 Open Access
 Authors : Mansour Aljohani , Raed Althomali
 Paper ID : IJERTV10IS080233
 Volume & Issue : Volume 10, Issue 08 (August 2021)
 Published (First Online): 02092021
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Target Detection using Marine Radar Data and A Maximum Likelihood Estimator
Target Detection using Marine Radar Data and A Maximum Likelihood Estimator
Mansour Aljohani
Electrical and Electronic Technology Department Yanbu Industrial College, Royal Commission Yanbu Industrial City, Saudi Arabia
Raed Althomali
Electrical and Electronic Technology Department Yanbu Industrial College, Royal Commission Yanbu Industrial City, Saudi Arabia
Abstract Marine radar data has been applied to the detection algorithm and signal preprocessing. Analyzing the statistical output where the signal to clutter ratio has been reduced by using the Maximum likelihood estimation (MLE) algorithm. Some other estimation methods have been applied to the marine radar row data.
KeywordsMarine Radar; Likelihood; DSP

INTRODUCTION
An Xband radar Marine radar is a reliable and durable kind of radar. It is the overshelf radar at a low cost. Marine radar is a highpower radar with variable pulse width. These types of radars contain magnetron to generate a sequence of highpower pulses. These pulses are random phases, which constrain the radar to generate a high resolution image. The whole coherent process is describing in detail in [1].
Phases unwrap method (coarse to fine or C2F) proposed in [2] [6] which transform the shorter baseline data to longest baseline data. The advantage of C2F methods is to sustain the signal information keeping the longest baseline unwrapped phase, and the other baseline information will be ignored.
The Maximumlikelihood [7] [10] method computes the phase of the marine radar data, which is a single look complex (SLC) based on the model. By computing the phase, the noise and reliability level will be improved from the radar data. Weighted least squares [11] [16] and least squares and other methods are helpful to compute the unwrapped phase.
The main purpose of this article is to improve the radar data by reducing the noise level using the Maximum likelihood method. Real radar data has been processed using MATLAB in a highperformance computer.

MEASUREMENT SETUP
The rotating antenna has been replaced by a fixed antenna at the top of the radar. The antenna pointed to the target, which is one steel rod. The target in the turntable rotating 360 degrees. Each rotates the radar sends a transmitter signal and receives it back. The received signal will be downconverting using an RF component. The RF components are connected to a highperformance computer to process the data and record it. The data needs a high
capacity storage system. The radar system parameter is showing in table 1.
TABLE 1. MARINE RADAR SYSTEM PARAMETER
Marine Radar Parameters
Parameters
Values
Opereating Frequency
9.14 GHz
PRF
3000 Hz
Bandwidth
50 KHz
Dutiy Cycle
12%
CPI
9.12 s
Transmitter Power
25 kW

SIGNAL PREPROCESSING
After receiving the signal by the monostatic marine radar, the signal downconverted from 9.14 GHz to 100 Hz by RF component (Mixer). Then the 3D data matrix is created by MATLAB for further processes. Transmit signal acquired from signal leakage to perform the cross correlation process between the transmit and receive signal.
Figure 1 depicts the result of the crosscorrelation between the transmit and received signal without any filtering process. Then to reduce the clutter from the antenna we averaged the 3D radar data cube.
A low pass filter is applied to each signal to reduce the highfrequency component from the signal. The characteristic of the low pass filter is the cutoff frequency at 100 Hz. So, it will pass what is below 100 Hz and block what is above 100 Hz.
Before storing the data, the data has been tested using Signatech analogtodigital converter (ADC) and then the output of the ADC which is a digital signal is the input of a highperformance Keysight digital oscilloscope. the connection between the ADC and the oscilloscope was a long highly sensitive coaxial cable. So, between each coaxial cable, we add an amplifier to reduce or eliminate the signal losses or distortion.

DETECTION USING (MLE) ALGORITHM The Marine radar data has been tested in two situations:
First: the data acquired for noise and clutter only and the absence of the target.
Second: the data acquired the target, noise, and clutter.
The MLE algorithm is to estimate the content of the noise covariance and mean clutter for both situations as seen in figures 14, 15, and 16. We used the first situation, in which the target is not present, as reference data see figure 4.

