Linear prediction method based on orthogonal compensation

Abstract

The invention discloses a linear prediction method based on orthogonal compensation, and relates to the technical field of data processing and machine learning, and the method comprises the steps: calculating a pixel offset between two adjacent frames of images after initial linear compensation and corresponding offset data through an image matching algorithm; taking the offset data as a difference value between an observed value and a predicted value, and taking a reference value of the reference data as a state quantity; if the offset is normal, fusing the real-time data by using an interactive multi-model proposed based on the Bayesian theory, and performing model training on the system linear data error model by taking a predicted value of the interactive multi-model proposed based on the Bayesian theory as input and a linear data error of the system linear error model as output; and if the offset is abnormal, a scale value is calculated by using the compensated data information, so that the corrected data information is obtained, and the situation that the recognition capability is greatly reduced and even missing detection or false detection may occur when an abnormal type which is not fully learned is encountered is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and machine learning, and in particular to a linear prediction method based on orthogonal compensation. Background Art

[0002] With the continuous advancement of science and technology, data processing and machine learning, as important sources of training model data for modern electronic equipment, have become crucial for the detection of performance and safety. In the performance analysis of machines and equipment, accurately identifying energy loss anomalies is a key factor in evaluating their quality and service life. However, traditional data processing and machine learning analysis methods often rely on a large number of learning samples, and when faced with complex and changeable abnormal situations, the recognition accuracy and efficiency are subject to certain limitations. Although existing data processing and machine learning analysis technologies can identify anomalies by comparing and analyzing data processing and machine learning data, in actual operations, due to the complexity and diversity of machine learning performance data, these technologies often find it difficult to accurately capture all types of abnormal losses. In addition, when encountering anomaly types that have not been fully learned, the recognition ability of existing technologies will drop significantly, and even missed detection or false detection may occur. Based on this, a linear prediction method based on orthogonal compensation is proposed to solve the above problems. Summary of the invention

[0003] The object of the present invention is to provide a linear prediction method based on orthogonal compensation to solve the problems raised in the above background technology.

[0004] By adopting the above technical solution, a linear prediction method based on orthogonal compensation includes:

[0005] S1: Collect real-time data and set the benchmark value of the benchmark data;

[0006] Benchmark values ​​based on benchmark data;

[0007] S2: padding the predicted data until the data length of the time domain data to be predicted reaches the expected target length;

[0008] S3: Calculate the compensation value of the initial inertial data through a linear compensation algorithm and perform initial linear compensation;

[0009] S4: calculating the pixel offset between two adjacent frames of images after initial linear compensation and the corresponding offset data through an image matching algorithm;

[0010] The offset data is taken as the difference between the observed value and the predicted value, and the benchmark value of the benchmark data is taken as the state quantity; if the offset is normal, the interactive multi-model proposed based on Bayesian theory is used to fuse the real-time data, and the predicted value of the interactive multi-model proposed based on Bayesian theory is taken as input, and the linear data error of the system linear error model is taken as output, and the system linear data error model is trained;

[0011] If the offset is abnormal, a recursive orthogonal search algorithm is used to build an online system linear data error model, and the linear data error of the system linear data error model is estimated in combination with the predicted value of the interactive multi-model to compensate for the output data information, and the standard value is calculated using the compensated data information to obtain the corrected data information.

[0012] According to the above technical solution, the interactive multi-model is obtained by the following method:

[0013] Acquire a plurality of first time domain data;

[0014] Extracting part of the data in each of the first time domain data, respectively constructing a plurality of second time domain data corresponding to the plurality of first time domain data, wherein the data length of each second time domain data is the same as the data length of the corresponding first time domain data;

[0015] Performing Fourier transform processing on the plurality of first time domain data to obtain a plurality of first frequency domain data, and performing Fourier transform processing on the plurality of second time domain data to obtain a plurality of second frequency domain data;

[0016] The plurality of second frequency domain data and the plurality of first frequency domain data are used as a training set of a preset convolutional neural network model, and the preset convolutional neural network model is trained; when the number of training times of the preset convolutional neural network model exceeds a preset number of times, the training is stopped; the second frequency domain data in the training set is input into the convolutional neural network model after the training is stopped, and the output third frequency domain data is obtained;

[0017] Perform inverse Fourier transform processing on the third frequency domain data to obtain third time domain data.

