In real-world scenarios, the identification accuracy of radar pulse data is often hampered by limited recognition features, a large volume of data samples, and unknown signals. To tackle these issues, a Sample Recommendation-based Multidimensional View Collaborative (SR-MDVC) domain adaptation algorithmis proposed. Initially, by integrating transfer learning theory and the hierarchical emission model, unknown and known pulse signals are defined from the perspective of probability density functions （PDFs）. Subsequently，a sample recommendation algorithm based on user collaborative filtering algorithm is introduced. This algorithm effectively reduces the disparity in the PDFs of unknown and known radar pulses. Moreover, a collaborative domain adaptation algorithm for multidimensional perspective model is developed. The model adapts data in the statistical feature dimension and the geometric feature dimension, and is optimized by iterating results across different dimensions. Finally, the domain - adapted data is fed into the least squares support vector machines (LSSVM) to complete the identification of unknown radar pulses. The effectiveness of this method is verified by semi-physical experiments, which demonstrate the excellent recognition performance of the proposed method, with a recognition accuracy of more than 0.9 for unknown and unbalanced data, and a 4-fold improvement in computational efficiency.

Detecting weak motion targets in complex environment by using radar sensors is of great importance for Internet of Things (IoT) applications. However, the complex environment with strong clutter and low-signal-to-clutter ratio (SCR) usually poses formidable challenges for achieving effective detection. To deal with this problem, different from the conventional energy-based method, this article proposes a novel detection technique based on the theory of information geometry (IG), which captures the nonlinear correlation feature (CF) of observed signals to effectively distinguish the target from clutter. Specifically, inherit the advantages of IG, we formulate a CF manifold and the geometric distances are derived to measure the dissimilarity of target and clutter. Then, a CF-based IG (CF-IG) detector is proposed. Moreover, we consider a multiframe detection strategy to further enhance the detection performance. A multiframe track-before-detect (TBD) method on the CF manifold is performed. Experimental results conducted on IPIX radar data and real measured drone target data validate the effectiveness and superiority of the proposed method.

In this letter, a novel radar forward-looking imaging technique based on beam pattern modulation and beam scanning is presented. First, the chirp beam, which presents quadratic varying phases within the main lobe, is generated and scans as a chirp pulse propagating along the azimuth direction by differentially exciting each element of a uniform linear array (ULA). Second, the target distribution is successfully reconstructed using 2-D pulse compression, and a theoretical analysis of the azimuth resolution is conducted. Finally, the sparse representation (SR) technique is employed to enhance the imaging performance. Simulation and experimental results validate the effectiveness and potential of the proposed method for acquiring high-resolution forward-looking images. This work holds promise for advancing the development of radar forward-looking methods and systems.

Towards the intelligent perception of radar intra-pulse modulation signals in complex environments, this paper proposes a method for fast recognition and SINR estimation of radar intra-pulse modulation signals. First, the adaptive random magnitude channel pruning algorithm (ARMCP) is designed to compress the YOLOv8n baseline model, coupled with channel-wise knowledge distillation (CWD) using the YOLOv8x large model as a teacher to restore the pruned YOLOv8n’s accuracy. Subsequently, a hybrid approach integrating hyper color segmenter and MobileNetv2 is employed for feature extraction within identified regions of interest (ROIs), facilitating the training of an XGBoost estimator for precise SINR prediction. The experimental results show that even at the pruning rate of 80%, the intra-pulse modulation recognition accuracy can still reach 99.07%, with a recognition speed of 3.2 ms, surpassing existing mainstream models such as YOLOv10 and YOLOv11. Additionally, the SINR estimation error is minimized to 0.958%, offering a novel paradigm and technical reference for advanced radar signal processing.

Target detection, especially in heterogeneous environments with strong clutter, is a difficult problem in signal processing. The existence of strong clutter in the heterogeneous environment usually causes the estimate of the clutter covariance matrix (CCM) to deviate from the truth, resulting in a degradation in detection performance. To address this problem, a novel CCM estimate method based on heterogeneous clutter clustering is proposed to improve the performance of radar detectors. First, it uses the distribution characteristics of clutter samples on the Hermitian positive-definite (HPD) manifold to segregate the clutter into clusters and select representative samples. Second, based on the cluster centers, a covariance matrix that can enhance the discrimination between the target echo and clutter on the HPD manifold is constructed. The proposed method can effectively enhance the discrimination between target and clutter because the reconstructed covariance matrix can reflect the clutter characteristics better. The advantage of the proposed method is demonstrated by experimental results on both simulated and real data. Compared with the conventional CCM estimation method based on the Log-Euclidean (LE) mean, the proposed method can improve the detection probability by 29%.

