CN108960334A - A kind of multi-sensor data Weighted Fusion method - Google Patents

Abstract

The invention discloses a kind of multi-sensor data Weighted Fusion methods for distributed multi-sensor detection system, this method is filtered amendment to the actual measurement data of each sensor first, improve the confidence level of single-sensor measurement data, secondly based on the data fusion weight between data acquisition sensor after amendment, improve the reasonability of data fusion weight distribution, the Weighted Fusion of multi-sensor data is finally carried out using amendment data and based on data fusion weight between the sensor for correcting data acquisition, therefore, the distributed multi-sensor fusion results obtained using data weighting fusion method of the invention are closer to actual conditions, confidence level is higher.

Description

A weighted fusion method for multi-sensor data

technical field

The invention belongs to the field of information fusion, in particular to a sensor data fusion.

Background technique

In a distributed multi-sensor detection system, multiple sensors are often used to measure the same parameter repeatedly, and the detection accuracy of the system is improved through data fusion. There are many methods for data fusion. Compared with the subjective arbitrariness and modeling limitations in the practical application of Bayesian decision-making and neural network and other fusion technologies, the weighted fusion method does not require prior information and has relatively high fusion accuracy. advantages have received widespread attention. However, the multi-sensor data weighted fusion methods in the prior art all use the original measurement data of the sensors for inter-sensor data fusion, and do not consider filtering the original measurement data of the sensors, resulting in low reliability of the fused data. Therefore, how to further improve the credibility after data fusion is a technical problem urgently needed to be solved by those skilled in the art.

Contents of the invention

The technical problem mainly solved by the present invention is to provide a multi-sensor data weighted fusion method to solve the technical problem that the reliability of the data fusion method in the existing distributed multi-sensor detection system in the prior art is not high.

In order to solve the above-mentioned technical problems, a technical solution adopted by the present invention is: a multi-sensor data weighted fusion method, including step S1, performing filter correction on each sensor data in the plurality of sensors to obtain the multi-sensor data The corrected data; step S2, calculate the inter-sensor fusion weight of the data fusion between different sensors among the multiple sensors; step S3, perform secondary weighted fusion based on the corrected data and the inter-sensor fusion weight, to obtain The final fusion result.

In another embodiment of the multi-sensor data weighted fusion method of the present invention, filter correction is performed on each sensor data of the plurality of sensors, and obtaining the corrected data of the plurality of sensors includes:

Step S21, setting sensor data is the jth measurement data of the i-th sensor among the plurality of sensors, where 1≤i≤N, N is the number of the plurality of sensors, 1≤j≤M, and M is the number of the plurality of sensors total number of measurements per sensor; obtain all basic data sets for each of the plurality of sensors where B(i,j) represents the j-th basic group of the i-th sensor;

Step S22, calculating mean value EB(i,j) and variance DB(i,j) of each said basic data set B(i,j);

Step S23, calculating the intragroup probability PB(i, j)=[pb(i, j) of each element in the basic data set B(i, j) appearing in the basic data set B(i, j). 1),pb(i,2),...,pb(i,j)];

Step S24, according to the basic data set B(i,j) and the primary intra-group probability PB(i,j) for the jth measurement data of the i-th sensor Make corrections to obtain the corrected data X=[X ₁ ,X ₂ ,...,X _N ] of all sensors, where Xi _i =[ _xi (1), _xi (2),..., _xi (M )] is the corrected data of the i-th sensor.

In another embodiment of the multi-sensor data weighted fusion method of the present invention, the calculation of inter-sensor fusion weights for data fusion between different sensors among the multiple sensors includes:

Step S31, calculating the mean value EX _i and variance DX _i of the corrected data of each sensor in the plurality of sensors;

Step S32, calculating inter-sensor fusion weight W=[w ₁ ,w ₂ ,…,w _N ] for fusion of different sensor data, where Indicates the data fusion weight of the i-th sensor.

