The recognition of 2D partially occluded objects using curve moment invariants

This paper puts forward an algorithm to recognize 2D partially occluded objects by using curve moment invariants. In this approach, images of both model objects and partially occluded unknown objects are firstly pre-processed. Then the objects in images are represented by their different order moment invariants based on boundary partitioning of the silhouettes. Afterwards, the match between the model objects and partially occluded unknown objects are implemented. Finally, the unknown objects can be recognized based on the feature calculating and matching. The experiment results show that this algorithm is effective in the recognition of partially occluded objects and robust in invariance to the transformation, scaling and rotation of the objects in images.