Insect multi-band two-dimensional body axis orientation fusion method based on orientation distribution prior

Abstract

The application discloses an insect multi-frequency band two-dimensional body axis orientation fusion method based on an orientation distribution prior, and the basic idea is as follows: firstly, the two-dimensional orientation of the insect is estimated based on a single-frequency band polarization scattering matrix; secondly, the distribution of the single-frequency estimation result of the two-dimensional orientation of the insect is simulated and analyzed under the condition that the radar measurement noise obeys Gaussian distribution, and the result shows that the single-frequency estimation result of the two-dimensional orientation obeys Gaussian distribution with the mean value being the true value of the orientation; finally, the maximum likelihood estimation of the two-dimensional orientation based on the two-dimensional orientation measurement result of the multi-frequency band is derived based on the Gaussian distribution of the single-frequency two-dimensional orientation estimation, and the effective fusion of the two-dimensional orientation estimation result of the multi-frequency band is realized.

Description

Technical Field

[0001] This invention belongs to the field of insect radar technology, specifically relating to a method for fusion of two-dimensional body axis orientation of insects based on orientation distribution priors in multiple frequency bands. Background Technology

[0002] Insect radar is an effective tool for studying migratory insects. Based on the measurement results of insect radar, parameters such as insect body length, weight, body width, wingbeat frequency, horizontal speed, and orientation can be retrieved. Based on the retrieved insect parameters, species identification and trajectory prediction of migratory insects can be performed, which is of great significance for studying the behavior of migratory insects and preventing migratory pests and diseases.

[0003] The orientation of an insect's body axis is an important behavioral parameter for migratory insects, and accurate orientation estimation is of great significance for the precise prediction of the trajectory of migratory insects.

[0004] Traditional methods use single-band polarization information measured by radar to estimate the two-dimensional orientation of insects. For insect radars capable of measuring multi-band polarization information, this method suffers from information waste. Furthermore, using only single-band polarization information for orientation estimation results in low accuracy and poor robustness. Effectively fusing multiple orientation measurements from radar across multiple bands can improve the accuracy and robustness of orientation estimation. Therefore, there is an urgent need to develop a multi-band two-dimensional body axis orientation fusion method for insects. Summary of the Invention

[0005] In view of this, the present invention provides a method for fusing the two-dimensional body axis orientation of insects in multiple frequency bands based on orientation distribution priors, which can achieve high-precision and robust estimation of the two-dimensional orientation of insects, which helps to achieve accurate prediction of the trajectory of migratory pests.

[0006] This invention is a method for fusing the two-dimensional body axis orientation of insects based on prior knowledge of orientation distribution, providing a high-precision and robust estimation method for the two-dimensional orientation of insects. Specific steps include:

[0007] Step 1: Estimate the two-dimensional orientation of insects based on the single-band polarization scattering matrix;

[0008] Step 2: Simulation of the distribution of two-dimensional orientation estimation results for single-band insects

[0009] Simulation analysis was conducted on the distribution of single-frequency estimation results of two-dimensional orientation of insects under the condition that the radar measurement noise follows a Gaussian distribution. The results show that the single-frequency estimation results of two-dimensional orientation follow a Gaussian distribution with the mean being the true orientation value.

[0010] Step 3: Fusion of multi-band two-dimensional orientation estimation results

[0011] Based on the Gaussian distribution of single-frequency two-dimensional orientation estimation, the maximum likelihood estimation of two-dimensional orientation based on multi-frequency two-dimensional orientation measurement results is derived, thereby achieving effective fusion of multi-frequency two-dimensional orientation estimation results. Attached Figure Description

[0012] Figure 1 The distribution of single-frequency estimation results for the two-dimensional orientation of insects under the condition that the radar measurement noise follows a Gaussian distribution in the simulation;

[0013] Figure 2 This is a comparison of the method of this invention and conventional methods for processing radar measured data. Figure 2 (a) and Figure 2 (b) The solid dots represent the two-dimensional orientation estimates of the traditional single-frequency two-dimensional orientation estimation method at 9.5 GHz and 11.5 GHz, respectively. The hollow circles are the fusion results of the two frequency band orientations obtained by the method of the present invention, and the straight line is the true value of the two-dimensional body axis orientation of the insect. Detailed Implementation

[0014] This invention provides a method for fusing the two-dimensional body axis orientation of insects across multiple frequency bands based on prior orientation distribution. The basic idea is as follows: First, estimate the two-dimensional orientation of the insect based on a single-band polarization scattering matrix. Second, simulate and analyze the distribution of the single-frequency estimation results of the two-dimensional orientation when the radar measurement noise follows a Gaussian distribution. The results show that the single-frequency estimation results of the two-dimensional orientation follow a Gaussian distribution with the mean being the true orientation value. Finally, based on the Gaussian distribution of the single-frequency two-dimensional orientation estimation, derive the maximum likelihood estimation of the two-dimensional orientation based on the multi-band two-dimensional orientation measurement results, thus achieving effective fusion of the multi-band two-dimensional orientation estimation results.

