Research on integrated training-inference technology for intelligent interpretation based on aerospace big data

In recent years, the scale of space-ground systems has been continuously expanding, and data acquisition capabilities have experienced explosive growth. The large-scale Aerospace observation data provide the research foundation for the application of artificial intelligence technologies in the field of intelligent interpretation. However, the current business workflow and practical applications of Aerospace Intelligent Interpretation still have significant shortcomings, such as insufficient timeliness and weak target detection capabilities, which severely restrict the efficiency of Aerospace information acquisition. To address the current issues, this study proposes an Integrated Training-Inference Architecture for intelligent interpretation based on Aerospace Big Data. By constructing sample database, algorithm library, and visualization system supporting human-machine collaboration, this research proposes workflow which including algorithm recommendation based on collaborative-filtering, continuous iteration optimization and autonomous training of algorithm models, intelligent interpretation led by human-machine collaboration, and multi-task-oriented performance evaluation for intelligent interpretation. The process is verified through typical Aerospace Information Intelligent Application Tasks, demonstrating the feasibility of the Training-Inference Integration architecture in intelligent interpretation domain. By employing this method, the capability and efficiency of Aerospace Big Data Intelligent Interpretation have been effectively enhanced, enabling the proactive release of intelligent algorithms services and achieving human-machine collaboration mode in intelligent interpretation tasks.