Study On Some Issues Of Wideband Radar Detection

The fundamental advantage offered by wide radar bandwidth is increased information about the presence, location, and identity of targets. By enhancing the signal bandwidth, it can be realize the imaging and mapping, enhance the capability of target detection and get more accurate measurements of the location of the targets. But the characteristics of the wideband radar signals reflected from targets are different from the conventional narrowband radar signals, so it has been a new research field to suppress clutter, detect target by target returns and then to realize the integration of detection, tracing, imaging and recognizing. Especially, Research on the wideband radar detection is significant in both the theory and application.This dissertation is addressed on the range-spread target detection in the wideband radar system, with emphasis on non-coherent integration, time-domain coherent integration , frequency-domain coherent integration and robust detection. The main work of this dissertation can be summarized as the following:1. The radar equation, target return models and clutter models of the wideband radar are analyzed. Firstly, the radar equation and target radar cross section (RCS) of the wideband radar are discussed; then the precise models of target returns from multiple pulse are deduced, which is further simplified to three different approximate models based on the rotation velocity of the target; finally, several clutter models and their generation method are introduced. Based on the EM algorithm, a parameter estimator for the K distribution clutter is designed. Particularly for small number of data snapshots, the proposed algorithm is illustrated via simulation results to provide better performance than conventional algorithms and therefore is more appropriate for real-world applications.2. The non-coherent detection of wideband radar is researched. Firstly, the previous algorithms for non-coherent wideband target detection are reviewed, where the merits and shortcomings of these methods are analyzed, By utilizing the goodness-of-fit (GOF) technique, a single-pulse based detection algorithm is developed. In contrast to earlier energy-integration based methods, our algorithm fully exploits the impact of the target echoes on statistical properties of background clutter, and hence improves the detection performance in the cases of low signal-to-clutter ratio (SCR) and large variation of the clutter amplitute. Secondly, a non-coherent integration method for multi-pulse detection problem is proposed on the basis of the viterb method. This method uses an improved bi-directional viterb algorithm to estimate the velocity of the target and the locations of strong scatterers, around which the range cells of interest are selected for non-coherent energy integration. Finally, examples from the measured data are used to illustrate the better detection performance and increased robustness of the proposed method.3. The time-domain coherent integration detection of range-walking target during a coherent processing interval (CPI) is studied. Firstly, the analysis based on the measured data is presented to illustrate the necessity of considering range walking of the target. To solve for this problem, the parameters of target echoes from each range cells are estimated for different expected velocities by using the generalized likelihood function (GLF) and maximum likelihood (ML) criterions, and then a two-step generalized likelihood rate test based detector (2S-GLRT-BD) is obtained and its constant false alarm rate (CFAR) property with respect to the noise power is proved. Secondly, by using the property of strong scattering cells around the target and the correlation between the two adjacent given velocities, an improved three-step GLRT-based detector (I3S-GLRT-BD) is developed to avoid the performance degradation by the range bins that only contain noise. Finally, with the two detectors mentioned above, a range-Doppler spread target GLRT-based detector(RDST-GLRT-BD) is developed to in the situation of the Doppler-spread caused by the rotation of targets. This method detects the Doppler spread and the number of Doppler spread for different range cells, estimates the parameters using the expectation maximization (EM) and atomic decomposition (AD) algorithms, and therefore, achieves the time-domain conherent integration in the presence of Doppler spread.4. The coherent integration detection of wideband radar in the frequency domain is researched. The models of the frequency-domain target returns and heterogeneous Gaussian clutter are analyzed. It can be seen that the problem of target range-walking during a CPI in the time domain can be equivalently represented by the linear variation of the Doppler frequency between two adjacent frenquency cells. We also show that under the suboptimal assumption, the time-domain heterogeneous Gaussian clutter becomes homogeneous in the frequency domain. Based on the above two conclusions, a frequency-domain GLRT-based detector (FD-GLRT-BD) is proposed and its statistical properties are discussed. Finally, numerical examples are provided to illustrate that the proposed detector can achieve multi-pulse coherent integration in the presence of heterogeneous clutter.5. The robust detection problem of wideband radar is studied. Firstly, the robust GLRT (RGLRT) detection methods are introduced and the impact of the steering mismatches on detection performance is analyzed. Then, a novel robust detection algorithm with low computational complexity is developed by using the Lagrange multiplier method. Experimental results with measured data illustrate its significant performance improvement over the conventional GLRT methods in the presence of steering mismatches. Secondly, via analyzing the measured data, we point out that the limitation of conventional robust detection algorithm by given searching area , to overcome these problems, we propose a robust detection algorithm based on the Semi-definite Programming. By fitting the detection threshold curve varying with the searching area, and then comparing the coherent integration energy to the fitted threshold using Semi-definite Programming, the proposed method can provide more robust detection. Moreover, a bi-threshold technique is also adopted to ease the computational support. Finally, simulation results with the measured data validate the effectiveness of the proposed algorithm and illustrate its enhanced detection performance as compared with conventional as well as robust GLRT methods.