[0015] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0016] Step 1: Estimating the two-dimensional orientation of insects based on the single-band polarization scattering matrix

[0017] A fully polarimetric radar can directly measure the echoes from the target's four polarization channels: HH, HV, VH, and VV. After complex signal and data processing, the target polarization scattering matrix (PSM) is obtained. Assuming the insect's two-dimensional body axis is oriented as θ, its PSM can be represented as S. θ :

[0018]

[0019] Rotate it The result was:

[0020]

[0021] in

[0022]

[0023] Because insects are symmetrical targets, when hour, It is a diagonal matrix, that is, it exists:

[0024]

[0025] Therefore, the orientation of the insect's two-dimensional body axis can be estimated:

[0026]

[0027] because So It contains the period of two tangent functions, therefore, There are two values, which correspond to the direction of the insect's two-dimensional body axis and the direction perpendicular to the two-dimensional body axis, respectively. Their discrimination can be achieved according to the method provided in the patent ("A Method for Distinguishing Parallel and Perpendicular Insects Based on Feature Phase", Patent No.: ZL201911203473.0).

[0028] Step 2: Simulation of the distribution of two-dimensional orientation estimation results for single-band insects

[0029] In actual radar measurements, due to the influence of noise, the insect PSM measured by insect radar can be expressed as:

[0030]

[0031] Among them, S θ It is the true scattering matrix of the insect, and N is a 2*2 complex noise matrix.

[0032] Due to the presence of noise, based on the noisy scattering matrix M θ Estimating the orientation of an insect's two-dimensional body axis is not accurate; rather, the results are affected by measurement noise and exhibit a certain distribution.

[0033] Assuming the noise in all four polarization channels of the radar follows a Gaussian distribution, the probability distribution of the estimated two-dimensional body axis orientation of an insect can be analyzed through simulation. The specific simulation steps are as follows:

[0034] Step 1: Rotate the scattering matrix of an insect measured in a microwave anechoic chamber to obtain four true scattering matrices when its true two-dimensional body axis orientation is 0, 10, 20, and 30 degrees respectively;

[0035] Step 2: Randomly generate a 2*2 complex Gaussian white noise matrix with fixed power and add it to the four real scattering matrices respectively. Based on the four scattering matrices with added noise, calculate the four two-dimensional orientation estimates according to formula (6).

[0036] Step 3: Repeat step 20,000 times to obtain the probability density distribution of the estimated value of the insect's two-dimensional body axis orientation under four conditions.

[0037] like Figure 1 As shown, the probability density distribution of the estimated orientation in four cases is presented (bar chart in the figure). The observation results show that the distribution exhibits a clear Gaussian shape, and the center (mean) of the distribution is the true value of the two-dimensional body axis orientation. Therefore, the Gaussian function was used to fit the four distributions respectively (curves in the figure). In all four cases, the goodness of fit exceeded 0.98 (goodness of fit is used to measure the fitting effect, and the distribution range is 0 to 1. The closer it is to 1, the better the fitting effect). This indicates that when the radar measurement noise is Gaussian distributed, the measured two-dimensional orientation of the insect follows a Gaussian distribution with the mean being the true value of the orientation.

[0038] Step 3: Fusion of multi-band two-dimensional orientation estimation results

[0039] The above analysis shows that, under the condition that the radar measurement noise is Gaussian distributed, the orientation of the insect's two-dimensional body axis estimated by formula (6) follows a Gaussian distribution with the mean being the true orientation value.

[0040] Assuming that the insect radar can simultaneously measure the insect polarization scattering matrix in K frequency bands, then K two-dimensional orientation estimation results can be obtained according to formula (6). Let z be the two-dimensional orientation estimation value of the k-th frequency band. k It follows the following Gaussian distribution:

[0041]

[0042] Where θ represents the orientation truth value. Let Variance be the variance.

[0043] Then z k The probability density distribution can be expressed as:

[0044]

[0045] Let z = [z1 z2 … z k ] T Then the probability density distribution of z can be expressed as:

[0046]

[0047] Taking the logarithm of the above expression yields

[0048]

[0049] Taking the derivative with respect to θ, we get

[0050]

[0051] Setting the above equation to 0, we obtain the maximum likelihood estimate of the two-dimensional volume axis orientation as follows:

[0052]

[0053] This enables the fusion of multi-band two-dimensional orientation.

