Research On Small Target Detection Technology Under The Background Of Sea Clutter

Sea clutter is the backward reflection echo of radar signal from the sea surface,which is affected by complex sea conditions and has a huge obstacle to detect small targets on the sea surface.At the same time,the radar cross-sectional area of small targets is small,the target signal is very weak,and it is easy to be submerged in sea clutter and noise,which leads to the increase of false alarm rate.It has become a hot spot and difficult point in radar signal processing field how to realize accurate and time-sensitive detection of small eye elevation in complex sea conditions.To detect small targets in the background of sea clutter,the characteristics of sea clutter need to be studied and the differences in the characteristics of sea clutter signals and small target signals need to be analysed for small target detection.According to the characteristics of chaos and fractal of sea clutter,the basic theory and characteristics of chaos are studied in this paper.This paper studies the difference between chaos and noise,and proposes a sea clutter denoising algorithm based on MEEMD algorithm and SG filter.This paper not only studies the relationship between chaos and fractal,but also studies the fractal characteristics of sea clutter,and proposes a small target detection method based on multifractal parameters of sea clutter in FRFT domain.The specific research is as follows:A least squares support vector machine(LSSVM)is used to construct a sea clutter single-step prediction model,which confirms the effectiveness of LSSVM applications in the field of sea clutter target detection.The improved MEEMD algorithm,which combines CEEMD with permutation entropy,is selected to decompose sea clutter by CEEMD to avoid modal confusion.Determine the threshold range of permutation entropy,and calculate the permutation entropy of IMF components decomposed by CEEMD each time.If the permutation entropy is greater than the maximum permutation entropy threshold,it is considered to be an abnormal component,and the decomposition will be stopped until the permutation entropy of the decomposed IMF component is less than the minimum permutation entropy threshold.The IMF component within the range of permutation entropy threshold is considered as the signal component dominated by noise,and SG filtering is carried out for such signals.The filtered components are signal reconstructed with the first IMF component that is smaller than the threshold range and the remaining signal to obtain the denoised sea clutter data.MEEMD does not need to decompose the sea clutter completely.As long as the permutation entropy of IMF components is below the threshold,the decomposition can be stopped,which reduces the operation time and improves the timeliness.The LSSVM single-step prediction model,with the root mean square error(RMSE)chosen as the judge of the denoising effect,takes IPIX radar #17 sea clutter as an example,the RMSE before denoising is 0.0064 and after denoising is 0.000968,the experimental result is improved by an order of magnitude,and the small target signals that were not detected before denoising can be detected,improving the accuracy.The effect of complex sea conditions on small target detection in the background of sea clutter,the traditional time-domain fractal analysis method does not consider the low signal-to-clutter ratio(SCR)of sea clutter.A high-order multi-fractal parameter detection method based on FRFT is proposed.Fractional Brownian Motion(FBM)is used to model the sea clutter data and substitute it into FRFT transform.It is found that there is no longer strict self-recognition.In addition to the FRFT transformation order,it is also related to the scale coefficient.The multi-fractal parameter is determined by MF-DFA.The measured sea clutter data are studied in different polarization,different sea conditions and different distance gates' multi-fractal parameters.The simulation results show that FRFT can effectively compensate the energy variables of small targets,and expand the linear range of multi-fractal parameters.HH polarization is increased by 10%,VV polarization is increased by 14.29%.The high-scale fractal parameters are normalized and threshold is set.The detection accuracy of HH polarization is 92.86%,and VV polarization is improved by main target detection.The detection accuracy is 78.57%,which verifies that the high-scale fractal parameters can be applied to the detection of small targets on the sea surface.The slope of fitting curve with normalized negative high-scale multifractal parameters and fractal parameters in the range of-5 to 5.Logistic regression model was used to predict whether there were small targets.The accuracy of prediction was 83.42%.By analyzing the characteristic difference between sea clutter and small target signal,combining the theory of empirical mode decomposition,support vector machine,fractal and fractional Fourier transform,a small target detection model of sea clutter background is established,which can alleviate the dependence of small target detection on complex sea condition,and has certain application value.