Aiming at scattered data, proposes an adaptive implicit surface reconstruction algorithm driven by fitting-error. Adaptive fitting is processed based on the fitting-error of quadric surface, which is different from traditional method that starts fitting from top-level node of octree, and the fitting times is reduced and reconstruction speed is accelerated, meanwhile, weight function is modified with density function, which weaken the influence caused by asymmetrical scattered data and improve the quality of reconstruction surface. The experimental results indicate that the algorithm is simple with fast reconstruction speed and good reconstruction surface.