Real-time astronaut visual tracking is the most important prerequisite for flying assistant robot to follow and assist the served astronaut in the space station. In this paper, an astronaut visual tracking algorithm which is based on deep learning and probabilistic model is proposed. Fine-tuned with feature extraction layers’ parameters being initialized by ready-made model, an improved SSD (Single Shot Multibox Detector) network was proposed for robust astronaut detection in color image. Associating the detection results with synchronized depth image measured by RGB-D camera, a probabilistic model is presented to ensure accurate and consecutive tracking of the certain served astronaut. The algorithm runs 10 fps at Jetson TX2, and it was extensively validated by several datasets which contain most instances of astronaut activities. The experimental results indicate that our proposed algorithm achieves not only robust tracking of the specified person with diverse postures or dressings but also effective occlusion detection for avoiding mistaken tracking.

Multi-satellite imaging mission planning has been difficult in various scenarios due to the complex constraints of multi-satellite imaging, the wide area covered by target points, and the difficulty of achieving different mission requirements in a short period of time with limited satellite resources. In addressing this challenge, this work investigates multi-satellite imaging mission planning based on the Unified Plan Model and Improved Adaptive Differential Evolution algorithm (UPM-IADE). First, a unified model is built based on two scenarios: a large-scale imaging mission and an emergency support mission, and then a mission assignment framework is adaptively selected based on mission priority. Second, a monorail task synthesis method based on visible time windows is created to clarify the execution relationship between the satellite and the target point. Finally, an individual weight ranking rule is developed, and the weight is used to combine the fitness value ranking and diversity ranking into a final fitness value ranking, which is used to select individuals that satisfy the mutation requirements into the mutation strategy pool for adaptive mutation strategy selection. Experiments 1, 2, 3, and 4 have demonstrated that UPM-IADE can successfully resolve the imaging satellite mission planning for both scenarios while providing remarkable performance in terms of high mission benefit and rapid response.

A person-following robot is under development for astronaut assistance on the Chinese Space Station. Real-time astronaut detection and tracking are the most important prerequisites for in-cabin flying assistant robots so that they can follow a specific astronaut and offer him/her assistance. In the limited space in the space station cabin, astronauts stand close to each other when working collaboratively; thus, large regions of their bodies tend to overlap in the image. In addition, because astronauts wear the same clothes most of the time, it is difficult to distinguish an individual astronaut using human body features. In this paper, we distinguish the astronauts by tracking their heads in the image. A deep learning model trained using big data is proposed for effective head detection. In addition, a motion model based on spatial clues is combined with the head detection results to track astronauts in the scene. A complete pipeline of the algorithm has been implemented and run efficiently on the Tegra X2 embedded AI microprocessor. A set of experiments were carried out and successfully validated the effectiveness of the proposed tracking algorithm. This algorithm is a step toward the implementation of robot assistants, especially in resource-limited environments.

The remote sensing satellite constellation based on micro-satellites is an important means to construct a global and all-sky earth observation system in the future. Therefore, realizing the recognition of aircraft activities on video micro-satellites is a key technology that needs to be solved urgently. In this paper, an efficient algorithm for aircraft activity recognition that can be deployed on video micro-satellites was proposed. First, aircraft detection was performed on the first incoming remote sensing image using a robust DCNN-based object detection model. Then, a multi-target tracking model incorporating geospatial information was built for aircraft tracking and activity recognition. The algorithm was deployed on an embedded AI computer which was a COTS component. The algorithm was verified using remote sensing videos from commercial micro-satellites. Experimental results show that the algorithm can process aircraft targets of different sizes, and is equally effective even with complex environmental backgrounds, lighting conditions, and various movements of the aircraft, such as turning, entering, and exiting. Based on aircraft tracking results and geospatial information, the motion speed of each aircraft can be obtained, and its activity can be divided into parking, taxiing, or flying. The scheme proposed in this paper has good application prospects in the realization of on-orbit event recognition in micro-satellites with limited computing and memory resources.

In recent years, object detection based on deep learning has made great progress in remote sensing. By designing different detection heads, multi-scale features are used for localization regression and classification. But more advanced detection heads were developed, primitive multi-scale features limited accurate detection of oriented objects. Therefore, this paper first proposes a channel attention module, named ESF, to enhance the feature representation of the network. ESF is a simple and effective lightweight module to highlight the importance of channel features by using spatial feature relations. The typical method used to assign weights is to pool the average or maximum value of multi-scale feature. In contrast, ESF makes use of the similarity of feature distribution to further converge the significant features. It enhances the importance of features with universal similarity in across channels, enabling the model to focus on more universally similar features adaptively. Many experimental results on the dataset HRSC2016 show that ESF combined with advanced directional target detection technology can enhance the detection and classification performance of the model. This demonstrates the effectiveness and wide applicability of ESF.