Abstract
This paper addresses three problems in the field of hyperspectral image segmentation: the fact that the way an image must be segmented is related to what the user requires and the application; the lack and cost of appropriately labeled reference images; and, finally, the information loss problem that arises in many algorithms when high dimensional images are projected onto lower dimensional spaces before starting the segmentation process. To address these issues, the Multi-Gradient based Cellular Automaton (MGCA) structure is proposed to segment multidimensional images without projecting them to lower dimensional spaces. The MGCA structure is coupled with an evolutionary algorithm (ECAS-II) in order to produce the transition rule sets required by MGCA segmenters. These sets are customized to specific segmentation needs as a function of a set of low dimensional training images in which the user expresses his segmentation requirements. Constructing high dimensional image segmenters from low dimensional training sets alleviates the problem of lack of labeled training images. These can be generated online based on a parametrization of the desired segmentation extracted from a set of examples. The strategy has been tested in experiments carried out using synthetic and real hyperspectral images, and it has been compared to state-of-the-art segmentation approaches over benchmark images in the area of remote sensing hyperspectral imaging.
Highlights
Even though quite a few methods have been proposed in the last two decades for hyperspectral image classification, addressing this task continues to be a challenging problem
We propose a complete methodology for conducting a multi-stage classification procedure that adheres to this last approach that is, it uses a segmentation step followed by a subsequent classification stage
> f th, f th, if rj, otherwise, where s0i denotes the updated spectrum and si the original one for cell i; sij represents the spectrum of a neighboring cell j; n is the number of neighboring cells considered for the state update; r j is the distance between cell j and Pi ; wij corresponds to the weight for the spectrum of cell j; wi is the weight for the spectrum of cell i and, f r (r j ) represents a function that assigns weights depending on the distance between the points, r j
Summary
ECAS, version I [58] and II [59,60], are evolutionary approaches for the generation of Cellular. They provide the parameters or transition rules that govern their operation and behavior. EAs work on a population of solutions instead of a single point This makes them less likely to be misled by local optima when searching for the optimum set of transition rules. A new population of possible solutions is generated every generation by means of the selection and variation of the members of the population as a function of their evaluation according to a fitness function. This process is iterated until an individual with sufficient fitness (candidate solution) is found or when a previously selected computational limit is reached. A weighted spectral average of the spectrum of cell i and that of some of its neighbors is performed considering all the spectral bands
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
