With the rapid development of remote sensing technology, optical remote sensing images are increasingly being used in areas such as military reconnaissance, environmental monitoring, and urban planning. Due to the small number of pixels, fuzzy features, and complex background, it is difficult for conventional convolutions to effectively extract features from small objects. To address this problem, we propose to use multi-scale dilated convolutions to increase the receptive field size of the model to adapt to changes in object size, capture multi-scale contextual information of the feature map, and extract richer object features. First, we propose a Dilated Convolutional Residual (DCR) module for high-level feature extraction in the network. Second, the context aggregation (CONTEXT) module uses remote interaction to perform associative computation on images using contextual aggregation, allowing the model to understand the global semantic information of the image. We propose a novel object detection method, DCN-YOLO, which achieves an AP50 of 56.6 on the AI-TOD dataset, effectively improving the detection accuracy and robustness of small objects in remote sensing images. It provides a new technical approach to the detection of small objects in remote sensing.

With the rapid development of space programs in various countries, the number of satellites in space is rising continuously, which makes the space environment increasingly complex. In this context, it is essential to improve space object identification technology. Herein, it is proposed to perform intelligent detection of space objects by means of deep learning. To be specific, 49 authentic 3D satellite models with 16 scenarios involved are applied to generate a dataset comprising 17,942 images, including over 500 actual satellite Palatino images. Then, the five components are labeled for each satellite. Additionally, a substantial amount of annotated data is collected through semi-automatic labeling, which reduces the labor cost significantly. Finally, a total of 39,000 labels are obtained. On this dataset, RepPoint is employed to replace the 3 × 3 convolution of the ElAN backbone in YOLOv7, which leads to YOLOv7-R. According to the experimental results, the accuracy reaches 0.983 at a maximum. Compared to other algorithms, the precision of the proposed method is at least 1.9% higher. This provides an effective solution to intelligent recognition for spatial target components.

This paper proposes a novel robust control scheme for tip trajectory tracking of a lightweight flexible single-link arm. The developed control scheme deals with the influence of tip payload changes and disturbances during the working process of the flexible arm, thus realizing the accurate tracking for the tip reference trajectory. The robust control scheme is composed of an inner loop and an outer loop. The inner loop adopts the traditional PD control, and an active disturbance rejection control (ADRC) with a sliding mode (SM) compensation is designed in the outer loop. Moreover, the sliding mode compensation is mainly used to cope with the disturbance estimation error from the extended state observer (ESO), by which the insensitivity to tip payload variations and strong disturbance resistance is achieved. Finally, some numerical simulations are performed to support the theoretical analysis. The results show that the system is more robust to the tip mass variations of the arm and more resistant to the external torque after adding the sliding mode robustness term to the ADRC.

Beam pointing control based on Risley prisms is of great significance in wide-angle, high-precision application scenarios, such as laser communication, but its inherent nonlinear system characteristics seriously restrict the performance of beam pointing control, such as accuracy. For this reason, this paper combines the theory of fully actuated control with reinforcement learning methods and designs a fully actuated control method based on reinforcement learning compensation: suppressing the influence of system nonlinearity through fully actuated control, using reinforcement learning to estimate system perturbations and nonlinearities, and then outputting a compensated control quantity using the low-dimensional output of fully actuated control as the reference input of reinforcement learning reduces the complexity of learning and realises the end-to-end uncertainty estimation. Finally, the stability of the method is theoretically analyzed, and the effectiveness of the method is verified by experimental analysis, which can further improve the beam pointing accuracy of the Risley prism system.

To achieve precise measurement without landing, the high-mobility vehicle-mounted theodolite needs to be leveled quickly with high precision and ensure sufficient support stability before work. After the measurement, it is also necessary to ensure that the high-mobility vehicle-mounted theodolite can be quickly withdrawn. Therefore, this paper proposes a hierarchical automatic leveling strategy and establishes a two-stage electromechanical automatic leveling mechanism model. Using coarse leveling of the first-stage automatic leveling mechanism and fine leveling of the second-stage automatic leveling mechanism, the model realizes high-precision and fast leveling of the vehicle-mounted theodolites. Then, the leveling control method based on repeated positioning is proposed for the first-stage automatic leveling mechanism. To realize the rapid withdrawal for high-mobility vehicle-mounted theodolites, the method ensures the coincidence of spatial movement paths when the structural parts are unfolded and withdrawn. Next, the leg static balance equation is constructed in the leveling state, and the support force detection method is discussed in realizing the stable support for vehicle-mounted theodolites. Furthermore, a mathematical model for “false leg” detection is established furtherly, and a “false leg” detection scheme based on the support force detection method is analyzed to significantly improve the support stability of vehicle-mounted theodolites. Finally, an experimental platform is constructed to perform the performance test for automatic leveling mechanisms. The experimental results show that the leveling accuracy of established two-stage electromechanical automatic leveling mechanism can reach 3.6″, and the leveling time is no more than 2 mins. The maximum support force error of the support force detection method is less than 15%, and the average support force error is less than 10%. In contrast, the maximum support force error of the drive motor torque detection method reaches 80.12%, and its leg support stability is much less than the support force detection method. The model and analysis method proposed in this paper can also be used for vehicle-mounted radar, vehicle-mounted laser measurement devices, vehicle-mounted artillery launchers and other types of vehicle-mounted equipment with high-precision and high-mobility working requirements.