Abstract: During the attachment process of small body spacecraft, there are scale, viewpoint and illumination variations in the image, making it difficult for traditional feature-matching algorithms to obtain accurate matches. In this paper, a small-body cross-scale illumination invariant matching algorithm is proposed. To address the problem of scale changes in the image during the attachment process, the global attention mechanism is combined with the dilated convolution to construct a scale adaptive adjustment module; the viewpoint invariant feature extraction module is designed to solve the problem of low matching accuracy under the large viewpoint changes in the feature matching algorithm; the self-attention mechanism is combined with the inter-attention mechanism to establish the feature dependency relationship, and the illumination invariant features are extracted. Experimental validation is carried out using the real images of Ceres and Bennu, and the results show that the proposed algorithm achieves an accuracy of more than 89% under large scale, view angle and illumination changes.