In this paper, the shape of the object is reconstructed based on the fiber Bragg grating (FBG) flexible sensor encapsulated by polydimethylsiloxane (PDMS) material, and the influence of the number and layout of the sensors on the reconstruction accuracy of the object is studied. The COMSOL simulation software is used to verify the accuracy of the algorithm, and four sensors are prepared to measure and reconstruct the bent aluminum plate under set conditions. The maximum relative error of the shape reconstruction span is 1.203%, and the maximum relative error of the height is 2.802%. The research results provide the application basis for the shape detection of soft robots.