Shape sensing technology is a new research direction in recent years and has a wide range of application prospects. In this paper, a shape reconstruction method based on strain mode mode and error compensation is proposed. By measuring the strain data of some position points of the object, the modal theory was used to realize the strain-displacement transformation, and then the shape of the object was reconstructed. In this paper, titanium alloy plates with length, width and height of 1 000, 1 000 and 0.5 mm were taken as the research object. The displacement mode and strain mode modes were obtained by ANSYS workbench 18 finite element simulation software. According to the modal similarity of position points in the finite element simulation, the K-means++ clustering algorithm was used to optimize the position of strain measuring points, and the strain deformation was generated by applying 400 N force on the upper surface of the alloy plate. The shape reconstruction error of the proposed algorithm is smaller than that of the conventional uniform distribution algorithm. A Radial Basis Function Neural Network (RBFNN) was used to train the error and reconstructed displacement data set. According to reconstructed displacement prediction error, the fitting error was less than 3.5%.