Recent advances in deep learning have shown exciting promise in low-level artificial intelligence tasks such as image classification, speech recognition, object detection, and semantic segmentation, etc. Artificial intelligence has made an important contribution to autopilot, which is a complex high-level intelligence task. However, the real autopilot scene is quite complicated. The first accident of autopilot occurred in 2016. It resulted in a fatal crash where the white side of a vehicle appeared similar to a brightly lit sky. The root of the problem is that the autopilot vision system cannot identify the part of a vehicle when the part is similar to the background. A method called DIDA was first proposed based on the deep learning network to see the hidden part. DIDA cascades the following steps: object detection, scaling, image inpainting assuming a hidden part beside the car, object re-detection from inpainted image, zooming back to the original size, and setting an alarm region by comparing two detected regions. DIDA was tested in a similar scene and achieved exciting results. This method solves the aforementioned problem only by using optical signals. Additionally, the vehicle dataset captured in Xi’an, China can be used in subsequent research.

Two-dimensional distributed source is a more realistic model to represent the target signal propagation in such fields as radar, sonar, and wireless communication. This paper presents a novel uniform circular sub-array (UCSA) configuration, which can be used to accurately estimate parameters of multiple two-dimensional incoherently distributed (TDID) sources. By sequentially utilizing the mapping relationship between the generalized array manifold, which has the spatial translation invariant relationship, and the signal subspace of TDID source, a closed-form solution of four parameters (central and spread parameters within azimuth and elevation dimension) of every TDID source can be obtained with low computational complexity. Furthermore, by properly using the signal subspace, the estimation accuracy is improved significantly. Numerical simulations illustrate the properties of the proposed method and show that it can simultaneously improve the robustness with respect to the scattering size and uncertainty of the spatial distribution. The proposed method has the potential to be used for fast and accurate multi-TDID source localization with arrays of small aperture and few array elements.

To solve the frequency estimation deterioration caused by independent and identically distributed (i.i.d.) noise, this letter proposes an exact and robust frequency estimation algorithm based on the accumulated phase-difference power on Kullback-Leibler divergence (APP-KLD). First, the frequency estimation problem is modeled by the accumulated power of phase-difference between the receiving signal and Fourier basis function. Furthermore, the true frequency is estimated iteratively by maximizing the Kullback–Leibler divergence, which measures the distance between the accumulated power and a constant reference vector. Simulations demonstrate the robust performance of the APP-KLD algorithm at various levels of heavy-tailed noise and sample numbers.

With the rapid advancement of deep learning techniques, side-scan sonar image segmentation has become a crucial task in underwater scene understanding. However, the complex and variable underwater environment poses significant challenges for salient object detection, with traditional deep learning approaches often suffering from inadequate feature representation and the loss of global context during downsampling, thus compromising the segmentation accuracy of fine structures. To address these issues, we propose SonarNet, a Global Feature-Based Hybrid Attention Network specifically designed for side-scan sonar image segmentation. SonarNet features a dual-encoder architecture that leverages residual blocks and a self-attention mechanism to simultaneously capture both global structural and local contextual information. In addition, an adaptive hybrid attention module is introduced to effectively integrate channel and spatial features, while a global enhancement block fuses multi-scale global and spatial representations from the dual encoders, mitigating information loss throughout the network. Comprehensive experiments on a dedicated underwater sonar dataset demonstrate that SonarNet outperforms ten state-of-the-art saliency detection methods, achieving a mean absolute error as low as 2.35%. These results highlight the superior performance of SonarNet in challenging sonar image segmentation tasks.

Coordinated swarms of low-cost commercial unmanned aerial vehicles (UAVs) pose a growing security threat to civil infrastructure, public events, and restricted airspace. Effective counter-UAV defense requires the real-time, cooperative scheduling of heterogeneous directed-energy systems, where high-energy lasers provide precise single-target engagement and high-power microwaves provide wide-area suppression. To address this threat, we formulate the defensive scheduling problem as a partially observable Markov game and propose the Heterogeneous Dual-Graph Multi-Agent Actor–Critic (HDG-MAAC). HDG-MAAC employs two graphs: a fixed-size local graph encoded by a graph attention network (GAT) actor for decentralized threat selection and a shared global graph with three semantic edge types encoded by a graph convolutional network (GCN) critic for stable credit assignment. This asymmetric design is tailored to decision-making and evaluation, respectively. In a two-system, four-target simulation benchmark with direct, zig-zag, and feint maneuver patterns, HDG-MAAC converges stably across eight seeds, with a final-stage return of $33.6\pm 4.0$, and neutralizes $2.79\pm 0.26$ targets over 1000 deterministic episodes, outperforming DDPG (2.20), MADDPG (2.13), and H2G-MAAC (1.64) with $p<0.01$. In analytically tractable geometric scenarios, the rule-based scheduler with exact geometric pointing still yields more raw interceptions. The learning method’s contribution is therefore limited to training stability, interpretability, and applicability beyond such scenarios, rather than the raw interception count. Ablation studies indicate that this robustness originates from the GCN critic, whose contribution lies in training stability rather than in raising the final performance ceiling. An attention analysis yields an overall target association rate of 42.0%, with 47.7% and 47.1% for direct and zig-zag maneuvers, respectively, compared to a 25% chance baseline, as well as a drop to 27.1% under feints. Without an explicit coordination reward, coordinated target convergence emerges in 34% of firing frames. HDG-MAAC provides a stable and interpretable decision support framework for civil security operators safeguarding public infrastructure from rogue drone groups. The framework is intended solely for protective civilian use, with responsible use considerations.