Learning General Feature Descriptor for Visual Measurement with Hierarchical View Consistency