Abstract: A line laser scanning in-situ 3D measurement system was built up. Image denoising and gray centroid algorithms were adopted to suppress the interference induced by arc light and oxide layers during laser stripe feature extraction. A multi-layer weld bead point cloud stitching and fusion algorithm integrating downsampling, weld bead feature-based coarse registration and iterative closest point (ICP) precise registration was proposed to address the problem of sidewall information loss in linear laser scanning, and the extraction of weld bead dimensional features as well as 3D morphological reconstruction for the WAAM forming process was realized. The morphological evolution laws of the weld beads on aluminum alloy thin-walled parts fabricated via TIG-WAAM were analyzed with the constructed system. The results found that the deteriorated cooling condition of top bead along with the deposition height and the overlap of multi-layer beads' formation are the main reasons for the defects of thin-wall structure including the excess height at head, change of bead height and width, and the downflow metal on sidewalls.