[0054] In practical applications, it is assumed that the radar can measure N scattering matrices in a single frequency band, obtaining N two-dimensional orientation estimates. The nth estimate is θ. n Then the orientation estimate and orientation variance for this frequency band are:

[0055]

[0056] To verify the aforementioned insect multi-band two-dimensional body axis orientation fusion method based on orientation distribution prior, and to improve the accuracy and robustness of orientation estimation after fusion of single-band orientation measured by traditional methods, an insect radar with multi-band PSM measurement capability was used in the field to measure the two-dimensional body axis orientation of an insect under different servo azimuths. This method was then compared with a traditional single-band orientation estimation method ("A High-Precision Insect Body Axis Orientation Extraction Method Based on Polarization Scattering Matrix Estimation," Patent No.: ZL201710137290.8). Figure 2 As shown, the two-dimensional orientation of insects was measured in two frequency bands (9.5 GHz and 11.5 GHz) using traditional methods. Figure 2 (a) and Figure 2 (b) As shown by the solid dots, the fusion result of the two frequency band orientations obtained by the method of the present invention is: Figure 2 As shown in the hollow circle diagram, the straight line represents the true value of the insect's two-dimensional body axis orientation. It can be seen that the orientation fusion result obtained by this method for the two frequency bands is closer to the true value. The root mean square error (RMSE) of two-dimensional orientation estimation for different methods under all servo azimuth angles was statistically analyzed. The results show that the traditional method has an RMS error of 3.43 degrees and 3.44 degrees at 9.5 GHz and 11.5 GHz, respectively, while the method of this invention has an RMS error of 2.68 degrees. Therefore, this verifies the performance improvement of this invention compared to the traditional method.

[0057] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for fusing the orientation of insects in two-dimensional multi-band body axes based on orientation distribution priors, characterized in that, Includes the following steps: Step 1: Estimate the two-dimensional orientation of insects based on the single-band polarization scattering matrix; Step 2: Simulation of the distribution of two-dimensional orientation estimation results for single-band insects Simulation analysis was conducted on the distribution of single-frequency estimation results of two-dimensional orientation of insects under the condition that the radar measurement noise follows a Gaussian distribution. The results show that the single-frequency estimation results of two-dimensional orientation follow a Gaussian distribution with the mean being the true orientation value. Step 3: Fusion of multi-band two-dimensional orientation estimation results Based on the Gaussian distribution of single-frequency two-dimensional orientation estimation, a maximum likelihood estimate of two-dimensional orientation based on multi-band two-dimensional orientation measurement results is derived, achieving effective fusion of multi-band two-dimensional orientation estimation results. The method for fusing multi-band two-dimensional orientation estimation results is as follows: assuming that the insect radar can simultaneously measure the insect polarization scattering matrix of K frequency bands, according to the formula... We obtain K two-dimensional orientation estimation results. Let's assume the two-dimensional orientation estimation value of the k-th frequency band is... It follows the following Gaussian distribution: ； in, To move toward the truth value, For variance; but The probability density distribution is expressed as: ； make ,So The probability density distribution is expressed as: ； Taking the logarithm of the above expression, we get: ； Again Differentiating, we get: ； Setting the above equation to 0, we obtain the maximum likelihood estimate of the two-dimensional volume axis orientation as follows: ； This enables the fusion of multi-band two-dimensional orientation.

2. The method for fusing the orientation of insects in two-dimensional multi-band body axes based on orientation distribution priors as described in claim 1, characterized in that: In step two, the insect PSM actually measured by the insect radar is expressed as follows: ； in, It is the true scattering matrix of the insect, and N is a 2*2 complex noise matrix.

3. The method for fusing the orientation of insects in two-dimensional multi-band bodies based on orientation distribution priors as described in claim 1, characterized in that: In step two, the specific simulation steps for analyzing the probability distribution of the estimated two-dimensional body axis orientation of the insect are as follows: Step 1: Rotate the scattering matrix of an insect measured in a microwave anechoic chamber to obtain four true scattering matrices when its true two-dimensional body axis orientation is 0, 10, 20, and 30 degrees respectively; Step 2: Randomly generate a 2*2 complex Gaussian white noise matrix with fixed power and add it to the four real scattering matrices respectively. Based on the four scattering matrices with added noise, according to the formula... Calculate the four two-dimensional orientation estimates; Step 3: Repeat step 20,000 times to obtain the probability density distribution of the estimated value of the insect's two-dimensional body axis orientation under four conditions.

4. The method for fusing the orientation of insects in two-dimensional multi-band body axes based on orientation distribution priors as described in claim 1, characterized in that: In step three, it is assumed that the radar can measure N scattering matrices in a single frequency band, obtaining N two-dimensional orientation estimates; where, the first... n The estimated value is Then the orientation estimate and orientation variance for this frequency band are: ； 。

Images