a current moment based on navigation state information vector estimation and a navigation state covariance information matrix at a previous moment; constructing an augmented information matrix of a diagonal structure based on the augmented state vector and obtaining an augmented information vector; and calculating a measurement information matrix and an information vector increment corresponding to each delay measurement, and executing additive updating operation to construct a newly-added wide information matrix and a newly-added wide information vector with delay measurement at the current moment so as to recover the navigation state at the current moment by utilizing a Schur compensation method. According to the method, the Markov characteristics of the navigation system are utilized to construct the augmented information matrix with a block diagonal structure, the sparsity of the matrix is fully mined, and the real-time performance and numerical stability requirements of the underwater navigation system are ensured by utilizing the advantages of information filtering in the aspect of measurement updating.

The application discloses a short-time fractional Fourier transform-based super-large array near-field space non-stationary channel estimation method and system, belongs to the technical field of signal and information transmission, and solves the problem that existing technologies cannot accurately estimate channel parameters and have large pilot overhead when estimating a near-field space non-stationary channel. The application comprises pilot signal receiving and processing; according to a preset near-field coverage distance and an angle range, a range of a frequency modulation slope of an angular frequency is calculated, the range is uniformly sampled to obtain a frequency modulation slope set, and a fractional order rotation angle set is obtained through mapping; a short-time fractional inverse Fourier transform sub-codebook is constructed and integrated and arranged to obtain a total codebook; a sparse set and sparse coefficients of the near-field space non-stationary channel in a short-time fractional Fourier transform domain are iteratively estimated; and a near-field space non-stationary channel vector is reconstructed in combination with the total codebook. The application is suitable for a super-large array near-field space non-stationary channel estimation scene.

The invention discloses a noise estimation method and device for an orthogonal frequency division multiplexing communication system, and the method comprises the following steps: S1, calculating respective initial noise estimation matrixes on subcarriers based on a received training sequence; s2, on the basis of the initial noise estimation matrix, a measurement parameter vector used for representing noise characteristics of each subcarrier is obtained through calculation, and the dimension of the measurement parameter vector is equal to the number of the subcarriers; s3, dynamically and adaptively dividing all the subcarriers into one or more continuous frequency segments according to the measurement parameter vector, and determining the dividing boundary by the change characteristic of the measurement parameter vector in the frequency domain; and S4, for each frequency segment obtained by division, based on the initial noise estimation matrix corresponding to the subcarriers contained in the segment, calculating a noise covariance matrix of the segment, so that colored noise can be more effectively converted into white noise, the detection error of the signal is remarkably reduced, and the receiving performance and demodulation reliability of the whole communication system are improved.

This invention discloses a method and system for estimating the main-trailer angle by asynchronously fusing the main vehicle trajectory and trailer wheel speed, belonging to the field of automotive electronics technology. The method includes: acquiring distance information from two detection points at the rear of the main vehicle to the ground; acquiring the main vehicle's position, speed, heading angle, and attitude angle; calculating the absolute coordinates of the landing points of the detection points and continuously acquiring data to form the main vehicle's motion trajectory; establishing a kinematic model; calculating the observed yaw rate of the trailer; using an extended Kalman filter, estimating the trailer's heading angle in real time with the kinematic model as the prediction equation; performing time alignment correction on the state estimation based on the asynchronous relationship between the timestamps of the main vehicle's motion trajectory observation data and the wheel speed observation data; fusing the observation data into the state estimation to obtain the corrected trailer heading angle; and subtracting the corrected trailer heading angle from the main vehicle's heading angle to obtain the main-trailer angle, effectively suppressing the periodic accumulation and divergence of wheel speed integral errors.

The invention discloses an unmanned aerial vehicle monitoring system for forest carbon sink estimation. The unmanned aerial vehicle monitoring system comprises an unmanned aerial vehicle flight platform, a multi-sensor module, a data transmission module, an edge calculation module, a cloud server and a carbon reserve estimation module. The multi-sensor module is detachably connected to the unmanned aerial vehicle flight platform and used for synchronously collecting multi-dimensional original data of a forest ecosystem, and the multi-sensor module at least comprises a laser radar sensor, a hyperspectral sensor, a thermal infrared sensor and a visible light camera; the edge calculation module is integrated on the unmanned aerial vehicle flight platform. Through fusion of the multi-source sensor, multi-dimensional data such as a three-dimensional structure, biochemical characteristics, thermal characteristics and texture characteristics of a forest can be acquired, and reliable basic data is provided for carbon reserve estimation; local real-time processing of data can be realized through the edge calculation module, so that the processing delay is greatly reduced; and through the carbon reserve estimation module, the estimation precision in different scenes can be ensured according to a forest type optimization model.