In this paper, the power spectrum information geometry (PSIG) detector, which inherits the performance advantages of matrix information geometry (MIG) detectors in heterogeneous clutter backgrounds, is proposed. The PSIG detector can address two urgent problems in applications of MIG detectors, which are the expensive computation expense and unavailable acquisition ability of target velocity. Specifically, the PSIG detector utilizes power spectrums instead of high dimensional covariance matrices to characterize sample data and employs subband filter bank to extend the detection from range cells to range-Doppler cells, thus it requires less computation expense and can obtain the target velocity information according to the Doppler cell. Experiments based on the real data show the advantages of the proposed PSIG detectors in comparison with competitive methods. Especially, in the experiments with the real-recorded airborne radar data, the proposed method can effectively suppress the heterogeneous main-lobe clutter without any prior knowledge and provides detection probability improvement of more than 30% to the competitive methods with low false-alarm ratios.

In this letter, an effective parametric IF estimation method based on piece-wise sparse representation (PWSR) is proposed to enhance the precision of conventional time-frequency distribution (TFD)-based estimators. Firstly, the signal is divided into a series of short-time segments according to the piece-wise quadratic frequency modulation (QFM) model. Then, the corresponding QFM parameters of each segment are accurately estimated by solving a sparse representation (SR) problem. Moreover, a construction scheme enabling the QFM dictionary to vary adaptively with the analyzed short-time signal is also provided. Finally, the IF of each segment is reconstructed according to the SR solution individually and combined together to generate the complete IF estimates. A radar imaging example verifies that the proposed method achieves a notable improvement in estimation accuracy as compared to existing TFD-based approaches.

To improve target detection performance in non-Gaussian backgrounds, the joint design of transmit sequence and receive filter for multiple-input-multiple-output (MIMO) radar is studied. By approximating the probability density function of observed non-Gaussian data with the Gaussian mixture model, a Riemannian manifold of Gaussian mixture distribution is developed to depict the complicated background first. Then, maximizing the geometric distance on manifolds, which is converted by maximizing the discrimination between the target and clutter, is proposed as the criterion for the joint design of transmit sequence and receive filter. Thereby, under the constant-modulus constraint, the joint design problem can be transformed into an optimization problem. However, the proposed optimization problem is non-convex and constrained. To solve this problem, a Riemannian optimization framework is provided. By taking the advantage of the underlying geometric and algebraic structure of the constraint space, the original constrained optimization problem in Euclidean space can be transformed into the unconstraint optimization problem over Riemannian product manifolds. Moreover, to obtain the global optimal solution, the Riemannian gradient of the geometric distance cost is derived for the conjugate gradient algorithm. Experiments demonstrate that the proposed method shows advantages in detection performance compared with competitive methods.

This paper addresses the problem of detecting radar targets submerged into strong sea clutter background. In this study, a novel type of detection method based on matrix information geometry (MIG) is developed. Filtering process and manifold projection are incorporated into detector design. Firstly, a filtering scheme for correlation coefficients is established via subband decomposition, such that a subband Hermitian positive definite (HPD) manifold constructed by a set of subband HPD matrices is formulated. Accordingly, the detection is performed as discriminating the target and the clutter on the subband HPD manifold. Then, in order to enhance the discriminative power between the target and the strong clutter, a manifold projection method that maps the HPD manifold into a lower-dimensional and more discriminative one is devised. In this study, the manifold projection is formulated as an optimization problem on a Stiefel manifold according to the principle of maximizing signal-to-clutter ratio (SCR). Subsequently, a manifold projection based subband MIG detector is proposed. Extensive experiments based on simulated data and real radar data are carried out to verify the effectiveness of the proposed method. The experimental results demonstrate that the proposed method can efficiently suppress the strong sea clutter and achieve better detection performance than the competitors.

