SBIR-STTR Award

Laser radars (ladars) have been investigated for the detection, classification and terminal tracking of military targets. One area of concern is the performance of automatic target acquisition (ATA) algorithms on targets that are stochastically obscured by foliage. A program is proposed that will increase the understanding of target obscuration while increasing the set of non-proprietary ATA algorithms available to the weapon systems designer.The techniques developed during this program will benefit the field of computer vision. Specifically, robots that pick parts and perform quality control will be enhanced by the ability to effectively process occluded objects. Medical imaging will also benefit from the algotihms developed during this program

Image processing techniques derived from Partial Differential Equations (PDEs) are applied to ladar range images to resolve occlded targets. One approach will use model-driven anisotropic diffusion to replace an occluded area with the maximum a priori surface that is derived from the observable portions of the image. A second approach will align multiple ladar range images taken from different viewpoints and form a 3D surface from the combined data.

Keywords:
Ladar, Image Processing, 3d Surface Reconstruction, Anisotropic Diffusion