In another embodiment of the multi-sensor data weighted fusion method of the present invention, performing secondary weighted fusion based on the corrected data and the inter-sensor fusion weight to obtain a final fusion result includes:

Step S41, performing a fusion of the corrected data of the plurality of sensors and the inter-sensor fusion weight to obtain a fusion data set Z=[z ₁ ,z ₂ ,...,z _M ]=WX ^T ;

Step S42, obtaining the mean value EZ and variance DZ of the primary fusion data set Z;

Step S43, calculating the secondary within-group probability PZ=[pz ₁ ,pz ₂ ,...,pz _M ] of each element in the primary fusion data set Z appearing in the primary fusion data set Z;

Step S44, performing secondary data fusion on the primary fusion data set Z and the secondary intragroup probability PZ to obtain a final fusion result Y=(PZ)Z ^T .

The beneficial effects of the present invention are: the present invention discloses a multi-sensor data weighted fusion method for a distributed multi-sensor detection system, the method first performs filter correction on the actual measurement data of each sensor, and improves the accuracy of the single-sensor measurement data. Credibility, secondly based on the corrected data to obtain the data fusion weight between sensors, improve the rationality of the data fusion weight distribution, and finally use the corrected data and the inter-sensor data fusion weight based on the corrected data to carry out weighted fusion of multi-sensor data, so , the distributed multi-sensor fusion result obtained by using the data weighted fusion method of the present invention is closer to the actual situation and has higher reliability.

Description of drawings

Fig. 1 is the flow chart of an embodiment of multi-sensor data weighted fusion method of the present invention;

Fig. 2 is a flowchart of an embodiment of the method for filtering and correcting sensor data in the present invention;

Fig. 3 is a flowchart of an embodiment of the method for calculating inter-sensor fusion weights for data fusion between different sensors among multiple sensors in the present invention;

Fig. 4 is the flow chart of an embodiment of the method for twice weighted fusion of multi-sensor data in the present invention;

Fig. 5 is a flow chart of another embodiment of the multi-sensor data weighted fusion method of the present invention.

Detailed ways

In order to facilitate the understanding of the present invention, the present invention will be described in more detail below in conjunction with the accompanying drawings and specific embodiments. Preferred embodiments of the invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described in this specification. On the contrary, these embodiments are provided to make the understanding of the disclosure of the present invention more thorough and comprehensive.

It should be noted that, unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by those skilled in the technical field of the present invention. Terms used in the description of the present invention are only for the purpose of describing specific embodiments, and are not used to limit the present invention.

Fig. 1 is a flow chart of an embodiment of the multi-sensor data weighted fusion method of the present invention. In Figure 1, the method includes:

Step S1, filter and correct each sensor data in the plurality of sensors to obtain corrected data of the plurality of sensors. In this step, based on the measurement data of a certain sensor, all the measurement data of the sensor are respectively corrected to obtain the corrected data of all the measurements of the sensor, and so on to obtain the corrected data of all the sensors.

Step S2, calculating inter-sensor fusion weights for data fusion between different sensors among the plurality of sensors. In this step, based on the corrected data of all sensors obtained in step 1, the weight of data fusion between different sensors is calculated.

Step S3, performing quadratic weighted fusion based on the corrected data and the inter-sensor fusion weights to obtain a final fusion result. In this step, based on the corrected data of all sensors obtained in step S1 and the inter-sensor fusion weights obtained in step S2 for data fusion between different sensors, the secondary fusion of data is performed to obtain the final fusion result.

It can be seen from the above that the data weighted fusion method of the present invention first filters and corrects the actual measurement data of each sensor to improve the reliability of the single sensor measurement data, and secondly obtains the data fusion weight between sensors based on the corrected data, Improve the rationality of data fusion weight distribution, and finally use the correction data and the inter-sensor data fusion weights based on the correction data to carry out weighted fusion of multi-sensor data. Therefore, the multi-sensor fusion results obtained by using the data weighted fusion method of the present invention are closer to In reality, the credibility is higher.