STATISTICAL ANALYSIS OF THE LOGMLE DECISION VARIABLE PROPERTIES
The analysis of the real marine radar data demonstrates that the probability density function (pdf) is aligned with the variable of the decision history see figure 2.
While a number of the trial goes to infinite, the histogram looks like Gaussian and the curve is going to be smoother, and maximum likelihood estimate converge to the linear line. The estimate converges in probability variance is inversely proportional to N. The Max Likelihood Estimate is consistent as depicted in figures 9, 10, and 11.
It can be seen from figure 13, variance is inversely proportional to N, where N is the sample number, and the Variance of the estimate is decreasing when N is increased. The mean of radar data can be seen in figures 7, 8, and 12.
Figures 3and 5 shows the estimated value for A and B is very close to the real value of A and B. They are unbiased. There is a gap Maximum likelihood estimate and Cramer Rao Bound (CRB). The variance of an unbiased estimate is greater than the maximum likelihood estimate.
The CRB of the estimate of the variance corresponding to the probability density function for a sample of N independent identically distributed normal random variables.
The loglikelihood function associated with the MLE for the variance MLE is corresponding to the probability density function for a sample of N independent identically distributed normal random variables. MLE parameter estimation for the slope can be applied using vector parameter estimation and the Fisher information matrix.
CRB for the variance associated with the maximum likelihood estimated by using a noise variance of 100. Figure 6 shows the performance for various values of N from 2 to 200 and the results for each as a function of N. MLE of the parameters turn out to be the same as the least square solution (assuming the errors are Gaussian). Under Gaussian distribution assumption for the noise and the clutter could be driven theoretically when N is a small number.
Then after applying the theoretical part for the Gaussian assumption with different thresholds and different false alarm rates to the experimental data, the target detection can be seen in our data as portrayed in Figure 17.

CONCLUSION
The maximumlikelihood algorithm is used to improve the detection of the target. There are two situations of acquiring the radar data in this experiment. The purpose of these situations is to compute the mean clutter and the noise covariance. The probability of detection was computed to check the radar receiver operating characteristics (ROC).
The analyss of the pdf for the radar data describes that there is a huge difference in the histograms. The histograms spikes in the plot assumed to be the signal of the target is present in the examination history.
The result shows the difference between the maximum likelihood estimate and Cramer Rao Bound. There is improving in the results as we can see it in the figures of the real marine radar data.
REFERENCES

H. Abdelbagi, M. Aljohani, A. Mrebit and M. C. Wicks, "FPGA based coherent Doppler processor for marine radar applications," 2015 National Aerospace and Electronics Conference (NAECON), 2015, pp. 5869, doi: 10.1109/NAECON.2015.7443040.

H. Godrich, A. Petropulu and H. V. Poor, "Power allocation schemes for target localization in widely distributed MIMO radar systems," 2010 – MILCOM 2010 MILITARY COMMUNICATIONS CONFERENCE, 2010, pp. 846851, doi: 10.1109/MILCOM.2010.5680193.

Li, H. wan and Z. Wang, "Accurate Measurement of Radar Carrier Frequency," 2007 International Conference on Microwave and Millimeter Wave Technology, 2007, pp. 14, doi: 10.1109/ICMMT.2007.381513.

J. Abatzoglou, "Maximum Likelihood Estimation of Range of Polynomial Amplitude Modulated Complex Scatterers," 2006 Fortieth Asilomar Conference on Signals, Systems and Computers, 2006, pp. 12171221, doi: 10.1109/ACSSC.2006.354949.

Junghoon Kim and Joohwan Chun, "Superresolution range estimation in FMCW radar system," 2016 Progress in Electromagnetic Research Symposium (PIERS), 2016, pp. 2558 2562, doi: 10.1109/PIERS.2016.7735044.

J. Abatzoglou and B. C. Steakley, "Comparison of maximum likelihood estimators of mean wind velocity from radar/lidar returns," Conference Record of The Thirtieth Asilomar Conference on Signals, Systems and Computers, 1996, pp. 156160 vol.1, doi: 10.1109/ACSSC.1996.600848.

