| CPC B25J 9/1697 (2013.01) [B25J 9/163 (2013.01); B25J 9/1653 (2013.01); G06T 7/75 (2017.01); G06T 17/00 (2013.01)] | 18 Claims |
|
1. A method of identifying objects to be moved by an industrial automatic robotics picking system, comprising:
obtaining one or more first two-dimensional (2D) image data;
obtaining one or more first three-dimensional (3D) data, the one or more first 2D image data and the one or more 3D data corresponding to a first scene;
generating a first aligned model by aligning the one or more first 2D image data and the one or more first 3D data;
obtaining a second 2D image data;
obtaining a second 3D data, the second 2D image data and the second 3D data corresponding to a second scene;
generating a second aligned model by aligning the second 2D image data and the second 3D data;
generating a composite model by aligning the first aligned model and the second aligned model to a frame;
generating an intermediate mask model from the composite model by comparing the one or more first 2D image data to the second 2D image data and by comparing the one or more first 3D data to the second 3D data;
augmenting the intermediate mask model by applying 2D image processing techniques to the intermediate mask model;
generating a 2D object image data and a 3D object image data by applying an augmented mask model to the second 2D image data and the second 3D data, the 2D object image data and the 3D object image data identifying one or more objects associated with the second 2D image data and the second 3D data and that are not associated with the one or more first 2D image data and the one or more first 3D data; and
applying statistical analysis to identify pick points for picking the one or more objects identified in the 2D object image data and the 3D object image data.
|