Most existing methods for partially occluded object recognition are only suit for Euclid and similarity transformation. As a result, the performance would be degraded in the affine and perspective transformation. This paper focuses on partially occluded shape recognition under affine transformation. The recognition algorithms are as follows First, a new local invariant under affine transformation is given based on the invariants in affine geometry. Second, a new similarity function is established to measure the similarity between models and object to be recognized on the basis of the local invariant. And then, a transform function is designed to normalize the similarity value between 0 and 1, so it is convenient to select similarity threshold. Finally, a loss feature judged function is constructed to judge whether each local feature is lost, and similarity is calculated only use the local features which are not lost. By comparing similarity with pre-threshold, we can recognize object from a partially occluded line drawing. The similarity function and the loss feature judge function consider the noises and occlusion. As a result, the reliability of recognition is improved greatly. The experiment results show that the proposed algorithms are quite robust to shape variations, including noise and occlusion. Moreover, they establish one-to-one correspondence between model features and object features in a scene, and can recognize multiple objects.