[0018] According to the above technical solution, the first valid data of the third time domain data is compared with the second valid data of the corresponding second time domain data, and when the data length of the first valid data is greater than the data length of the second valid data, it is determined that the data lengths of the two meet the preset verification condition;

[0019] When the data length of the first valid data is less than or equal to the data length of the second valid data, it is determined that the data lengths of the two do not meet the preset verification conditions; wherein the valid data is data from part of the data extracted from the corresponding first time domain data; when the verification is successful, it is determined that the convolutional neural network model after stopping training has converged to obtain an orthogonal compensation data linear prediction model.

[0020] According to the above technical solution, the preset convolutional neural network model includes K convolutional layers, each convolutional layer has P pooling layers, the input of each convolutional layer is the output of all other layers before the convolutional layer, and the size of each convolution kernel is R. The input-output relationship of the convolutional neural network model can be defined as:

[0021] y=f(x,α);

[0022] Among them, x is the input layer of the convolutional neural network, y is the output layer of the convolutional neural network, x and y are both frequency domain data, ɑ is the parameter of the convolutional neural network model, and f(x,α) represents the nonlinear mapping relationship of the convolutional neural network.

[0023] According to the above technical solution, a large amount of easy-to-analyze frequency domain data is used as the input and output of the preset convolutional neural network model, and the preset convolutional neural network model is trained, which can improve the model training rate. The training is stopped when the model reaches the convergence condition, and a trained nuclear magnetic resonance data linear prediction model is obtained at the same time.

[0024] According to the above technical solution, the convergence condition of the convolutional neural network model also includes that the number of model training reaches a preset number threshold and satisfies the verification condition, and multiple second frequency domain data and multiple first frequency domain data are used as the training set of the preset convolutional neural network model, and the preset convolutional neural network model is trained. When the number of training times of the preset convolutional neural network model exceeds the preset number value, the training is stopped, and the convolutional neural network model after the training is stopped is verified by the training set. When the verification is successful, it is determined that the convolutional neural network model after the training is stopped has converged.

[0025] According to the above technical solution, the interactive multi-model tracking algorithm mainly includes the following four steps, namely, input model interaction, filter filtering, probability model update and new data fusion.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. A linear prediction method based on orthogonal compensation, which calculates the pixel offset and corresponding offset data between two adjacent frames of images after initial linear compensation through an image matching algorithm; the offset data is taken as the difference between the observed value and the predicted value, and the reference value of the reference data is taken as the state quantity; if the offset is normal, the interactive multi-model proposed based on Bayesian theory is used to fuse the real-time data, and the predicted value of the interactive multi-model proposed based on Bayesian theory is used as input, and the linear data error of the system linear error model is used as output to train the system linear data error model; if the offset is abnormal, the recursive orthogonal search algorithm is used to model the system linear data error model online, and the linear data error of the system linear data error model is estimated in combination with the predicted value of the interactive multi-model, so as to compensate for the output data information, and use the compensated data information to calculate the standard value, so as to obtain the corrected data information, so as to avoid a significant decrease in recognition ability when encountering anomaly types that have not been fully learned, and even possible missed detection or false detection.

[0028] 2. A linear prediction method based on orthogonal compensation, through an interactive multi-model tracking algorithm, mainly includes the following four steps, namely input model interaction, filter filtering, probability model updating and new data fusion. The interactive multi-model Kalman filter tracking algorithm is a filtering recursive process. Each time, the output of the previous interaction is used as the input value of the next interaction, thereby completing the cycle of the entire filtering tracking process, solving the problem of invalid linear motion compensation caused by inaccurate recognition ability estimation or inability to adopt estimation means in previous methods, so that data processing and machines can be automatically and quickly corrected in a passive state, so as to perform accurate linear motion compensation. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the structure of the interactive multi-model algorithm process of the present invention;

