This application relates to the field of smart
home appliance technology, and discloses a method and apparatus for cross-domain small-sample object image classification for smart terminals. The method includes: acquiring an image of food to be identified;
processing the image to extract basic feature vectors; mapping the basic feature vectors using a dual Riemannian manifold
processing module to obtain Euclidean space features and hyperbolic space features respectively; fusing the Euclidean space features and hyperbolic space features to obtain a comprehensive feature representation for
food classification; and determining the
food classification result based on the comprehensive feature representation. The dual Riemannian manifold
processing module includes Euclidean space branches and hyperbolic space branches. This method achieves multi-space
feature fusion by collaboratively extracting Euclidean space features and hyperbolic space features using dual Riemannian manifolds, thereby improving the feature representation capability and classification accuracy of food images.