Fig. 2 is a flowchart of an embodiment of the method for filtering and correcting sensor data according to the present invention. In Figure 2, the method includes:

Step S21, setting sensor data is the jth measurement data of the i-th sensor among the plurality of sensors, where 1≤i≤N, N is the number of the plurality of sensors, 1≤j≤M, and M is the number of the plurality of sensors total number of measurements per sensor; obtain all basic data sets for each of the plurality of sensors where B(i,j) represents the j-th basic group of the i-th sensor;

Step S22, calculating mean value EB(i,j) and variance DB(i,j) of each said basic data set B(i,j);

Step S23, calculating the intragroup probability PB(i, j)=[pb(i, j) of each element in the basic data set B(i, j) appearing in the basic data set B(i, j). 1),pb(i,2),...,pb(i,j)];

Step S24, according to the basic data set B(i,j) and the primary intra-group probability PB(i,j) for the jth measurement data of the i-th sensor Make corrections to obtain the corrected data X=[X ₁ ,X ₂ ,...,X _N ] of all sensors, where Xi _i =[ _xi (1), _xi (2),..., _xi (M )] is the corrected data of the i-th sensor.

Specifically, in the present invention, it is assumed that there are N mutually independent sensors in a distributed detection system to measure the parameters of a certain object to be measured, and the measurement equation is:

In formula (1), is the parameter measurement value of the i-th sensor for the kth time, x _i is the real value of the parameter, δ _i (k) is the measurement noise of the i-th sensor for the k-th time, and its mean value is zero and the variance is The Gaussian distribution of where k≥1.

And it can be seen from formula (1) that the measurement data of the sensor obeys the mean value of _xi and the variance of The Gaussian distribution of

For the k-th measurement data of the i-th sensor The k measurement data of the sensor before the kth time and the kth time Statistical analysis was performed as a set of data.

According to the concept of probability density function, it can be known that the jth measurement data of the i-th sensor There is the following relationship with the real value x _i of the parameter to be measured and the standard deviation σ _i of the measurement noise:

According to the maximum likelihood estimation theory, the k measurement data of the i-th sensor before the kth time and the kth time of the sensor can be obtained The maximum likelihood estimate of the mean and standard deviation of is:

Considering that k is a finite value, the maximum likelihood estimation of σ _i is modified as:

Then, calculate the k measurement data of the i-th sensor before the kth time and the kth time of the sensor The probability that each measurement data in the group appears in all the measurement data in this group.

According to the probability density function of normal distribution, when k>1, the jth (1≤j≤k) measured value of sensor i The probability of occurrence is:

In the formula, are the k measured values of the i-th sensor before the kth time and the kth time of the sensor mean and standard deviation of .

Then the j-th measurement data of the i-th sensor The probability of occurrence in all measurement data in this group is:

Finally, the k-th measurement data of the i-th sensor is corrected, and the above steps are repeated until all the measurement data of the i-th sensor and the measurement data of all sensors are corrected.

FIG. 3 is a flow chart of an embodiment of the method for calculating inter-sensor fusion weights for data fusion between different sensors among multiple sensors according to the present invention. In Figure 3, the method includes:

Step S31, calculating the mean value EX _i and variance DX _i of the corrected data of each sensor in the plurality of sensors;

Step S32, calculating inter-sensor fusion weight W=[w ₁ ,w ₂ ,…,w _N ] for fusion of different sensor data, where Indicates the data fusion weight of the i-th sensor.

