In a processing system, video images of moving objects are processed to model
the objects in a 3D computer model. Video from multiple cameras is processed to
separate objects from their shadows, and to test whether an object is made up of
separate objects, which are then modelled separately. Each object is modelled using
vertical planes whose bases approximate the object's ground footprint, using planes
based on object surface planes identified in the image data, or using a single
vertical plane. Pixel data from the video images is rendered onto the planes in
the models. The video for rendering is selected based on the viewer's viewing direction,
the camera viewing directions, and quality characteristics of the cameras and image
data. If the viewer's viewing direction is close to vertical or a plane of an object,
a schematic of the objects' positions is displayed. To account for image data from
different cameras being used, successive images are tested for visual discontinuous,
and are modified if necessary. Information indicating the accuracy/reliability
of the rendered image is displayed.
