OCCLUSION HANDLING FOR COMPUTER VISION

摘要

Disclosed are a system, apparatus, and method for performing occlusion handling for simultaneous localization and mapping. Occluded map points may be detected according to a depth-mask created according to an image keyframe. Dividing a scene into sections may optimize the depth-mask. Size of depth-mask points may be adjusted according to intensity. Visibility may be verified with an optimized subset of possible map points. Visibility may be propagated to nearby points in response to determining an initial visibility of a first point's surrounding image patch. Visibility may also be organized and optimized according to a grid.

主权项

1. A method for computer vision object occlusion handling, the method comprising:
projecting map points observable from a 3D map to points of a keyframe, wherein the keyframe comprises an image with an associated camera viewpoint; creating, from each of the points of the keyframe, a depth map comprising a plurality of depth map points, wherein each depth map point comprises a value representing a distance from the keyframe's associated camera viewpoint to a respective map point from the 3D map, and wherein if a depth map point represents a plurality of respective map point distances, the value comprises a smallest of the plurality of respective map point distances; identifying potentially visible points in the keyframe, the potentially visible points comprising the points of the keyframe having a representative distance within the depth map; selecting a plurality of the potentially visible points to test for visibility in the image; testing the selected plurality of the potentially visible points for visibility, wherein the testing comprises determining whether each of the plurality of the potentially visible points comprises a surrounding image patch from the keyframe that corresponds to an image patch from the 3D map; and mapping an environment according to results from the testing for visibility.