CN107122764B - Maritime ship target detection method based on KpN model - Google Patents

Abstract

The invention provides a KpN model-based ship target detection method at sea. The technical solution is: Statistically model the obtained SAR image using KpN distribution and use the logarithmic cumulant of the SAR image to estimate the parameters of the KpN model, calculate the CFAR detection threshold according to the estimated value of the KpN model parameters, and use CFAR detection to realize the Detection of ship targets at sea. The invention can realize more accurate estimation of the shape parameter, scale parameter and noise power in the KpN model, and enhances the detection performance of the ship target at sea. At the same time, the invention does not need to set additional parameters or conditions, and is simple and easy to implement.

Description

Maritime ship target detection method based on KpN model

technical field

The invention belongs to the technical field of SAR (synthetic aperture radar, synthetic aperture radar), and relates to a method for detecting a ship target at sea based on a KpN model.

Background technique

Maritime ship target detection is an important field of SAR application. It has a wide range of applications in the fields of military intelligence surveillance, illegal immigration supervision, and large-scale marine traffic supervision. CFAR (constant false alarm rate, constant false alarm rate) detection is currently the most commonly used detection method for maritime ship targets. The core of CFAR detection lies in sea clutter modeling. At present, the commonly used sea clutter models are mainly K distribution model and G ⁰ distribution model, but these two models do not take into account the influence of channel noise. The KpN (K plus Noise, K plus noise) model adds the influence of channel noise on the basis of the K distribution model, which can more accurately fit sea clutter (references: KDWard and RJATough, "Radar detection performance in sea clutter and discrete spikes," Radar, 2002, pp. 253-257). However, the estimation accuracy of the KpN model parameters in the existing methods is not high enough, which limits the role of the KpN model in ship target detection at sea to a certain extent.

Contents of the invention

The invention provides a KpN model-based ship target detection method at sea. The method adopts KpN distribution to statistically model SAR images and uses logarithmic cumulants to estimate the parameters of the KpN model, and realizes the detection of ship targets at sea.

Technical scheme of the present invention is:

The obtained SAR image is statistically modeled using the KpN distribution and the parameters of the KpN model are estimated using the logarithmic cumulant of the SAR image, and the CFAR (constant false alarm rate, constant false alarm rate) detection is calculated according to the estimated value of the KpN model parameters Threshold, using CFAR detection to realize the detection of marine ship targets. Among them, the estimated value of the shape parameter v in the KpN model is obtained by solving the following formula Estimated value of the scale parameter b and an estimate of the noise power p _n

Among them, Ψ() is the psi function, Ψ(,) is the polygamma function, N is the equivalent visual number, and the specific expressions of the parameters A, B, C, and D are shown in the following formula:

Moreover, the specific calculation formula of the CFAR detection threshold T is as follows:

in is the probability density function of the KpN model; P _fa represents the false alarm rate, which is usually set artificially according to actual needs.

The beneficial effects of the present invention are:

1. Compared with existing methods, using logarithmic cumulants to estimate KpN parameters can achieve more accurate estimation of shape parameters, scale parameters and noise power in the KpN model, and enhance the detection performance of marine ship targets.

2. The KpN parameter estimation method using the logarithmic cumulant proposed by the present invention does not need to set additional parameters or conditions, and is simple and easy to implement.

Description of drawings

Fig. 1 is a flowchart of the present invention;

Fig. 2 is experimental data of the present invention;

Fig. 3 is the experimental result figure of the present invention;

Figure 4, Figure 5, and Figure 6 are the results of theoretical verification.

Detailed ways

Fig. 1 is a flowchart of the present invention, and concrete implementation steps are as follows:

The obtained SAR image is statistically modeled using the KpN distribution and the parameters of the KpN model are estimated using the logarithmic cumulant: that is, the SAR image is first considered to be in line with the KpN model distribution, and then the parameters of the KpN model are estimated, which can include the following two steps:

The first step is to calculate the image logarithmic cumulant according to the original SAR image, and its calculation method is shown in formula 1:

in represents the first-order image logarithmic cumulant, represents the second-order image logarithmic cumulant, Indicates the logarithmic cumulant of the third-order image, M indicates the total number of pixels in the image, x _i is the gray value of the i-th pixel in the image, i∈[1,M].

In the second step, the estimated value of the shape parameter v in the KpN model is obtained by numerically solving the formula 2 Estimated value of the scale parameter b and an estimate of the noise power p _n Its specific expression is as follows.

