With the development of digital twin technology, the intelligent measurement degree of aerospace assembly products is getting higher and higher. Assembly of large parts is the basis of product performance, and assembly accuracy is an important index to ensure product quality. During the process of skin assembly, the deformation of the assembly force structure with the change of state will result in interference between thin-walled parts, and the forced positioning will affect the stability and safety of the product structure. Digital twin technology has the characteristics of multi-source, heterogeneous, massive, real-time and so on. The twin model gathers all the data obtained by acquisition, reading and fusion. It is particularly important to measure and perceive the product structure to obtain the real state information of components. Therefore, real-time sensing of structural deformation not only meets the characteristics of dynamic updating of the digital twin assembly model but also plays an important role in realizing the high-precision assembly of products. In this paper, based on the characteristics of real-time perception of product form and large-scale measurement field, combined with different measurement advantages of laser tracking, visual measurement and fiber monitoring system, a multi-source system fusion measurement model is constructed. The structural morphology sensing technology based on multi-source system fusion has the characteristics of real-time monitoring, which is not limited by the assembly state, and can effectively reduce the frequency of visual measurement deformation. The high-fidelity data obtained can provide a reference for constructing the twin assembly model. Firstly, the multi-station coordinate unified model of the multi-source system was established, and the pose relationship between multi-source systems was established based on the global coordinate system, so as to realize the measurement network fusion of multi-source system. Secondly, the heterogeneous data fusion model was established, and the heterogeneous data of optical fiber monitoring wavelength and spatial point coordinates were unified based on the surface interpolation reconstruction idea. The multi-source data fusion was realized based on the Gaussian process to predict the deformation point cloud and realize the product structure shape perception. Thirdly, the accuracy evaluation index is listed based on the measurement model, and the accuracy evaluation of the multi-source system fusion structure morphometry model is realized comprehensively. To verify the feasibility of the measurement model, the accuracy of the global measurement network was verified by the spatial coordinate component errors before and after the fusion of the common transfer stations. The accuracy of the fusion model was verified by comparing the predicted shape variables of multi-source data fusion, the shape variables obtained by the interpolation algorithm and the actual shape variables. The deformation point monitoring selection is determined based on ANSYS simulation analysis. Finally, the thin-walled structure with skin was taken as an example to simulate the assembly deformation experiment. The experimental results show that the absolute error of the deformation data obtained by the multi-source data fusion method is kept at 0.016 mm, and the average relative error is 4.66%, which is about 4% lower than that obtained by the interpolation algorithm. The multi-source system fusion structure shape sensing technology predicts the deformation model based on high-precision measurement field fusion heterogeneous data. The deformable point cloud contains the physical attribute information of the parts, which makes the surface detail reaction more complete. The measurement method of multi-source system fusion realizes the real-time monitoring of structural deformation, and the measurement data has the characteristics of high fidelity. The digital twin model is based on the interactive mapping of real-time data, which can monitor and predict the assembly process status of precision products, such as aerospace products, through measurement adjustment and finally achieve high-quality assembly. High-fidelity information sensing is the key to build the twin preassembly model. Therefore, the structure and morphology sensing technology based on multi-source system fusion can provide support for the establishment of the twin model.