Nonstationary moving target detection in spiky sea clutter is a challenging task due to the high-power, time-varying, and target-like properties of sea spikes. In this letter, we propose a time-frequency correlation (TFC)-based constant false alarm rate (TFC-CFAR) detection method on the time-frequency manifold, and apply it to the nonstationary moving target detection in spiky sea clutter. The data samples in each range cell are modeled as a TFC matrix that captures the correlation between two frequency components of the time-frequency distribution. The clutter covariance matrix is estimated by the geometric mean of a set of TFC matrices in reference cells. Three geometric metrics are employed to measure the dissimilarity between the clutter and target signals. Based on these geometric measures, three TFC-CFAR detectors are compared. Experiments performed on a real IPIX radar dataset confirm that the TFC can be used for identifying and eliminating sea spikes, while the TFC-CFAR detector achieves better detection performance than the conventional detectors.

Moving targets in a heterogeneous environment can be extracted through traditional robust principal component analysis (RPCA). Competitive RPCA algorithms address a nonconvex constraint by relaxing it to a fixed rank. However, the solution does not necessarily reach a global optimum. To address this problem, a moving target detection method by RPCA in the topological space of low-rank matrices is proposed to obtain superior target detection performance. First, RPCA is considered in the topological space of low-rank matrices, which is the closure of the fixed-rank manifold. Then, combined with manifold optimization and the proximal gradient, the RPCA-PGTSLr algorithm is applied to solve the problem caused by a non-differentiable sparsity term, so that the target can be precisely extracted. Experiments performed on measured data demonstrate that the proposed method exhibits advantages in detection performance over competitive methods in a heterogeneous environment.

From the perspective of feature extraction and classification, deep neural networks are widely used in radar target detection. However, in heterogeneous environment, the traditional deep neural networks are difficult to extract a robust feature, which leads to the degradation of network detection performance. In order to address this problem, a radar target detection method with multi-task learning in heterogeneous environment is proposed. Considering the influence of heterogeneous data distribution, the proposed method designs a contrastive learning module added to a multi-task autoencoder. It can learn a compact and distinguishable feature representation, which enhances the feature separability between the clutter and the target. Simultaneously, a classifier is introduced to realize a binary detection in the feature representation. Comprehensive experiments are carried out to show that the proposed method guarantees a good detection performance in heterogeneous environment and solves the issue of over-fitting to a certain extent. Compared with some classical detectors, the proposed method shows better performance.

Ship target detection in synthetic aperture radar (SAR) imagery is of great significance in the field of ocean monitoring. Classical constant false alarm rate (CFAR) detectors and emerging information geometry methods are model-driven essentially, requiring precise modeling of the sea clutter distribution. In the complex and changeable ocean scenes, the performance of these two types of detectors is limited. To solve this problem, a ship target detection algorithm in SAR imagery based on the maximum eigenvalue of the sample covariance matrix is proposed in this letter. Without seeking the distribution model of clutter backgrounds, the difference between the target and the clutter background is fully captured by constructing the sample covariance matrix, and its maximum eigenvalue is utilized as the test statistic. Experimental results on measured SAR images show that the proposed method achieves better detection performance and faster calculation speed compared with the existing typical methods.

Accurate time-frequency (TF) feature extraction of moving target is a challenging task due to the poor resolution and serious cross-terms of conventional TF analysis (TFA) methods. In this letter, an effective TFA algorithm based on the adaptive short-time sparse representation (ASTSR) is proposed to enhance the TF feature of moving target. Firstly, the limitation of the Fourier transform-based short-time TFA is revealed from the motion approximation perspective. Then, in order to achieve accurate motion approximation, the width of the analysis window is determined adaptively by minimizing the bandwidth of each short-time signal individually. Finally, the TF representation (TFR) with high energy concentration is obtained by utilizing the sparsity of these signal segments in the chirp dictionary. Comparisons indicate that the ASTSR provides high-resolution TFRs without producing interference terms at an acceptable computational cost while performing well in weak component expressing and signal denoising. Furthermore, a ISAR imaging example confirms the potential of the proposed method.

