The straightness perception of scraper conveyor is a technology that helps to achieve intelligent operation of coal mine working faces. However, there is a problem of accuracy degradation caused by sensor torsion based on fiber Bragg grating (FBG) shape sensing technology. This paper proposes an accuracy compensation method based on torsion angle, and verifies the effectiveness of accuracy compensation through simulation and experimental research. Experimental tests have shown that when the torsion angle of the sensing substrate is 10° and 20°, the end accuracy of the reconstructed curve after accuracy compensation is increased by 2.77% and 1.93% compared to before compensation, respectively. This study indicates that compensation based on the torsion angle can significantly improve the accuracy of two-dimensional (2D) curve reconstruction, and it has broad engineering application prospects in fields such as straightness perception of scraper conveyors and monitoring of underwater pipelines.