CN103743435A - Multi-sensor data fusion method - Google Patents

Abstract

The invention discloses a multi-sensor data fusion method. Multi-sensors acquire data signals and the data signals are converted through an A/D converter so as to obtain digital signals; the digital signals undergo data filtering and pretreatment; and finally, feature extraction and algorithm fusion are carried out to obtain a result after multi-sensor data fusion. During the data-level multi-sensor data fusion, two same sensors simultaneously measure a same indoor environment quality factor. Thus, real-time performance and measurement accuracy of sensors of the same type are simultaneously raised. By the method, original measurement data of sensors is fully utilized; and fusion of information, such as error of mean square, measurement accuracy and the like, of sensors is carried out. It is not required to know any experiment knowledge about sensor measurement data. Drifting and noise of sensors are inhibited to some extent, and measurement accuracy of a system is raised.

Description

A kind of multi-Sensor Information Fusion Approach

Technical field

The present invention relates to a kind of multi-Sensor Information Fusion Approach.

Background technology

At present, along with the raising of the development of sensor Data Fusion and the complexity of problem to be solved, the limitation exposed day by day of single blending algorithm, is combined into several data blending algorithm the development trend of sensor Data Fusion.And current adaptive weight fusion estimated algorithm is the algorithms most in use to same type of sensor data fusion.In the measuring process of each factor of Indoor Environmental Quality, there is noise error in the actual measured value obtaining, and passes judgment on the levels of precision of actual measurement data through conventional square error.

Summary of the invention

The defect that the present invention seeks to exist for prior art provides a kind of multi-Sensor Information Fusion Approach.

The present invention for achieving the above object, adopts following technical scheme: a kind of multi-Sensor Information Fusion Approach, and multisensor obtains data-signal, then by A/D converter, changes, and obtains digital signal; Then carry out data filtering and pre-service, and then obtain the result after Fusion through feature extraction and algorithm fusion.

Further, described multisensor is for adopting two sensors, and the variance of two sensors is respectively σ ₁ ², σ ₂ ², the true value that estimate is X, the measured value of sensor is respectively x ₁, x ₂, be mutually independent, and

be X without inclined to one side estimation, weighting factor is respectively w ₁, w ₂, the weighting factor after fusion and value meet respectively:

w ₁+w ₂=1 （2-1）

$\hat{x}=\left[\begin{array}{cc}{w}_1& w_2\end{array}\right]{\left[\begin{array}{cc}{x}_1& x_2\end{array}\right]}^T=\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_i---(2-2)$

From formula (2-2)

for X without inclined to one side estimation, and fusion value is the linear function of each measurement value sensor;

Each sensor variance is:

${\mathrm{\σ}}^2=E\left[\frac{1}{4}\underset{j=1}{\overset{4}{\mathrm{\Σ}}}{(x_j-\stackrel{\‾}{x})}^2\right]---(2-3)$

In formula (2-3): i is for measuring number of times;

Total mean square deviation is:

${\mathrm{\σ}}^2=E\left[{(x-\hat{x})}^2\right]=E[\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i^2{(x-x_i)}^2+2\underset{\underset{i\≠j}{i=1,j=1}}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_j(x-x_i)(x-x_j)---(2-4)$

From formula (2-4), total meansquaredeviationσ ²be the polynary secondary letter about each weighting factor, certainly exist minimum value, known according to Lagrange's theorem extreme value theory, at total meansquaredeviationσ ²a hour corresponding weighting factor is:

$w_i^{\′}=\frac{1}{{\mathrm{\σ}}_i^2\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-5)$

Now corresponding Minimum Mean Square Error is:

${{\mathrm{\σ}}_{\min }}^2=\frac{1}{\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-6).$

Beneficial effect of the present invention: the present invention adopts two identical sensors to measure same Indoor Environmental Quality factor in data level Fusion simultaneously, has improved real-time and the measuring accuracy of same type of sensor so simultaneously.The method takes full advantage of the raw measurement data of sensor, the information such as the square error, measuring accuracy of sensor of saying merge, do not require any experimental knowledge of knowing sensor measurement data, it has suppressed drift and the noise of sensor to a certain extent, has improved the measuring accuracy of system.

Embodiment

The present invention relates to a kind of multi-Sensor Information Fusion Approach, multisensor obtains data-signal, then by A/D converter, changes, and obtains digital signal; Then carry out data filtering and pre-service, and then obtain the result after Fusion through feature extraction and algorithm fusion.