[0030] Figure 2 It is the simulation operation diagram of CV and CA of the present invention;

[0031] Figure 3 This is a weight change diagram of the model of the present invention;

[0032] Figure 4 It is the probability change diagram of CV and CA of the present invention;

[0033] Figure 5 FIG1 is a comparison between the MM algorithm of the present invention and the Kalman filter algorithm;

[0034] Figure 6 FIG2 is a comparison of the MM algorithm of the present invention and the Kalman filter algorithm. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Embodiment 1:

[0037] A linear prediction method based on orthogonal compensation provided by the present invention is preferably implemented as follows: Figures 1 to 6 As shown: A linear prediction method based on orthogonal compensation, comprising:

[0038] S1: Collect real-time data and set the benchmark value of the benchmark data;

[0039] Benchmark values ​​based on benchmark data;

[0040] S2: padding the predicted data until the data length of the time domain data to be predicted reaches the expected target length;

[0041] S3: Calculate the compensation value of the initial inertial data through a linear compensation algorithm and perform initial linear compensation;

[0042] S4: calculating the pixel offset between two adjacent frames of images after initial linear compensation and the corresponding offset data through an image matching algorithm;

[0043] The offset data is taken as the difference between the observed value and the predicted value, and the benchmark value of the benchmark data is taken as the state quantity; if the offset is normal, the interactive multi-model proposed based on Bayesian theory is used to fuse the real-time data, and the predicted value of the interactive multi-model proposed based on Bayesian theory is taken as input, and the linear data error of the system linear error model is taken as output, and the system linear data error model is trained;

[0044] If the offset is abnormal, a recursive orthogonal search algorithm is used to build an online system linear data error model, and the linear data error of the system linear data error model is estimated in combination with the predicted value of the interactive multi-model to compensate for the output data information, and the standard value is calculated using the compensated data information to obtain the corrected data information.

[0045] Specifically in the above technical solution, .

[0046] Further,.

[0047] in,.

[0048] The data lengths of the plurality of first time domain data are the same and are L, and the first time domain data is defined as F1, and the expression of F1 is as follows:

[0049]

[0050] Where N represents the number of spectrum resonance peaks, A n represents the amplitude of the nth resonance peak, Tn represents the relaxation time of the nth resonance peak, t represents the sampling time, and w n Represents the frequency of the nth formant.

[0051] Part of the data in each of the first time domain data is extracted, and a plurality of second time domain data corresponding to the plurality of first time domain data are respectively constructed.

[0052] The data length of each second time domain data is the same as the data length of the corresponding first time domain data, the extracted partial data is continuous data, and the continuous data is extracted from the first bit of the first time domain data, and the data length of the continuous data is smaller than the data length of the corresponding first time domain data, the data length of the second time domain data is the same as the data length of the first time domain data, and the partial data extracted from the first time domain data is half of the data of the first time domain data, that is, the second time domain data only contains the first half of the data of the first time domain data.

[0053] Corresponding data is obtained to obtain multiple extracted data, and data padding is performed on the multiple extracted data so that the data length of each extracted data reaches the data length of the corresponding first time domain data, to obtain multiple second time domain data; wherein the preset threshold range is the range starting from the first bit of the time domain data, and the process of data padding is the process of increasing the value to 0.

[0054] The first half of the data in the second time domain data is extracted from the first time domain data, and its amplitude is the same as the first half of the corresponding first time domain data. In order to control the range of linear prediction and improve the accuracy of model training, the data length of the second time domain data is increased to the same as the data length of the corresponding first time domain data, and the data value of the increased part is set to 0, as shown in the input second time domain data.

[0055] The data length of the first time domain data is 10. The first time domain data includes (1, 2, 3, 4, 5, 6, 7, 8, 9.10). The preset threshold range is 1-5, then the extracted data is (1, 2, 3, 4, 5), and then the extracted data is filled with data with a value of 0 so that the data length of the extracted data reaches the corresponding data length of the first time domain data, 10, to obtain the second time domain data, then the second time domain data is (1, 2, 3, 4, 5, 0, 0, 0, 0, 0).