Jiang, J. Huang and Y. Hou, "MIMO radar joint estimation of target location and velocity with multiple subcarrier signals," 2010 IEEE International Symposium on Phased Array Systems and Technology, 2010, pp. 398402, doi: 10.1109/ARRAY.2010.5613338.

H. Godrich, A. Petropulu and H. V. Poor, "Resource allocation schemes for target localization in distributed multiple radar architectures," 2010 18th European Signal Processing Conference, 2010, pp. 12391243.

A. M. Zelnio and B. D. Rigling, "Detectionbased localization of passive radar receivers," 2012 IEEE Radar Conference, 2012, pp. 01730177, doi: 10.1109/RADAR.2012.6212132.

M. J. Sheinvald, M. Wax and A. J. Weiss, "On maximumlikelihood localization of coherent signals," in IEEE Transactions on Signal Processing, vol. 44, no. 10, pp. 24752482, Oct. 1996, doi: 10.1109/78.539032.

J. N. Jiang and N. Zhang, "Expectation MaximizationBased Target Localization From Range Measurements in Multiplicative Noise Environments," in IEEE Communications Letters, vol. 25, no. 5, pp. 15241528, May 2021, doi: 10.1109/LCOMM.2021.3050455.

I. O. Besson and F. Galtier, "Estimating particles velocity from laser measurements: maximum likelihood and CramerRao bounds," in IEEE Transactions on Signal Processing, vol. 44, no. 12, pp. 30563068, Dec. 1996, doi: 10.1109/78.553479.

H. Lee and J. Chun, "Virtual Array Response Vector for Angle Estimation of MIMO Radar With a WideBand Interleaved OFDM Signal," in IEEE Communications Letters, vol. 25, no. 5, pp. 1539 1543, May 2021, doi: 10.1109/LCOMM.2021.3049355.

J. C. Ye, Y. Bresler and P. Moulin, "CramerRao bounds for 2D target shape estimation in nonlinear inverse scattering problems with application to passive radar," in IEEE Transactions on Antennas and Propagation, vol. 49, no. 5, pp. 771783, May 2001, doi: 10.1109/8.929632.

X. Zhang, F. Wang, H. Li and B. Himed, "Maximum Likelihood and IRLS Based Moving Source Localization With Distributed Sensors," in IEEE Transactions on Aerospace and Electronic Systems, vol. 57, no. 1, pp. 448461, Feb. 2021, doi: 10.1109/TAES.2020.3021809.

T. J. Abatzoglou and G. O. Gheen, "Range, radial velocity, and acceleration MLE using radar LFM pulse train," in IEEE Transactions on Aerospace and Electronic Systems, vol. 34, no. 4, pp. 10701083, Oct. 1998, doi: 10.1109/7.722676
Fig. 1. CrossCorrelation Result before the filtering
Nonfluctuating Coherent
Receiver Operating Characteristic (ROC) Curves
1
0.9 SNR=9.4dB
0.8
SNR=8dB
0.7
0.6 SNR=4dB
0.5 SNR=2dB
0.4
0.3
0.2
0.1
0
1010 108 106 104 102 100
Pfa
Fig. 2. PDF of the radra data.
Fig. 3. The variance of estimate B for various N
Fig. 4. Maximum Likelihood of the data.
Fig. 5. The variance of estimate A for various N
Fig. 6. Comparison to a theoretical estimate of variance.
Fig. 7. Mean of estimate A for various N
Fig. 8. Mean of estimate B for various N
Fig. 9. Histogram of estimates of variance for 200 Samples
Fig. 10. Histogram of estimates of variance for 1000 Samples
Fig. 11. Histogram of estimates of variance for 10000 Samples.
Fig. 12. Mean of estimate various for N
Fig. 14. Line Solution, N=10
Fig. 13. The variance of estimate for various N
Fig. 15. Line solution for, N=100.
Fig. 16. Line solution for, N=1000.
Fig. 17. Crosscorrelation after applying the Maximum Likelihood algorithm.