Specifically, it is assumed that the i-th sensor acquires a total of M measurement data within the specified time (M>1), after the above steps, the correction value is Then the mean and variance of the M corrected data of the i-th sensor are:

According to the idea of weighted fusion, if the data of the i- _{th sensor is Xi and its weight is ω i ,} then the result of weighted fusion of all N sensors is:

make Then Z _i obeys the standard normal distribution, and the fusion result in formula (9) can be expressed as:

It can be seen from formula (10) that the fusion result Y obeys the mean variance normal distribution of . Using the method of least squares, when When , the variance of the fusion result Y is the smallest

Fig. 4 is a flow chart of an embodiment of the method for twice-weighted fusion of multi-sensor data in the present invention, the method comprising:

Step S41, performing a fusion of the corrected data of the plurality of sensors and the inter-sensor fusion weight to obtain a fusion data set Z=[z ₁ ,z ₂ ,...,z _M ]=WX ^T ;

Step S42, obtaining the mean value EZ and variance DZ of the primary fusion data set Z;

Step S43, calculating the secondary within-group probability PZ=[pz ₁ ,pz ₂ ,...,pz _M ] of each element in the primary fusion data set Z appearing in the primary fusion data Z;

Step S44, performing secondary data fusion on the primary fusion data set Z and the secondary intragroup probability PZ to obtain a final fusion result Y=(PZ)Z ^T .

Specifically, firstly, weighted fusion is performed according to the formula (11) using the sensor correction data and the weights of data fusion between different sensors.

Then, take the fusion data as a group to form a primary fusion data group, calculate the mean value and variance of the data group, and the secondary intra-group probability of each element in the group for data fusion, and then perform weighted fusion to obtain the final result of data fusion. Wherein, the method for calculating the secondary intra-group probability of each element in the group for data fusion is the same as the method for calculating the occurrence probability of each element in the group when calculating the Mth measurement data correction of a certain sensor.

Specifically, the present invention also discloses an embodiment of using the multi-sensor data weighted fusion method to perform data fusion on thermostat parameters. Assume that in a distributed multi-sensor detection system, three thermocouple sensors detect the incubator six times. Among them, Table 1 shows the multiple measured values of each sensor on the incubator.

Table 1 The six measured values of each sensor on the incubator

(1) Take the current measurement and historical measurement data of different sensors as a basic data group, and calculate the mean value and variance of the sensor data group.

For sensor S1:

Since there is no measurement data before the first measurement, the mean value of the first basic data set of sensor S1 is 899.5 of its measurement value, and the variance is 0;

Since there is the first measurement before the second measurement, the first and second measurement values are used as the second basic data set, and can be obtained through formula (3) and formula (4), the second basic data set of sensor S1 The mean is 902.4 and the variance is 16.82;

Similarly, the mean and variance of the third, fourth, fifth and sixth basic data sets of the sensor S1 can be calculated. Table 2 is the mean and variance of the six basic data sets of sensor S1.

Table 2 The mean and variance of the six basic data sets of sensor S1

According to the above steps, the mean value and variance of the first to sixth basic data groups of the sensor S2 and the sensor S3 can be calculated. Table 3 is the mean value and variance of the six basic data groups of sensor S2 and sensor S3.

Table 3 The mean and variance of the six basic data sets of sensor S2 and sensor S3

(2) On the basis of step (1), calculate the probability that each measurement data in the corresponding basic data group of the sensor appears in all the measurement data in the group.

For sensor S1:

Since there is no measurement data before the first measurement, the variance of the first basic data set is 0, therefore, the probability that the first measurement data appears in all the measurement data in the first basic data set is 1;

Since there is the first measurement before the second measurement, according to formula (5) and formula (6), the first and second measurement data in the second basic data set of sensor S1 are all in the second basic data set The probability of occurrence in the measured data is

In the same way, the probability of each measurement data in the third, fourth, fifth and sixth basic data groups of sensor S1 appearing in all the measurement data in the corresponding group can be calculated. Table 4 is the six basic data groups of sensor S1. The probability that each measurement data occurs in all data in the corresponding group.

Table 4. Each measurement data in the six basic data groups of sensor S1 is all in the corresponding group

probability of occurrence in the data

According to the above steps, the probability of each measurement data in the first to sixth basic data groups of sensor S2 and sensor S3 appearing in all the data in the corresponding group can be calculated. The probability that each measurement data appears in all the data in the corresponding group, Table 6 shows the probability that each measurement data in the six basic data groups of sensor S3 appears in all the data in the corresponding group.

