The clutter selection strategy based on sliding window in the conventional constant false alarm rate (CFAR) algorithm leads to different clutter qualities between pixels of the same target in a complex environment. To solve the problem, this letter proposes an improved CFAR algorithm based on density clustering. First, a two-parameter CFAR is used to detect ship targets. Then, density clustering is performed on each detected target pixel based on spatial distance and detection threshold to improve the target detection accuracy. Finally, false alarms caused by speckle noise are eliminated by using the number of times a pixel is clustered. The experimental results show that compared with the conventional CFAR algorithm and the superpixel-level CFAR detectors for ship detection in synthetic aperture radar (SAR) imagery (SP-CFAR), the proposed algorithm achieves a detection accuracy improvement of over 14.8% in heterogeneous clutter scenarios and dense target scenarios, while maintaining a low false alarm rate no higher than 0.13% in strong noise environments.

This paper discusses the problem of detecting subspace signals in structured interference plus compound Gaussian sea clutter with persymmetric structure. The sea clutter is represented by a compound Gaussian process wherein the texture obeys the inverse Gaussian distribution. The structured interference lies in a known subspace, which is independent with the target signal subspace. By resorting to the two-step generalized likelihood ratio test, two-step Rao, and two-step Wald design criteria, three adaptive subspace signal detectors are proposed. Moreover, the constant false-alarm rate property of the proposed detectors is proved. The experimental results based on IPIX real sea clutter data and simulated data illustrate that the proposed detectors outperform their counterparts.

The exact knowledge of the signal steering vector is not always known, which may result in detection performance degradation when a signal mismatch occurs. In this paper, we discuss the problem of designing a robust radar target detector in the background of Gaussian noise whose covariance matrix is unknown. To improve robustness to mismatched signals, a random perturbation that follows the complex normal distribution is added under the alternative hypothesis. Since traditional detectors that divide complex parameters into real parts and imaginary parts are sometimes difficult to obtain, a new robust, complex parameter gradient test is derived directly from the complex data. Moreover, the CFAR property of the new detector is proven. The performance assessment indicates that the gradient detector exhibits suitable robustness to the mismatched signals.

Convolutional neural network (CNN)-based detectors for small targets on the sea surface have proven effective, yet their increasingly complex structures and high energy demands pose challenges to deployment on resource-limited devices. To address this issue, this letter proposes a residual spiking network (RSN), which is developed within the advanced framework of spiking neural networks (SNNs). By using spiking neurons (SNs) as core computational units, the RSN can efficiently transmit features extracted from time-frequency graphs through discrete spikes in the backbone network. Meanwhile, the output is derived from the membrane voltage of the SNs, enabling reliable control over false alarms. Experimental results from the IPIX dataset validate the RSN’s stability in controlling the probability of false alarms (PFA). With an observation time of 1.024s and a PFA of 0.001, the average probability of detection (PD) is 0.8266. The RSN demonstrates a PD comparable to that of ResNet18 while consuming only 70% of its energy, highlighting its potential for practical applications in resource-constrained environments.

The weight vector in space-time adaptive processing (STAP) algorithm will lead to notches at the position of the interfering targets when there are interfering targets in the training data. If these interfering targets are close to the target of interest on the space-time spectrum, the target signal self-nulling occurs. To deal with this problem, a machine learning-aided non-homogeneity detection (ML-NHD) method is proposed. More specifically, the subaperture smoothing technique is first performed on each training data to obtain the subaperture sample covariance matrices (SCMs). We prove that when the airborne radar works in side-looking mode and the clutter foldover factor is an integer, the numbers of large eigenvalues of the subaperture SCMs are different for the ordinary training data samples and outlier training data samples. Then, four features are constructed based on the differences in the characteristics of eigenvalues and eigenvectors of the subaperture SCMs. Finally, a binary classifier based on support vector machine (SVM) is trained to classify the ordinary training data and the outlier training data. The performance assessment shows that the ML-NHD method can detect the outlier training data effectively and achieves better performance of clutter suppression compared with the conventional methods.

Airborne forward-looking radar (AFLR) has been more and more impoatant due to its wide application in the military and civilian fields, such as automatic driving, sea surveillance, airport surveillance and guidance. Recently, sparse deconvolution technique has been paid much attention in AFLR. However, the azimuth resolution performance gradually decreases with the complexity of the imaging scene. In this paper, a data-driven airborne Bayesian forward-looking superresolution imaging algorithm based on generalized gaussian distribution (GGD- Bayesian) for complex imaging scene is proposed. The generalized gaussian distribution is utilized to describe the sparsity information of the imaging scene, which is quite essential to adaptively fit different imaging scenes. Moreover, the mathematical model for forward-looking imaging was established under the maximum a posteriori (MAP) criterion based on the Bayesian framework. To solve the above optimization problem, quasi-Newton algorithm is derived and used. The main contribution of the paper is the automatic selection for the sparsity parameter in the process of forward-looking imaging. The performance assessment with simulated data has demonstrated the effectiveness of our proposed GGD- Bayesian algorithm under complex scenarios.