Method and apparatus for detecting obstacle based on monocular camera

An apparatus for detecting an obstacle based on a monocular camera, comprising:

• an obtaining module configured to obtain a target frame image and its adjacent frame image shot by the monocular camera;
• a pre-tracking module configured to delete, based on a result obtained by tracking an initial feature point set of the adjacent frame image in a transformed image of the adjacent frame image, an unstable feature point from the initial feature point set of the adjacent frame image, to obtain a preferred feature point set of the adjacent frame image, wherein the transformed image of the adjacent frame image is calculated based on a preset transformation model;
• a subset obtaining module configured to divide a target feature point set to obtain several target feature point subsets, wherein the target feature point set is obtained by tracking the preferred feature point set of the adjacent frame image in the target frame image;
• a judging module configured to judge, for each of the target feature point subsets, whether the target feature point subset corresponds to an obstacle, based on a change of a distance between points within a ground projection point set of the target feature point subset from a time instant when the adjacent frame image is shot by the monocular camera to a time instant when the target frame image is shot by the monocular camera;
• and
• a determining module configured to determine a union of all the target feature point subsets which are judged as corresponding to obstacles as an obstacle point set of the target frame image.