Continuum robots have been widely used in natural cavity intervention therapy, but the current flexible continuum robots have poor positioning accuracy and lack of accurate navigation means. In response, a multi-segment continuum robot shape reconstruction method based on electromagnetic positioning information was proposed, aimed at achieving accurate end navigation and positioning. An efficient two-stage active structure shape reconstruction method based on third-order Bessel curve was proposed. The solution of Bessel curve was optimized to nonlinear least squares problem, which was solved quickly and accurately by particle swarm algorithm with linear decreasing weights. The proposed method can accurately fit the shape of the two-segment continuum and has good real-time performance. The results show that within the bending angle process, the average root-mean-square error of shape reconstruction is 0.48 mm, which is 0.8% of the length of the continuum bending segment. The accuracy and reliability of the proposed shape reconstruction method for intracavitary interventions using micro-continuum robots are demonstrated through detailed modeling and validation results.