Development of a Multi-Modal Object Detection System for Space Station Applications

The fast growth of space exploration and orbital infrastructure has generated the necessity of smart systems of monitoring, which can function in an autonomous limited setting like that of space stations.Object detection models based on deep learning have proven to be an efficient solution to automated visual monitoring and analysis.In this paper, the authors have developed an object detection system that is used to detect and localize important objects in a simulated space-station environment.The system proposed uses a deep learning-based detection model with an image processing pipeline to detect objects in real time and show bounding box predictions.It was developed in Python and up-to-date computer vision packages, which allowed detecting several object classes in the input images or video streams.The project illustrates how convolutional neural network-based models of detecting objects may be put to practice in space-environment monitoring missions like equipment recognition, safety checks, and automated inspections.Experimental results show that the model successfully identifies objects with satisfactory accuracy under controlled testing conditions.The suggested solution emphasizes the use of the current computer vision systems in the construction of lightweight intelligent monitoring systems that can be applied in the aerospace industry.Although the system is currently used as a prototype to be experimented with in academia, it shows that it is feasible to deploy AI-based-visual-perception systems to support astronauts and autonomous robots in the space station operations.