Target Detection And Clutter Rejection Methods In Heavy-tailed Clutter Background

It has always been a hot issue in the field of radar signal processing to obtain the target of interest in clutter background.With the more complex radar detection environment and more advanced radar equipment,the statistical models of radar clutter tend to be diversified.In practice,the statistical histogram of many measured clutter data shows heavy tailing,and the traditional detector designed for Gaussian clutter model is no longer applicable.In addition,in some complex clutter environments,due to the existence of isolated clutter,clutter false alarms inevitably appear in the detection results.After detection,it is necessary to design a reasonable rejection algorithm to eliminate the clutter false alarms.Based on the introduction and analysis of various common radar clutter models and detectors,the problems of target detection and clutter rejection in heavy-tailed clutter are studied in this paper.The main contents are as follows:1.For two special models of alpha-stable distribution clutter model,the positive alpha-stable(P?S)distribution model and the sub-Gaussian symmetric alpha-stable(SGS?S)distribution model,two kinds of detection algorithms are deduced and analyzed.Because the probability density function(PDF)and cumulative probability density function(CDF)of P?S SGS?S model does not have a closed-form expression with respect to elementary function,the research on detection method for these two clutter model is limited.(1)For the P?S clutter model: This model has been proved to be able to model the power of some measured clutter data well.Firstly,with the help of H-function,the closed-form expression of the CDF of P?S model with respect to H-function is derived in Chapter 3.Then the false alarm and detection probabilities expressions of Greater Of,Small Of,Order Statistic and Censored Mean Level detectors are derived,the constant false alarm rate characteristics of these detectors in P?S clutter background are explored,and the their detection performances are analyzed and compared.(2)For the SGS?S clutter model: This model is suitable for the modeling of coherent radar clutter vector.With the help of H-function,the closed-form expression of PDF of SGS?S model with respect to the H-function is derived.On this basis,a detector based on two-step generalized likelihood ratio test(GLRT)for SGS?S clutter background,which is called the GLRT-SGS?S detector is proposed.The experimental results show that the GLRT-SGS?S detector has better detection performance in SGS?S clutter background compared with other traditional detectors.2.Due to the small size of the range resolution cell of high resolution radar,the translational motion of the target in the radial direction of the radar may lead to the problem that the target signal moves across the range cells between multiple pulses.In addition,the rotation of the target may also lead to the expansion of the target signal in the Doppler domain.At this time,the traditional rank-1 target signal model cannot accurately describe the actual target signal model.To solve this problem,a range-spread target detection method is proposed in the background of compound-Gaussian clutter with inverse gamma texture in Chapter 4.In this method,the long-term coherent accumulation process is divided into several short-term sub coherent accumulation processes and then the GLRT-LTD is used to obtain the output of each sub coherent accumulation process,which overcomes the problem of target model mismatch;Meanwhile,the high-order cross-correlation accumulation method is used to realize long-term accumulation on the basis of motion compensation of the target,so as to obtain better detection performance.Experiments show that the detection performance of this method is better than traditional range-spread target detection methods.3.In the ground detection scene,due to the existence of isolated clutter,there will be much clutter false alarms in the detection results,which will lead to the decline of radar target detection performance and have adverse effects on the subsequent recognition process.Therefore,it is necessary to add a rejection process after the detection process to eliminate clutter false alarms.At present,there are few researches on rejection methods for high resolution range profile data.It is important to design effective clutter rejection methods for improving radar performance.Based on the above analyses,we propose a clutter rejection method based on Hausdorff distance and K-center one-class classifier.This method extracts the point set composed of the location and intensity of strong scattering scatterers as the rejection feature,which can better reflect the structural differences between clutter and target and has good robustness against noise.At the same time,the Hausdorff distance is used to replace the traditional Euclidean distance in K-center one-class classifier to better measure the similarity between features,and finally ensure the improvement of clutter rejection performance.The experimental results based on the real radar data show that the proposed rejection method has higher rejection accuracy rate than the traditional rejection methods,and can remove more clutter false alarms while retaining as many interested target samples as possible.