[0056] By setting the value of the padding data to 0, the model can mainly predict the data with the value of 0 during training, so that the model has the ability to convert the data with the value of 0 into predicted data.

[0057] The method of constructing the second time domain data also includes: keeping the first m points of the first time domain data F1 unchanged, setting the points after the m points to 0, and marking them as F2. The expression of F2 is as follows:

[0058] F2=F1;

[0059] ID2(m+1:1)=0;

[0060] Among them, 1 represents the index of the last element of F2.

[0061] A plurality of first time domain data are processed by Fourier transform to obtain a plurality of first frequency domain data, and a plurality of second time domain data are processed by Fourier transform to obtain a plurality of second frequency domain data.

[0062] By performing Fourier transform processing on multiple first time domain data and multiple second time domain data, the time domain data that is difficult to process can be transformed into frequency domain data that is easy to analyze, so that the frequency domain data that is easy to analyze can be directly processed or processed, thereby accelerating the prediction efficiency of the data. The multiple second frequency domain data are used as input data, and the multiple first frequency domain data are used as output data. The preset convolutional neural network model is trained until convergence to obtain a data linear prediction model.

[0063] The preset convolutional neural network model contains K convolutional layers, each with P pooling layers. The input of each convolutional layer is the output of all other layers before the convolutional layer. The size of each convolution kernel is R. The input-output relationship of the convolutional neural network model can be defined as:

[0064] y=f(x,α);

[0065] Among them, x is the input layer of the convolutional neural network, y is the output layer of the convolutional neural network, x and y are both frequency domain data, α is the parameter of the convolutional neural network model, and y=f(x,α) represents the nonlinear mapping relationship of the convolutional neural network.

[0066] A large amount of easy-to-analyze frequency domain data is used as the input and output of the preset convolutional neural network model, and the preset convolutional neural network model is trained to improve the model training rate. The training is stopped when the model reaches the convergence condition, and the trained nuclear magnetic resonance data linear prediction model is obtained at the same time.

[0067] The convergence condition of the convolutional neural network model includes that the accuracy of the model reaches a certain value. For example, when the output accuracy of the model reaches 98%, it is determined that the model has reached the convergence condition, and the training can be stopped. The model with an output accuracy of 98% can be used as the trained linear prediction model for nuclear magnetic resonance data.

[0068] The convergence condition of the convolutional neural network model also includes that the number of model training reaches a preset number threshold and meets the verification condition, multiple second frequency domain data and multiple first frequency domain data are used as the training set of the preset convolutional neural network model, and the preset convolutional neural network model is trained. When the number of training times of the preset convolutional neural network model exceeds the preset number value, the training is stopped, and the convolutional neural network model after the training is stopped is verified by the training set. When the verification is successful, it is determined that the convolutional neural network model after the training is stopped converges. When the model is trained 50,000 times, the training is stopped, and the model is verified by the training set. When the model passes the verification, it can be determined that the model meets the convergence condition.

[0069] Embodiment 2:

[0070] On the basis of the first embodiment, a preferred embodiment of a linear prediction method based on orthogonal compensation provided by the present invention is shown in the figure: a CV model is established, and the state space of the CV model can be expressed as:

[0071]

[0072] Then the transfer function is:

[0073]

[0074] Alternatively, the complete state space of the CV model can also be expressed as:

[0075]

[0076] The matrix is ​​represented as:

[0077]

[0078] Furthermore, the main idea of ​​the interactive multi-model IMM control algorithm is the automatic identification and switching between models based on the Bayesian theory: at any tracking moment, real-time maneuver model detection is performed by setting model filters corresponding to the number of possible target models, setting weight coefficients and the probability of model update for each filter, and finally weighted calculation to obtain the current optimal estimation state, thereby achieving the purpose of model adaptive tracking.

[0079] The following is the IMM algorithm process Figure 1 :

[0080] As can be seen from the figure, an interactive multi-model tracking algorithm mainly includes the following four steps, namely input model interaction, filter filtering, probability model update and new data fusion.

