Abstract: Due to the incomplete, irregular and discrete nature of the input point cloud data, it is difficult for the existing complementation methods to extract enough global information from the defective point cloud data to generate accurate and complete shapes. In this paper, a dual-sequence feature-enhanced coding and decoding method based on selective state space is proposed for repairing and completing 3D defective point cloud shapes. Firstly, two different sequences of the defective point cloud objects are obtained by using Hilbert filling curves and their dual-sequence features are extracted, which are fused together and integrated with the compression and hidden states of the multi-level selective state space to obtain the enhanced coding of the dual-sequence features of the defective point cloud; secondly, the point-by-point segmentation of the point cloud coding is carried out to capture the existing shapes and the missing structures, and then the low-density point cloud features that can represent the contours of the three-dimensional skeleton are generated. Next, the point cloud encoding is segmented point by point to capture existing shapes and missing structures, and generate low-density point cloud features that can represent the 3D skeleton contours. In the meantime, the point cloud global attention features are used to remove outliers that deviate from the overall shape, and the grouped local features are used to remove noisy points that deviate from the local structures; next, the key features and query features are used to construct a multi-stage point cloud up-sampling decoder, and a complete high-density 3D point cloud is obtained with the supplemented shapes. Experimental results on three datasets with 11 mainstream point cloud complementation methods show that the evaluation metrics CD-1, CD-2 and EMD of the proposed method decrease by 1%-10%, 1%-22%, 3%-23%, and increase by 8%-11%, respectively, compared with the baseline network, which verifies the validity of the method.