Table 5 The probability of each measurement data in the six basic data groups of sensor S2 appearing in all the data in the corresponding group

Table 6 The probability of each measurement data in the six basic data groups of sensor S3 appearing in all the data in the corresponding group

(3) On the basis of step (2), correct the current measurement data in the basic data set of the sensor, and repeat steps 1 and 2 until all the measurement data of all the sensors are corrected.

For sensor S1:

According to the formula (7), the corrected data of the first measurement is 899.5; the corrected data of the second measurement is 899.5×0.5+905.3×0.5=902.4.

Similarly, the corrected data of the third, fourth, fifth and sixth measurements of the sensor S1 can be calculated. Table 7 shows the corrected data of all the measurement data of the sensor S1.

Table 7 Corrected data of all measured data of sensor S1

According to the above steps, the corrected data for each measurement of the sensor S2 and the sensor S3 can be calculated, and the corrected data of all the measured data of the sensor S2 and the sensor S3 can be calculated in Table 8.

Table 8 Corrected data of all measurement data of sensor S2 and sensor S3

(4) On the basis of step (3), calculate the weight of data fusion between different sensors.

According to the formula (8), the mean and variance of all the corrected data of different sensors are calculated, and Table 9 shows the mean and variance of all the corrected data of different sensors.

Table 9 Mean and variance of all corrected data for different sensors

sensor S1 S2 S3 average 901.0857 890.8015 899.6151 variance 1.1856 20.4217 3.4289

Then the inter-sensor fusion weight of sensor S1 participating in data fusion is The inter-sensor fusion weight of sensor S2 participating in data fusion is The inter-sensor fusion weight of sensor S3 participating in data fusion is

(5) On the basis of step (4), weighted fusion is performed by using the corrected sensor data and the inter-sensor fusion weights of data fusion between different sensors.

According to formula (11), the fusion data set can be obtained as:

(6) On the basis of step (5), calculate the mean value and variance of the fusion data group Y, and the secondary intra-group probability of each element in the data group Y for data fusion, and then perform weighted fusion to obtain the final data fusion result.

According to formula (3) and formula (4), the mean value of the fusion data set Y is 900.2982 and the variance is 1.0241.

According to the formula (5) and formula (6), the probability of each data in the data Y appearing in all the data can be obtained. Table 10 shows the secondary intra-group probability of each data in the fusion data group Y appearing in all the data in the group. Among them, in the method of calculating the probability of each data appearing in all data in the fusion data group Y and calculating the probability of each data appearing in the group in the 6th basic data group in sensors S1, S2, S3 The probability approach is the same.

Table 10 The secondary intra-group probability of each data in the fusion data group appearing in all the data in the group

data 898.7607 901.6072 900.9032 900.7955 900.0160 899.7063 probability 0.0738 0.1013 0.1956 0.2073 0.2250 0.1971

According to formula (9), the final result of data fusion is 900.3587.

Fig. 5 is a flow chart of another embodiment of the multi-sensor data weighted fusion method of the present invention. In Figure 5, the method includes:

Step S51, taking the current measurement data of each sensor and all historical measurement data before the current measurement data as a basic group, and calculating the mean value and variance of the basic group measurement data of the sensor.

Specifically, assuming that N mutually independent sensors measure the parameters of an object to be measured, the measurement equation is:

In formula (12), is the parameter measurement value obtained by the i-th sensor for the kth time, _xi is the real value of the parameter, δ _i (k) is the measurement noise of the i-th sensor for the k-th time, and its mean value is zero and the variance is The Gaussian distribution of where k≥1.

It can be seen from formula (12) that the measured value of the sensor obeys the mean value of x _i and the variance of The Gaussian distribution of

For the k-th measurement value of the i-th sensor The k measured values of the sensor before the kth time and the kth time Statistical analysis was performed as a basic data set.