Further, described multisensor is for adopting two sensors, and the variance of two sensors is respectively σ ₁ ², σ ₂ ², the true value that estimate is X, the measured value of sensor is respectively x ₁, x ₂, be mutually independent, and x be X without inclined to one side estimation, weighting factor is respectively w ₁, w ₂, the weighting factor after fusion and value meet respectively:

w ₁+w ₂=1 （2-1）

$\hat{x}=\left[\begin{array}{cc}{w}_1& w_2\end{array}\right]{\left[\begin{array}{cc}{x}_1& x_2\end{array}\right]}^T=\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_i---(2-2)$

From formula (2-2)

for X without inclined to one side estimation, and fusion value is the linear function of each measurement value sensor;

Each sensor variance is:

${\mathrm{\σ}}^2=E\left[\frac{1}{4}\underset{j=1}{\overset{4}{\mathrm{\Σ}}}{(x_j-\stackrel{\‾}{x})}^2\right]---(2-3)$

In formula (2-3): i is for measuring number of times;

Total mean square deviation is:

${\mathrm{\σ}}^2=E\left[{(x-\hat{x})}^2\right]=E[\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i^2{(x-x_i)}^2+2\underset{\underset{i\≠j}{i=1,j=1}}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_j(x-x_i)(x-x_j)---(2-4)$

From formula (2-4), total meansquaredeviationσ ²be the polynary secondary letter about each weighting factor, certainly exist minimum value, known according to Lagrange's theorem extreme value theory, at total meansquaredeviationσ ²a hour corresponding weighting factor is:

$w_i^{\′}=\frac{1}{{\mathrm{\σ}}_i^2\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-5)$

Now corresponding Minimum Mean Square Error is:

${{\mathrm{\σ}}_{\min }}^2=\frac{1}{\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-6).$

The foregoing is only preferred embodiment of the present invention, in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (2)

1. a multi-Sensor Information Fusion Approach, is characterized in that, multisensor obtains data-signal, then by A/D converter, changes, and obtains digital signal; Then carry out data filtering and pre-service, and then obtain the result after Fusion through feature extraction and algorithm fusion.

2. a kind of multi-Sensor Information Fusion Approach as claimed in claim 1, is characterized in that, described multisensor is for adopting two sensors, and the variance of two sensors is respectively σ ₁ ², σ ₂ ², the true value that estimate is X, the measured value of sensor is respectively x ₁, x ₂, be mutually independent, and x be X without inclined to one side estimation, weighting factor is respectively w ₁, w ₂, the weighting factor after fusion and value meet respectively:

w ₁+w ₂=1 （2-1）

$\hat{x}=\left[\begin{array}{cc}{w}_1& w_2\end{array}\right]{\left[\begin{array}{cc}{x}_1& x_2\end{array}\right]}^T=\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_i---(2-2)$

From formula (2-2) for X without inclined to one side estimation, and fusion value is the linear function of each measurement value sensor;

Each sensor variance is:

${\mathrm{\σ}}^2=E\left[\frac{1}{4}\underset{j=1}{\overset{4}{\mathrm{\Σ}}}{(x_j-\stackrel{\‾}{x})}^2\right]---(2-3)$

In formula (2-3): i is for measuring number of times;

Total mean square deviation is:

${\mathrm{\σ}}^2=E\left[{(x-\hat{x})}^2\right]=E[\underset{i=1}{\overset{2}{\mathrm{\Σ}}}{w}_i^2{(x-x_i)}^2+2\underset{\underset{i\≠j}{i=1,j=1}}{\overset{2}{\mathrm{\Σ}}}{w}_i{x}_j(x-x_i)(x-x_j)---(2-4)$

From formula (2-4), total meansquaredeviationσ ²be the polynary secondary letter about each weighting factor, certainly exist minimum value, known according to Lagrange's theorem extreme value theory, at total meansquaredeviationσ ²a hour corresponding weighting factor is:

$w_i^{\′}=\frac{1}{{\mathrm{\σ}}_i^2\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-5)$

Now corresponding Minimum Mean Square Error is:

${{\mathrm{\σ}}_{\min }}^2=\frac{1}{\underset{i=1}{\overset{2}{\mathrm{\Σ}}}\frac{1}{{\mathrm{\σ}}_i^2}}---(2-6).$