[0081] Assume that r motion models are set in the interactive multi-model algorithm, and their state equations are as follows:

[0082] X(k+1)=A j X(k)+W j (k); j = 1, 2, 3...

[0083] Where X(k) is the state vector of the system, A j is the state transfer matrix of the target, W j (k) has a mean of 0 and a covariance matrix of Q j Gaussian white noise.

[0084] The observation equation of any model j in the model set is:

[0085] Z(k)=H j X(k)+V j (k);

[0086] Where Z(k) is the measurement vector, H j is the observation matrix of model j, V j (k) has a mean of 0 and a covariance of R j The white noise matrix.

[0087] The model transformation matrix can be expressed as:

[0088]

[0089] Where P ij Represents the transfer matrix from model i to model j;

[0090] 1. Input interaction;

[0091] The current measurement value Z(k) is inferred based on the system state estimate and covariance estimate at the previous moment, and then the model is reinitialized. The new initial value is obtained through the Markov operation matrix between different models. The optimal state estimate and estimated covariance matrix of the model at time k-1 are set as follows:

[0092] X j (k-1|k-1) and p j (k-1|k-1);

[0093] Then the initial conditions of the model after interaction are further derived:

[0094]

[0095]

[0096] in:

[0097]

[0098]

[0099] 2. Filter filtering;

[0100] Since the Kalman filter algorithm has excellent tracking effect, it is usually used to filter and predict the target in multi-model interactive tracking. The input of the j model at time k-1 is the optimal estimate X oj (k-1|k-1) and update the covariance p 0j (k-1|k-1), the tracking process is similar to the single model Kalman filter algorithm. The Kalman filter process of the IMM algorithm is as follows:

[0101] Its Kalman filter first-order prediction equation and prediction covariance matrix are:

[0102] X I (k-1|k-1)=A j X oj (k-1|k-1);

[0103] p j (k|k-1)=A j p oj (k-1|k-1)A T j +Q j ;

[0104] The Kalman filter gain is:

[0105] K j (k) = p j (k|k-1)H T [Hp j (k|k-1)H T +Q j ] -1 ;

[0106] The system state estimate and covariance estimate corresponding to a single filter output at time k are:

[0107] X j (k|k)=X j (k|k-1)+K j (k)[Z(k)-HX j (k|k-1)];

[0108] p j(k|k)=[IK j (k)H]p j (k|k-1);

[0109] 3. Update of model probability;

[0110] In the IMM algorithm, model update directly affects the effectiveness of the algorithm. The algorithm uses the maximum likelihood function method to update the model: by calculating the similarity between the current model and the current state of the moving target, the weight of the most suitable tracking model is given.

[0111] The maximum likelihood function that best matches the j model at time k is:

[0112]

[0113] Among them, S j (k) = Hp j (k|k-1)H T +R,d j (k) = Z j (k)-H j X j (k|k-1);

[0114] Then the update probability of model j can be obtained as:

[0115]

[0116] Where c is the normalization constant, that is:

[0117]

[0118] 4. New fusion data;

[0119] The final output result of the interaction data at time k is given according to the tracking results calculated separately by each model and the weight of model matching, namely:

[0120] The final output of the interaction data at time k is given based on the tracking results calculated separately for each model and the weight of the model matching, namely: the overall estimated value:

[0121]

[0122] Population covariance:

[0123]

[0124] The above is a filtering recursive process of the interactive multi-model Kalman filter tracking algorithm. Each time, the previous interactive output is used as the next interactive input value, thus completing the cycle of the entire filtering tracking process.

