10 results about "Filter model" patented technology

The invention relates to the field of liquid level measurement, and discloses a liquid level measurement method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a liquid level signal obtained by monitoring a to-be-measured liquid level at present and / or the acceleration during liquid movement; the liquid level signal is processed through a preset first filtering model, a first measurement result of the liquid level is obtained, and the first filtering model processes the liquid level signal through historical liquid level signals in the sliding window; adjusting the size of the sliding window according to the acceleration to obtain a second filtering model, and obtaining a second measurement result of the liquid level by using the second filtering model; a third filtering model corresponding to the current liquid level state is determined according to the fluctuation condition of the liquid level signal, and a third measurement result of the liquid level is obtained through the model; and performing weighted fusion on at least two of the first measurement result, the second measurement result and the third measurement result to obtain a final measurement result. Different filtering models are fused, and the reliability of liquid level measurement is improved when the liquid is in a non-stationary state.

This application discloses a filtering method, a filtering model training method, and related apparatuses, belonging to the field of coding technology. The method includes determining K filtering model groups based on quantization parameters of a target image, determining a reconstruction block corresponding to a current coding block in the target image, determining a target filtering model from the K filtering model groups, and filtering the reconstruction block based on the target filtering model. The same filtering model group is applicable to coding blocks having the same coding quality, different filtering model groups are applicable to coding blocks having different coding qualities, and different filtering models within the same filtering model group are applicable to coding blocks having different contents. Therefore, for the reconstruction block corresponding to the current coding block, the target filtering model can be selected from the K filtering model groups with reference to the coding quality and content of the coding block, and then the reconstruction block is filtered based on the target filtering model to reduce coding distortion and improve filtering performance.

The application provides a kind of guided artillery shell dynamics parameter identification method based on double-layer game UKF, this method mainly includes filter model construction and UKF filtering improvement two parts, according to six degrees of freedom trajectory dynamics model and aerodynamic extension equation, filter model is constructed, so as to obtain system state equation, according to the parameter coupling relationship between guided artillery shell dynamics model and inertial device, decoupling is obtained gyro output angular velocity and accelerometer output specific force, so as to construct system observation equation;Based on the innovation of filter model and prior observation data, design a kind of double-layer game improved UKF method, inner game is composed of noise mapping method and Sage-Husa method, for evaluating the dynamic relationship between prediction and actual measurement;Outer game assesses two kinds of adaptive methods by calculating Mahalanobis distance, which reflects the game between the predicted observation covariance and the actual measurement covariance.The application can adapt to unknown noise environment, self-adapting adjustment observation noise covariance, improve guided artillery shell parameter identification precision.

A coding method, a bitstream, an encoder, a decoder, and a storage medium are disclosed. The method includes the following. Multiple candidate filtering models for the current block are determined. The multiple candidate filtering models are ordered to determine a target filtering-model-candidate list for the current block. A target filtering model for the current block is determined according to the target filtering-model-candidate list. Coefficients of the target filter are determined according to the target reference region. Intra prediction is performed on the current block according to the coefficients of the target filter to determine a prediction value of the current block. In this case, the candidate filtering models can be obtained through different combinations of reference regions and filters, so types of filtering models are expanded. Then, the candidate filtering models are ordered to obtain the filtering-model-candidate list, and the target filtering model is determined from the filtering-model-candidate list for intra prediction. In this way, intra prediction accuracy and coding performance can be improved.

The invention provides an attitude estimation method and device, equipment and a storage medium. Relates to the technical field of state estimation. The method comprises the following steps: based on fractional order extended Kalman filtering, introducing a fractional order margin, a fractional order covariance correction term and a fractional order residual, and constructing a filter module; wherein the filter module takes angle data as input and outputs an attitude estimation value; constructing a fitness function according to the error between the true value attitude and the attitude estimation value of the filter module, and optimizing the fractional order of the filter module to obtain the optimized fractional order; and configuring the filter module by using the optimized fractional order, and realizing attitude estimation based on the configured filter module. According to the method, by introducing mechanisms such as fractional order margin, fractional order covariance correction and fractional order residual error, the constructed filter module effectively enhances the utilization of the FEKF on historical information and nonlinear dynamic, and the accuracy and robustness of the filtering model are improved.

The invention discloses a dual-mode adaptive random model adjustment method, system and device and a medium, and the method comprises the steps: obtaining real-time state sensing data at each epoch of a Kalman filtering calculation process through monitoring a prediction error sequence of Kalman filtering and calculating a normalized prediction error square statistic; judging whether the monitoring point enters a high-dynamic mode or not; in response to the judgment of entering the high-dynamic mode, executing a model adjustment operation through a continuous adaptive control algorithm to obtain an adaptive filtering model for the high-dynamic mode; when the adaptive filtering model is adopted, judging whether a dynamic event is ended or not; in response to judging that the dynamic event is ended, recovering the filtering model to a state before adjustment; a filtering model in a current state is applied to original observation data for positioning calculation, a new prediction error sequence generated by calculation is fed back, and a continuous high-precision positioning result is obtained. According to the invention, timeliness and reliability of geological disaster monitoring and early warning are substantially improved.

The invention discloses a filtering method, a filtering model training method and a related device, and belongs to the technical field of coding and decoding. The method comprises the steps of determining K groups of filtering models according to quantization parameters of a target image, determining a reconstruction block corresponding to a current coding block in the target image, determining a target filtering model from the K groups of filtering models, and filtering the reconstruction block based on the target filtering model. Since the same group of filtering models is suitable for coding blocks with the same coding quality, different groups of filtering models are suitable for coding blocks with different coding qualities, and different filtering models in the same group of filtering models are suitable for coding blocks with different contents. Therefore, for the reconstruction block corresponding to the current coding block, the target filtering model can be selected from the K groups of filtering models in combination with the coding quality and the content of the coding block, and then the reconstruction block is filtered based on the target filtering model, so that the coding distortion is reduced, and the filtering performance is improved.

A method for modelling of a set of filters is provided. The method comprises acquiring a set of feature values each of which is associated with an index within an index range of the filters and dividing the index range into multiple segments using the acquired set of feature values. The method also comprises determining a filter model for at least one segment of the multiple segments and outputting the determined filter model.

The invention belongs to the technical field of artificial intelligence security, and particularly relates to a back door defense method based on confusion back door features, and the method comprises the steps: obtaining a credible sample set and a suspicious sample set based on a poisoning data set and prediction entropy; inputting the credible sample set and the suspicious sample set into a poisoning model, constructing a model loss function, and obtaining a filtering model based on the model loss function; inputting the poisoning data set into a filtering model, and obtaining a credible data set and a suspicious data set by judging the consistency between the output of the filtering model and the original label; and based on the trusted data set and the suspicious data set, adopting a semi-supervised learning method to obtain a benign model, thereby realizing defense of backdoor attacks. According to the invention, current mainstream backdoor attacks can be effectively defended, and the prediction accuracy of the model can be maintained; according to the method, a small part of benign samples are extracted from the poisoning data set through the prediction entropy, extra clean samples are not needed, and the method is more reasonable in an actual defense scene.

A method for modelling of a set of filters is provided. The method comprises acquiring a set of feature values each of which is associated with an index within an index range of the filters and dividing the index range into multiple segments using the acquired set of feature values. The method also comprises determining a filter model for at least one segment of the multiple segments and outputting the determined filter model.