To deal with the localization problem of robot equipped with laser sensor, a mobile robot Monte Carlo self-localization method based on adaptive curvature estimation for environmental features extracted was proposed. During the mobile robot localization, poses of the robot were predicted by motion-model, and then distribution of particles set was updated according to the similarity measurements of geometry by observation model, and robot self-localization was realized. Simulation experiments showed the proposed method could satisfy the requirements of the mobile robot self-localization in greenhouse.