Conducting materials science experiments in a microgravity environment mitigates the influence of gravity, enabling the study of intrinsic material growth mechanisms and the fabrication of materials with enhanced properties. The High-Temperature Materials Science Experimental Cabinet aboard the Chinese Space Station is equipped with an X-ray transmission imaging module, facilitating real-time imaging and observation of material solidification processes under microgravity. However, due to the constraints of the space station's experimental conditions, the X-ray images acquired by this module often exhibit blurriness, making direct observation of microstructures challenging. To address this issue, This paper proposes the GC-UNet++ image segmentation algorithm, specifically tailored for analyzing the microstructures formed during the solidification of Na
6Mo
11O
36 material. The algorithm's effectiveness is rigorously evaluated in terms of both image segmentation performance and its relevance to materials science applications. The experimental results demonstrate that in the image segmentation task, GC-UNet++ outperforms established algorithms such as UNet, UNet++, DC-UNet, UNet3+ and Pretrained-Microscopy-Models, achieving notable improvements across various image segmentation metrics. Furthermore, the microstructures formed during the growth of the Na
6Mo
11O
36 material can be segmented more accurately. This provides new ideas and methods for the study of the microstructure segmentation of materials, and has important application value.