The effective classification methods of the small target objects in the no-fly zone are of great significance to ensure safety in the no-fly zone. But, due to the differences of the color and texture for the small target objects in the sky, this may be unobvious, such as the birds, unmanned aerial vehicles (UAVs), and kites. In this paper, we introduced the higher layer visualizing feature extraction method based on the hybrid deep network model to obtain the higher layer feature through combining the Sparse Autoencoder (SAE) model, the Convolutional Neural Network (CNN) model, and the regression classifier model to classify the different types of the target object images. In addition, because the sample numbers of the small sample target objects in the sky may be not sufficient, we cannot obtain much more local features directly to realize the classification of the target objects based on the higher layer visualizing feature extraction; we introduced the transfer learning in the SAE model to gain the cross-domain higher layer local visualizing features and sent the cross-domain higher layer local visualizing features and the images of the target-domain small sample object images into the CNN model, to acquire the global visualizing features of the target objects. Experimental results have shown that the higher layer visualizing feature extraction and the transfer learning deep networks are effective for the classification of small sample target objects in the sky.

This paper focuses on the trajectory tracking control for general nonlinear single-input single-output (SISO) systems in which the output is not directly related to the control input. To address the tracking problem with the consideration of possible model uncertainty, external disturbance, and control input saturation, we employ the input-output feedback linearization technique and design a finite-time disturbance observer-based terminal sliding mode controller to improve the tracking performance and enhance the robustness. The stability analysis is carried out by using the Lyapunov method. To alleviate the chattering while achieving an acceptable control performance, a boundary layer method is adopted for the trade-off between the high-frequency control actions and the bounded unavoidable nonzero steady-state error. The proposed method is evaluated on the two typical nonlinear systems, which are fully linearizable and partially linearizable, respectively, and compared to the state-of-the-art method in terms of tracking and robustness through comprehensive numerical simulations. The results show that the proposed method not only renders the estimated disturbance error tends to be zero in finite time, but also has superiority in the fast reaction to disturbance and small tracking error without high-frequency chattering.

The performance descending of current congested link inference algorithms is obviously in dynamic routing IP network, such as the most classical algorithm CLINK. To overcome this problem, based on the assumptions of Markov property and time homogeneity, we build a kind of Variable Structure Discrete Dynamic Bayesian (VSDDB) network simplified model of dynamic routing IP network. Under the simplified VSDDB model, based on the Bayesian Maximum A Posteriori (BMAP) and Rest Bayesian Network Model (RBNM), we proposed an Improved CLINK (ICLINK) algorithm. Considering the concurrent phenomenon of multiple link congestion usually happens, we also proposed algorithm CLILRS (Congested Link Inference algorithm based on Lagrangian Relaxation Subgradient) to infer the set of congested links. We validated our results by the experiments of analogy, simulation, and actual Internet.

Due to the lack of other component information in traditional magnetic leakage signal defects and the low accuracy of prediction methods, this paper proposes an improved residual network for magnetic leakage detection defect recognition method that predicts defect size and different detection speeds. A new defect diagnosis method based on ResNet18 on the Convolutional Neural Network (CNN) is proposed in this study. This method transfers the pre-trained ResNet18 network and replaces the activation function in the transferred network structure. It extracts features from transformed two-dimensional images obtained by converting the original experimental signals and signals with added noise, removing the influence of manual features. The results demonstrated that the improved ResNet18 network model, after transfer learning, achieved 100% prediction accuracy for all 10,000 grayscale images generated with defect lengths of 50 mm; width of 2 mm; and depths of 2 mm, 4 mm, 6 mm, and 8 mm. Moreover, the prediction accuracies for the quasi-static, slow, compensated fast, and fast scanning speeds were 99.20%, 98.50%, 93.30%, and 94.00%, respectively, for defect depths of 2 mm, 4 mm, 6 mm, and 8 mm. These accuracies surpass those of other models, demonstrating the significant improvement in prediction accuracy achieved by this method.

In the domain of autonomous driving, 3D point cloud maps generated by light detection and ranging simultaneous localization and mapping are crucial for navigation, as they provide essential spatial constraints by leveraging static environmental features. However, residual traces of dynamic objects within these maps compromise their reliability. Existing methods for dynamic object removal often misclassify static points while eliminating dynamic ones, failing to achieve an optimal balance between preservation and deletion. To address this challenge, we propose Pandora, a novel approach that combines polar grids with adaptive binary height encoding to accurately distinguish dynamic objects while preserving high-quality static structures across diverse environments. The method employs an adaptive hierarchical strategy for point cloud localization and a bitwise matrix comparison mechanism to detect dynamic regions. Additionally, it identifies near-ground dynamic points through density analysis and incorporates a dual ray-casting module to mitigate misclassification of static points caused by occlusions. By leveraging parallel processing during per-frame comparisons, the computational overhead associated with coordinate transformations is significantly reduced. Experimental evaluation on the KITTI and semi-indoor datasets demonstrates that Pandora achieves superior overall performance, with dynamic and static point identification accuracy reaching 98.6%, while maintaining high processing efficiency.