According to the concept of probability density function, it can be obtained that the j-th measurement data of the i-th sensor There is the following relationship with the real value x _i of the parameter to be measured and the standard deviation σ _i of the measurement noise:

According to the maximum likelihood estimation theory, the k measurement data of the i-th sensor before the kth time and the kth time of the sensor can be obtained The maximum likelihood estimate of the mean and standard deviation of is:

Considering that k is a finite value, the maximum likelihood estimation of σ _i is modified as:

Step S52, calculating the probability that each measurement data in the basic group of the sensor appears in all the measurement data in the basic group.

According to the probability density function of normal distribution, when k>1, the jth (1≤j≤k) measured value of sensor i The probability of occurrence is:

In the formula, are the k measured values of the i-th sensor before the kth time and the kth time of the sensor mean and variance of .

Then the j-th measurement data of the i-th sensor The probability of occurrence in all measurement data in this group is:

Step S53, correcting the current measurement data in the sensor group to obtain the corrected current measurement data of the sensor.

On the basis of the above steps, the k-th measurement data of the i-th sensor is corrected.

Step S54, calculating inter-sensor fusion weights of the current measurement and corrected data among different sensors.

The k correction values of the i-th sensor before the k-th time and the k-th time As a group, according to formula (14) and formula (15), the mean value and variance of the k correction values of the i-th sensor before the kth time and the kth time can be obtained as:

According to the idea of weighted fusion, if the data of the i- _{th sensor is Xi and its weight is ω i ,} then the result of weighted fusion of all N sensors is:

make Then Z _i obeys the standard normal distribution, and the fusion result in formula (20) can be expressed as:

It can be seen from formula (21) that the fusion result Y obeys the mean variance normal distribution of . Using the method of least squares, when When , the variance of the fusion result Y is the smallest

Step S55 , performing weighted fusion based on the corrected current measurement data of different sensors and the inter-sensor fusion weights to obtain corrected current measurement fusion data.

The k-th weighted fusion is performed by using the corrected data measured by the sensor for the k-th time and the weight of the k-th data fusion between different sensors.

Step S56, judge whether all the measurement data of the sensor have been corrected, if not, take the next uncorrected measurement data of the sensor as the current measurement, and execute step S51; otherwise, execute step S57.

Step S57, combine the fusion data of all measured and corrected data into a primary fusion data set, and calculate the secondary intra-group probability of each data in the primary fusion data set appearing in all the data in the primary fusion data set, based on The primary fusion data group and the secondary intra-group probability are subjected to secondary weighted fusion to obtain a final result of data fusion.

It can be seen from the above that the multi-sensor data weighted fusion method of the present invention first filters and corrects the actual measurement data of each sensor to improve the reliability of the single-sensor measurement data, and secondly obtains data fusion between sensors based on the corrected data. Weight, improve the rationality of data fusion weight distribution, and finally use the correction data and the inter-sensor data fusion weight obtained based on the correction data to carry out weighted fusion of multi-sensor data. Therefore, the multi-sensor fusion result obtained by using the data weighted fusion method of the present invention The closer to the actual situation, the higher the credibility.

Specifically, the present invention also discloses an embodiment of using the multi-sensor data weighted fusion method to perform data fusion on thermostat parameters. Assuming that in a distributed multi-sensor detection system, three thermocouple sensors detect the incubator six times, the table 1 is the six measured values of each sensor on the incubator. Then use the multi-sensor data weighted fusion method of the present invention to include:

(1) Take the current measurement data and historical measurement data of different sensors as a basic data group, and calculate the mean value and variance of the basic data group of the sensors.

