Radar high resolution range profile target recognition based on T-mixture model

Radar high‐resolution range profiles target recognition based on stable dictionary learning

“…This means that conventional dictionary learning methods may be affected by perturbations, thus leading to instability of sparse representations and limiting the recognition performance. Some parametric statistical models, that is, AGC [3] and FA [11], are also compared with further validate the effectiveness of our method.…”

“…We examine the performance of SDL on the measured HRRP data from three real airplanes, which are extensively used in [1, 3–5, 8–13]. The parameters of the airplane targets and measurement radar are shown in Table 1 and the projections of target trajectories onto the ground plane are shown in Fig.…”

“…A high‐resolution range profile (HRRP) is the amplitude of the coherent summations of the complex time returns from target scatterers in each range cell, which contains abundant target structure signatures, such as target size, scatterer distribution, etc. Therefore, radar HRRP target recognition receives intensive attention from the radar automatic target recognition (RATR) community [1–13].…”

“…Another contribution of this paper is that efficient algorithms based on greedy strategy are proposed for the implementation of SDL. Experimental comparisons with three traditional HRRP‐based RATR methods: adaptive Gaussian classifier (AGC) [3], PCA‐based minimum reconstruction error approximation [10], FA model [11], and two dictionary learning methods: K‐SVD [13], Method of optimal directions (MOD) [14], demonstrate that the proposed method has a superior recognition performance on the measured radar HRRP dataset. Furthermore, SDL's mechanism makes it possible to appropriately share the information among samples from different target‐aspects and learn the global dictionary collectively, thus offering the potential to improve the overall recognition and generalisation performance even with small training data size.…”

Compressive Sensing of Stepped-Frequency Radar Based on Transfer Learning

“…Therefore, to estimate the pdf of this aspect-frame, we should first calculate the statistical weights of this aspect-frame. Since the posterior distributions and have been obtained, we can use the following posterior mean and covariance to represent the statistical property of from Gibbs sampling: (23) It is reasonable to assume that follows a complex Gaussian distribution with zero mean [34], and which can also be verified by the experimental results. Therefore, to simplify the model, the distribution of can be written as:…”