Long-distance falling detection early warning method and device based on IROS robot and medium

The invention provides an IROS-based robot long-distance falling detection early warning method and device and a medium, and belongs to the technical field of robot early warning. Comprising the following steps: fixing a depth camera at the front part of the robot, and downwards fixing the depth camera at an angle of 45-90 degrees with the horizontal direction; the IROS control module controls the depth camera to carry out data acquisition and obtains point cloud data of a road section in front of the robot; extracting non-ground point clouds in the point cloud data, classifying the non-ground point clouds into accurate point clouds, fuzzy point clouds and rejected point clouds, and rejecting the rejected point clouds; converting the fuzzy point cloud into laser format data, projecting the laser format data to a horizontal plane, and dividing the laser format data into sectors according to the horizontal plane angle every certain angle; and judging whether the road section in front of the robot collapses or n