For the first measurement:

Since there is no measurement data before the first measurement, the mean value of the first basic data set of sensor S1 is its measured value 899.5, and the variance is 0; the mean value of the first basic data set of sensor S2 is its measured value 898.3000, and its variance is 0; The mean of the first basic data set of sensor S3 is its measured value 896.7000, and the variance is 0; at this time, according to formula (22), the fusion value of the first measurement can be obtained as 898.1667.

For the second measurement:

Since there is the first measurement before the second measurement, the first and second measurement data are taken as the second basic data set, and according to formula (14) and formula (15), the second basic data set of sensor S1 The mean value is 902.4 and the variance is 16.82. The mean value of the second basic data set of sensor S2 is 887.1000 and the variance is 250.8800. The mean value of the second basic data set of sensor S3 is 901.7500 and the variance is 51.0050.

(2) On the basis of step (1), calculate the probability that a certain measurement data in the basic group of the sensor appears in all the measurement data in the basic group.

According to formula (16) and formula (17), the probabilities of the first and second measurement values of sensor S1 appearing in all the data in the second basic data set are respectively and The probabilities of the first and second measurement values of sensor S2 appearing in all data in the second basic data set are 0.5 and 0.5 respectively, and the first and second measurement values of sensor S3 are in all data in the second basic data set The probabilities of appearing in are 0.5 and 0.5, respectively.

(3) On the basis of step (2), correct the current measurement data in the basic group of the sensor.

According to formula (18), it can be obtained that the correction value of the second measurement data of sensor S1 is 899.5×0.5+905.3×0.5=902.4, the correction value of the second measurement data of sensor S2 is 887.1000, and the correction value of the second measurement data of sensor S3 The value is 901.7500.

(4) On the basis of step (3), calculate the fusion weight of the current measurement corrected data among different sensors.

According to formula (20), it can be obtained that the fusion weight of sensor S1's second measurement and correction data is 0.7159, the fusion weight of sensor S2's second measurement and correction data is 0.0480, and the fusion weight of sensor S3's second measurement and correction data is 0.2361.

(5) On the basis of step (4), perform weighted fusion based on the corrected data of the current measurement of different sensors and the fusion weight to obtain the primary fusion data after correction of the current measurement.

According to the formula (22), the fusion result of the corrected data of the second measurement of the three sensors can be obtained as follows:

y(2)=0.7159×902.4+0.048×887.1+0.2361×901.75=901.5122

(6) Repeat steps 1, 2, 3, 4, and 5 until the fusion of all observation data from different sensors is completed.

According to the calculation process of the above steps 1, 2, 3, 4, and 5, the fusion results of the corrected data of the third, fourth, fifth, and sixth measurements of the three sensors can be obtained. Fusion data after correction of 6 measurements.

Table 11 Fusion data after correction of 6 measurements among all sensors

Measurement times 1 2 3 4 5 6 fusion data 898.1667 901.5122 900.8111 900.8085 900.0013 899.7063

(7) On the basis of step (6), use all the fusion data after measurement correction as a primary fusion data set, and calculate the secondary group in which each data in the primary fusion data set appears in all the data in the group Probability, based on the primary fusion data group and the secondary intragroup probability of each data in the group appearing in all the data in the group, perform secondary weighted fusion to obtain the final result of data fusion.

According to formula (14) and formula (15), it can be obtained that the mean value of the first fusion data set after all the measurement corrections is 900.1677, and the variance is 1.3754.

According to formula (16) and formula (17), the secondary intra-group probability that each data in the first fusion data group appears in all data in the group can be obtained, and table 12 is that each data in the first fusion data group is in the group The quadratic within-group probability of occurrence in all data. Wherein in calculating the calculation method of the probability of each data appearing in the group in the primary fusion data group and calculating the probability of each data appearing in all the data in the group in the 6th basic data group in the sensors S1, S2, S3 method is consistent.

Table 12 The secondary intra-group probability of each data in the primary fusion data group appearing in all the data in the group

data 898.1667 901.5122 900.8111 900.8085 900.0013 899.7063 probability 0.0531 0.1181 0.1960 0.1963 0.2256 0.2109

According to formula (18), the final fusion data can be obtained as 900.3372.

