Camera system and methods for extracting 3D model of viewed object
AssigneeMartin Marietta Corporation
InventorRichard A. McClain
Life of this patent
Estimating the term…
Abstract
A system for extracting a three-dimensional model of an object viewed in a selected scene, includes at least one camera for acquiring scene information, each from a different spatial-temporal aspect of the total scene. A digitizer converts the scene video and directional information to digital data, for at least temporary storage in a mass memory associated with a computer for data processing. An interactive display provides a means by which a human operator can view the stored scene information and select at least one scene object for which the computer will extract a 3D model, for subsequent use by a designated output device.
Claims20
A system for generating data representative of a three-dimensional object from a plurality of views of a scene including said object, and for displaying a view of said scene, said system comprising:
- camera means for generating plural views of said scene, for generating data representative of said scene;
- display and selection means coupled for receiving said data representative of said scene, for generating a display of said plural views, and for receiving, and converting into data, user inputs relating to the locations of vertices associated with said object, the identification of faces of said object associated with said vertices, and geometric constraints including, where applicable, line lengths, line directions, line parallelism, and point or vertex coincidence, for thereby generating user-processed scene data, said user inputs being subject to errors so that, in said user-processed scene data, the locations of said vertices and lines may not exactly correspond among said views;
- processing means coupled to said camera means and to said display and selection means, for (a) processing said user-processed scene data by means of a dependency graph to reduce the number of said geometric constraints, (b) combining all said user-processed scene data relating to parameters of said scene and object into a single vector, and forming a single likelihood function from said vector, which likelihood function includes the locations and parameters of said camera means;
- (c) optimizing said likelihood function to obtain optimized data representing the most likely combination of said camera, vertex and line locations and parameters;
- and
- display means coupled for receiving said optimized data, for displaying a two-dimensional representation of said three-dimensional object.
A method for extracting a three-dimensional model of an object in a selected scene, comprising the steps of:
- acquiring, from a plurality of spatial-temporal views, each of which spatial-temporal views is associated with a viewing site, digital data representing at least that portion of the scene containing the object to be extracted, said digital data not including information relating to the location of at least one of said viewing sites;
- at least temporarily storing the acquired digital scene information data;
- displaying to a system user selected scenic information to allow the user to select from the displayed data a feature of the object to be extracted;
- processing a constrained least-squares optimization of at least the stored digital data for the feature selected via the display, to obtain a set of solution data describing the most likely location of said selected object based upon the most likely locations of said viewing sites;
- and
- storing the solution data set to provide output data for at least a display of the extracted object.
Dependencies are inferred from the claim text.