Plane estimation using object detection

A computer-implemented method comprising:

• obtaining at least two images of a real-world environment whose fields of view overlap at least partially;
• obtaining pose information indicative of corresponding camera poses from which the at least two images are captured, wherein the pose information is generated by a camera pose-tracking means that provides sixDegrees of Freedom (6DoF) tracking and comprises at least one of:
• an optics-based system with detectable markers, an inertial measurement unit (IMU or TIMU), or a simultaneous localization and mapping (SLAM) processor;
• detecting, in the at least two images, at least one object that is in contact with a given plane present in the real-world environment by executing an object-detection algorithm;
• identifying, in the at least two images, within portions of the images that represent an overlap of the fields of view, same features of the at least one object that lie on the given plane;
• determining poses of the same features, wherein a pose of a given same feature is determined by triangulation, based on a disparity in two-dimensional positions of the given same feature in the at least two images and the corresponding camera poses from which the at least two images are captured;
• and
• estimating the given plane based on the poses of the same features.

A computer program product comprising a non-transitory machine-readable data storage medium having stored thereon program instructions that, when executed by a processor, cause the processor to:

• obtain at least two images of a real-world environment whose fields of view overlap at least partially;
• obtain pose information indicative of corresponding camera poses from which the at least two images are captured, wherein the pose information is generated by a camera pose-tracking means that provides sixDegrees of Freedom (6DoF) tracking and comprises at least one of:
• an optics-based system with detectable markers, an inertial measurement unit (IMU or TIMU), or a simultaneous localization and mapping (SLAM) processor;
• detect, in the at least two images, at least one object that is in contact with a given plane present in the real-world environment by executing an object-detection algorithm;
• identify, in the at least two images, within portions of the images that represent an overlap of the fields of view, same features of the at least one object that lie on the given plane;
• determine poses of the same features, wherein a pose of a given same feature is determined by triangulation, based on a disparity in two-dimensional positions of the given same feature in the at least two images and the corresponding camera poses from which the at least two images are captured;
• and
• estimate the given plane based on the poses of the same features.