The invention relates to the technical field of positioning, navigation, semantic segmentation and target detection, in particular to an astronaut and scientific target cooperative positioning method and
system suitable for the lunar surface, and the method comprises the steps: achieving the position and posture
estimation of an astronaut through the collection of visual and inertial data, the construction of a feature point map, and the combination of VIO sliding window optimization, and generating a three-dimensional environment
point cloud; further performing three-dimensional target detection on the
point cloud, distinguishing a scientific target from a lunar
base station object, performing screening and constraint by using scientific target
prior information and a random sampling consistency
algorithm, constructing a target fitting geometric model into geometric constraint, feeding back the geometric constraint to a
pose optimization process, and introducing iterative optimization to realize
pose correction; meanwhile, the relative poses of the astronaut and the targets are calculated, and the environment map is updated. According to the method, the positioning precision and robustness are improved under the condition of no GPS, and the method is suitable for autonomous navigation and task execution in the complex lunar environment.