[0125] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions recorded in the aforementioned embodiments, or to replace some of the technical features therein by equivalents. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A linear prediction method based on orthogonal compensation, characterized in that: include: S1: Collect real-time data and set the benchmark value of the benchmark data; Benchmark values ​​based on benchmark data; S2: padding the predicted data until the data length of the time domain data to be predicted reaches the expected target length; S3: Calculate the compensation value of the initial inertial data through a linear compensation algorithm and perform initial linear compensation; S4: calculating the pixel offset between two adjacent frames of images after initial linear compensation and the corresponding offset data through an image matching algorithm; The offset data is taken as the difference between the observed value and the predicted value, and the benchmark value of the benchmark data is taken as the state quantity; If the offset is normal, the interactive multi-model proposed based on Bayesian theory is used to fuse the real-time data, and the predicted value of the interactive multi-model proposed based on Bayesian theory is used as input, and the linear data error of the system linear error model is used as output to train the system linear data error model; If the offset is abnormal, a recursive orthogonal search algorithm is used to build an online system linear data error model, and the linear data error of the system linear data error model is estimated in combination with the predicted value of the interactive multi-model to compensate for the output data information, and the standard value is calculated using the compensated data information to obtain the corrected data information.

2. The linear prediction method based on orthogonal compensation according to claim 1, characterized in that: The interactive multiple model is obtained by the following method: Acquire a plurality of first time domain data; Extracting part of the data in each of the first time domain data, respectively constructing a plurality of second time domain data corresponding to the plurality of first time domain data, wherein the data length of each second time domain data is the same as the data length of the corresponding first time domain data; Performing Fourier transform processing on the plurality of first time domain data to obtain a plurality of first frequency domain data, and performing Fourier transform processing on the plurality of second time domain data to obtain a plurality of second frequency domain data; The plurality of second frequency domain data and the plurality of first frequency domain data are used as a training set of a preset convolutional neural network model, and the preset convolutional neural network model is trained; when the number of training times of the preset convolutional neural network model exceeds a preset number of times, the training is stopped; the second frequency domain data in the training set is input into the convolutional neural network model after the training is stopped, and the output third frequency domain data is obtained; Perform inverse Fourier transform processing on the third frequency domain data to obtain third time domain data.

3. The linear prediction method based on orthogonal compensation according to claim 2, characterized in that: Compare the first valid data of the third time domain data with the second valid data of the corresponding second time domain data, and when the data length of the first valid data is greater than the data length of the second valid data, determine that the data lengths of the two meet a preset verification condition; When the data length of the first valid data is less than or equal to the data length of the second valid data, it is determined that the data lengths of the two do not meet the preset verification conditions; wherein the valid data is data from part of the data extracted from the corresponding first time domain data; when the verification is successful, it is determined that the convolutional neural network model after stopping training has converged to obtain an orthogonal compensation data linear prediction model.

4. The linear prediction method based on orthogonal compensation according to claim 2, characterized in that: The preset convolutional neural network model includes K convolutional layers, each convolutional layer has P pooling layers, the input of each convolutional layer is the output of all other layers before the convolutional layer, the size of each convolution kernel is R, and the input-output relationship of the convolutional neural network model can be defined as: y=f(x,α); Among them, x is the input layer of the convolutional neural network, y is the output layer of the convolutional neural network, x and y are both frequency domain data, ɑ is the parameter of the convolutional neural network model, and f(x,α) represents the nonlinear mapping relationship of the convolutional neural network.

5. The linear prediction method based on orthogonal compensation according to claim 4, characterized in that: A large amount of easy-to-analyze frequency domain data is used as the input and output of the preset convolutional neural network model. The preset convolutional neural network model is trained to improve the model training rate. The training is stopped when the model reaches the convergence condition, and a trained nuclear magnetic resonance data linear prediction model is obtained at the same time.

6. The linear prediction method based on orthogonal compensation according to claim 2, characterized in that: The convergence condition of the convolutional neural network model also includes that the number of model training reaches a preset number threshold and satisfies the verification condition, multiple second frequency domain data and multiple first frequency domain data are used as the training set of the preset convolutional neural network model, and the preset convolutional neural network model is trained. When the number of training times of the preset convolutional neural network model exceeds the preset number value, the training is stopped, and the convolutional neural network model after the training is stopped is verified by the training set. When the verification is successful, it is determined that the convolutional neural network model after the training is stopped has converged.

7. The linear prediction method based on orthogonal compensation according to claim 1, characterized in that: The interactive multi-model tracking algorithm mainly includes the following four steps: input model interaction, filter filtering, probability model update and new data fusion.

Images