In summary, the present invention discloses a multi-sensor data weighted fusion method, which first filters and corrects the actual measurement data of each sensor to improve the reliability of the single-sensor measurement data, and secondly obtains the inter-sensor relationship based on the corrected data. Data fusion weight, improve the rationality of data fusion weight distribution, and finally use the correction data and the inter-sensor data fusion weight obtained based on the correction data to perform weighted fusion of multi-sensor data. Therefore, the multi-sensor data obtained by using the data weighted fusion method of the present invention The fusion result is closer to the actual situation and has higher reliability.

The above are only embodiments of the present invention, and are not intended to limit the patent scope of the present invention. All equivalent structural transformations made by using the description of the present invention and the contents of the accompanying drawings, or directly or indirectly used in other related technical fields, are included in this patent. inventions within the scope of patent protection.

Claims (4)

1. a kind of multi-sensor data Weighted Fusion method, comprising:

Step S1 is filtered amendment to each sensing data in the multiple sensor, obtains the multiple sensor Amendment after data；

Step S2 is calculated in the multiple sensor and is merged weight between the sensor of data fusion between different sensors；

Step S3 carries out secondary weighted fusion based on weight is merged after the amendment between data and the sensor, to obtain most Whole fusion results.

2. multi-sensor data Weighted Fusion method according to claim 1, which is characterized in that the multiple sensor In each sensing data be filtered amendment, data include: after obtaining the amendment of the multiple sensor

Step S21, setting sensor dataFor the measurement data of i-th of sensor jth time in the multiple sensor, In 1≤i≤N, N be the multiple sensor number, 1≤j≤M, M be the multiple sensor in each sensor total survey Measure number；Obtain all master data groups of each sensor in the multiple sensorWherein basic group of j-th of i-th of sensor of B (i, j) expression；

Step S22 calculates the mean value EB (i, j) and variance DB (i, j) of each master data group B (i, j)；

Step S23 calculates what each element in the master data group B (i, j) occurred in the master data group B (i, j) The primary interior probability P B (i, j) of group=[pb (i, 1), pb (i, 2) ..., pb (i, j)]；

Step S24, according to the master data group B (i, j) and it is primary group described in probability P B (i, j) to i-th of sensor jth Secondary measurement dataIt is modified, obtains data X=[X after the amendment of all measurements of all the sensors₁,X₂,...,X_N], Wherein X_i=[x_i(1),x_i(2),…,x_iIt (M)] is data after the corresponding amendment of all measurements of i-th of sensor.

3. multi-sensor data Weighted Fusion method according to claim 2, which is characterized in that the calculating is the multiple Merging weight in sensor between different sensors between the sensor of data fusion includes:

Step S31 calculates the mean value EX of data after the amendment of each sensor in the multiple sensor_iWith variance DX_i；

Step S32 calculates fusion weight W=[w between the sensor of different sensors data fusion₁,w₂,…,w_N], whereinIndicate the data fusion weight of i-th of sensor.

4. multi-sensor data Weighted Fusion method according to claim 3, which is characterized in that described to be based on the amendment Weight is merged between data and the sensor afterwards and carries out secondary weighted fusion, includes: to obtain final fusion results

Step S41 will merge weight and once be melted between data and the sensor after the amendment of the multiple sensor It closes, obtains Single cell fusion data group Z=[z₁,z₂,…,z_M]=WX^T；

Step S42 obtains the mean value EZ and variance DZ of the Single cell fusion data group Z；

Step S43 calculates all fused datas in the Single cell fusion data group Z and occurs in the Single cell fusion data group Z Secondary group in probability P Z=[pz₁,pz₂,…,pz_M]；

Step S44, by the Single cell fusion data group Z and it is secondary group described in probability P Z carry out secondary data fusion, obtain number According to final fusion results Y=(PZ) Z of fusion^T。