Due to a serious shortage of training data, the performance of adaptive clutter suppression suffers remarkable degradation in heterogeneous environments. To address this problem, a novel clutter suppression method via affine transformation on manifolds is proposed. First, training samples in heterogeneous environments are characterized on an established manifold in which the distribution properties are analyzed. Then, a clutter classification scheme is proposed, whereby the KL divergence decision rule is derived to identify the training data as either homogenous or heterogeneous samples. Afterward, based on the distribution properties of samples and the clutter classification scheme, an affine transformation on manifolds is proposed for sample augmentation by transporting heterogeneous samples into the region of homogeneous samples. Finally, the clutter in the area of interest is suppressed on the manifold, which combines the transformed samples with the homogeneous samples, such that superior performance is obtained. Experiments on both simulated and real data validate the superiority of the proposed method in highly heterogeneous environments.

Recently, the matrix information geometry (MIG) detector, which characterizes sample data as a Hermitian positive definite (HPD) matrix located on the HPD manifold, was rapidly developed and demonstrated extraordinary performance in numerous applications, especially in heterogeneous clutter backgrounds. In this paper, the geodesic normal coordinate (GNC)-based manifold filter is proposed to improve the detection performance of the MIG detector in strong clutter backgrounds. Using the GNC system, the distribution of target echoes and clutter on the high-dimensional manifold can be visualized and analyzed. Moreover, by exploiting the information concerning the distribution of matrices, the manifold filter is proposed to enhance target echoes and suppress strong clutter. Then, the manifold-filter-based MIG detector is designed, and its superiority is theoretically analyzed. The actual clutter data is utilized to verify the effectiveness of the proposed method. The results show that the proposed manifold filter achieves a signal-to-clutter ratio improvement of more than 5 dB over the existing MIG detectors.

This letter proposes an enhanced matrix constant false alarm rate (CFAR) detection method that works on the lower-dimensional Riemannian manifold. Motivated by general matrix CFAR detection method and dimensionality reduction scheme of the Riemannian manifold, this method obtains a mapping by maximizing the geometric test statistic. Dimensionality reduction is formulated as an orthonormal constraint optimization problem on the Grassmann manifold. Moreover, an explicit mapping is obtained by solving the optimization problem via conjugate gradient approach. Performances of the proposed method are evaluated on the lower-dimensional Riemannian manifold. Experiments on simulated data and real sea clutter data demonstrate that our method leads to the robustness to outliers and the improvement of detection performance over classical methods.

Modern object detectors usually suffer from low accuracy issues, as foregrounds always drown in tons of back-grounds and become hard examples during training. Compared with those proposal-based ones, real-time detectors are in far more serious trouble since they renounce the use of region-proposing stage which is used to filter a majority of back-grounds for achieving real-time rates. Though foregrounds as hard examples are in urgent need of being mined from tons of backgrounds, a considerable number of state-of-the-art real-time detectors, like YOLO series, have yet to profit from existing hard example mining methods, as using these methods need detectors fit series of prerequisites. In this paper, we propose a general hard example mining method named Loss Rank Mining (LRM) to fill the gap. LRM is a general method for real-time detectors, as it utilizes the final feature map which exists in all real-time detectors to mine hard examples. By using LRM, some elements representing easy examples in final feature map are filtered and detectors are forced to concentrate on hard examples during training. Extensive experiments validate the effectiveness of our method. With our method, the improvements of YOLOv2 detector on auto-driving related dataset KITTI and more general dataset PASCAL VOC are over 5% and 2% mAP, respectively. In addition, LRM is the first hard example mining strategy which could fit YOLOv2 perfectly and make it better applied in series of real scenarios where both real-time rates and accurate detection are strongly demanded.

Corner points play an important role in moving object detection, especially in the case of free-moving camera. Corner points provide more accurate information than other pixels and reduce the computation which is unnecessary. Previous works only use intensity information to locate the corner points, however, the information that former and the last frames provided also can be used. We utilize the information to focus on more valuable area and ignore the invaluable area. The proposed algorithm is based on reinforcement learning, which regards the detection of corner points as a Markov process. In the Markov model, the video to be detected is regarded as environment, the selections of blocks for one corner point are regarded as actions and the performance of detection is regarded as state. Corner points are assigned to be the blocks which are seperated from original whole image. Experimentally, we select a conventional method which uses marching and Random Sample Consensus algorithm to obtain objects as the main framework and utilize our algorithm to improve the result. The comparison between the conventional method and the same one with our algorithm show that our algorithm reduce 70% of the false detection.