Diffusion maps and transfer subspace learning
View Patent ↗A method, computer program product, and customized image processing system provide aided target recognition (AiTR) using manifold transfer substance learning (MTSL). The method includes receiving image data. The method includes performing manifold learning technique comprising diffusion mapping on the received image data to transform the received image data. The method includes applying a transfer subspace learning technique comprising Transfer Fishers Linear Discriminative Analysis (TrFLDA) to the transformed data to recognize an object within the image data.
1. A method of aided target recognition (AiTR) capable of achieving high target recognition rates for cross-dataset conditions and cross-domain applications, the method comprising:
accessing source data comprising high-dimensional images of a first object;
receiving target data from a sensor of an airborne platform of a second object that is unknown, the target data comprising high-dimensional images of the one or more unknown objects;
normalizing the target data;
computing a kernel matrix using a Euclidean distance between points and a Gaussian kernel;
performing a manifold learning step;
performing a transfer subspace learning step;
computing the source data in a lower-dimensional subspace;
computing the target data in a lower-dimensional subspace;
performing a classification step by computing a classification rate for the target data in the lower-dimensional subspace based on the source data in a lower-dimensional subspace to identify at least one object in the target data;
determining a geospatial location of a real-world object that corresponds to the identified at least one object in the target data; and
directing a targeting system to target the real-world object
wherein performing the manifold learning step comprises computing diffusion maps coordinates for the source data and coordinates for the target data using a Nystrom extension.
2. The method of claim 1 , wherein the image data is a selected of: (i) an electro optical image; (ii) an infrared image; and (iii) a synthetic radar aperture image.
3. The method of claim 1 , wherein:
the manifold learning technique comprises a nonlinear dimensionality reduction technique of diffusion maps based on a heat diffusion analogy; and
the transfer subspace learning technique comprises transfer Fisher's linear discriminative analysis.
4. The method of claim 1 , wherein performing the transfer subspace learning step comprises computing a projection matrix that solves an optimization problem.