Among them, Ψ() is the psi function, Ψ(,) is the polygamma function, N is the equivalent visual number, and the specific expressions of the parameters A, B, C, and D are shown in formula three:

Calculate the CFAR detection threshold according to the estimated value of the KpN model parameters, and use the CFAR detection to realize the detection of the ship target at sea, that is, to realize the third step below:

In the third step, use the estimated value of the shape parameter obtained in the second step Estimated value of the scale parameter and an estimate of the noise power Calculate the CFAR detection threshold T. The specific calculation formula of CFAR detection threshold T is as follows:

in is the probability density function of the KpN model; P _fa represents the false alarm rate, which is usually set artificially according to actual needs.

The original SAR image is detected, and when the gray value of the detected pixel is greater than or equal to T, it is determined to be a ship target pixel. Otherwise, it is judged as a background pixel to realize the detection of the ship target at sea.

The experimental data of the present invention is the original SAR image. Figure 2 is the original SAR image, where the abscissa represents the azimuth direction, and the ordinate represents the distance direction, and the white pixels in the image are the ship targets to be detected. Fig. 3 is a diagram of detection results of ship targets at sea by using the present invention, the abscissa of Fig. 3 represents the azimuth direction, the ordinate represents the distance direction, and the white rectangular frame in the figure represents the detected ship target. Comparing Fig. 2 with Fig. 3, it can be seen that all 13 marine ship targets are detected well without false alarms, which verifies the effectiveness of the method of the present invention.

In order to further verify the validity of the logarithmic cumulant for parameter estimation of the KpN model, Matlab is used to generate random numbers that obey the KpN model. Fig. 4, Fig. 5, and Fig. 6 are diagrams comparing the results of the parameter estimation of the KpN model in the invention with the experimental results of the other two KpN model parameter estimation methods. Among them, Figure 4 is a comparison chart of the estimation results of the shape parameter v by the three methods, Figure 5 is a comparison chart of the estimation results of the noise parameter p _n by the three methods, and Figure 6 is a comparison chart of the estimation results of the scale parameter b by the three methods , Figure 4, Figure 5, and Figure 6. The abscissas represent the number of experiments, and the ordinates represent the estimated mean square error. The curve with a circle corresponds to the Home parameter estimation method, and the curve with the rice character corresponds to zlog(z) Parameter estimation method, the curve with squares corresponds to the parameter estimation method of the present invention. It can be found through observation that the mean square error estimated by the method of the present invention is smaller than that of the other two methods, which shows that the estimation accuracy of the method of the present invention is higher.

The probability density function of the KpN model that the present invention utilizes in Fig. 4 experiment is as shown in formula five:

Using Equation 5, the Mellin transformation φ _Z (s) expression of the KpN model can be obtained as follows:

Where W, () represents the Whittaker function.

According to formula 6, the expression of the second second type characteristic function ξ _Z (s) of the KpN model can be further obtained as follows:

Where U() represents the Tricomi function.

The expression of the theoretical logarithmic cumulant of the KpN model can be obtained through formula 7 as follows:

in Represents the i-th order theoretical logarithmic cumulant, and combining Formula 1 and Formula 8, the estimated expression such as Formula 2 can be obtained. In addition, other forms of probability density functions of the KpN model can also be used for the calculation of the present invention, without affecting the actual effect of the present invention.

Claims (1)

1. a sea ship target detection method based on KpN model, is characterized in that, comprises the steps: The KpN distribution is used for statistical modeling of the obtained synthetic aperture radar image, and the parameters of the KpN model are estimated by the logarithmic cumulant of the synthetic aperture radar image. The constant false alarm rate detection threshold is calculated according to the estimated value of the KpN model parameters. False alarm rate detection realizes the detection of ship targets at sea; Among them, the estimated value of the shape parameter v in the KpN model is obtained by solving the following formula Estimated value of the scale parameter b and an estimate of the noise power p _n Among them, Ψ() is the psi function, Ψ(,) is the polygamma function, N is the equivalent visual number, M represents the total number of pixels in the image, and the specific expressions of parameters A, B, C, D are shown in the following formula : Moreover, the specific calculation formula of the CFAR detection threshold T is as follows: in is the probability density function of the KpN model; P _fa represents the false alarm rate, which is artificially set according to actual needs, and the KpN model refers to the K plus noise model.