بهبود عملكرد آشكار‌ساز‌هاي هشدار غلط ثابت مبتني بر ميانگين هارمونيكي در كلاتر گوسي

In radar detectors, the target detection process relies on measured data, but the presence of outliers in these observations always acts as a destructive factor. The entry of these outlier an‎d anomalous values into detection algorithms causes statistical deviation an‎d a false increase in the decision threshold, which ultimately leads to the failure to detect targets. Dealing with this data in contexts related to Constant False Alarm Rate (CFAR) detectors, which is a key an‎d critical topic in the field of radar detectors, is no exception to this rule. Most algorithms developed in constant false alarm rate detectors are designed based on the homogeneity of the environment; this characteristic alone is sufficient to cause their performance to drastically degrade when facing environments contaminated with outliers. The conventional method employed in constant false alarm rate detectors to achieve a desired constant false alarm probability an‎d attain the CFAR property operates as follows: In the first stage, the power of clutter an‎d interference in reference cells is estimated using the energy of the received signal in the radar range cells. In the second stage, to detect targets in each Cell Under Test (CUT), its energy is compared with an adaptive threshold level. This adaptive threshold is a scaling factor of the clutter an‎d interference power estimate calculated from the energy of adjacent range cells in the first stage. Typically, in this class of detectors, it is assumed that the data in adjacent resolution cells possesses a statistical distribution similar to the clutter in the data under test; an assumption that is theoretically an‎d precisely equivalent to a homogeneous environment. However, this assumption is consistently violated by the non-identical clutter statistics of adjacent cells an‎d the cell under test, as well as by the presence of challenges such as clutter edges, interfering targets, an‎d electromagnetic interference, all of which lead to the creation of outliers an‎d environmental heterogeneity. On this basis, many studies strive to improve detector performance by introducing algorithms that are robust against outliers. Yet, many of these robust detectors utilize sorting-based methods on the data present in the reference window to identify an‎d eliminate outliers. Although this approach can be effective in isolating heterogeneous data, the sorting process, which requires repetitive comparisons between data samples, leads to a significant increase in computational complexity. This increase in complexity is considered a serious limitation, especially in applications where rapid processing speed is a critical parameter. In the first step, an innovative approach aimed at reducing the computational complexity of the data sorting process in the Harmonic Censoring detector is presented. The prominent feature of this approach is its generalizability an‎d applicability to all data-sorting-based detectors. Subsequently, we will discuss the performance of the proposed detectors not only in Gaussian clutter but also in non-Gaussian clutter, including cases with heavy-tailed distributions, considering two different methods. In the second step,as the concluding part of this thesis, a new detection method based on the Norm-P is introduced to effectively mitigate outliers while prioritizing speed an‎d simplicity. Finally, the proposed algorithmʹs superiority over the previous method is demonstrated using stan‎dard eva‎luation metrics.

محمدرضا تابان

محمد‌مهدي نقش , حامد نريماني