Automatic Extraction of Regions of Interest from Hyperspectral Imagery Based on Active Contour Model

Abstract

The invention discloses an active contour model-based method for automatically extracting regions of interest from hyperspectral images. First, a standard mixed vector of spectral reflectance of the object of interest is established according to the spectral reflectance vector of the known pure object pixel; and then, the calculation is performed. The correlation coefficient of the spectral reflectance standard mixture vector of the object of interest and the spectral vector of each pixel in the hyperspectral image to be processed is obtained to obtain the pixel correlation coefficient deviation matrix; finally, the C-V is constructed with the pixel correlation coefficient deviation matrix Active contour model, and then use the finite difference method to solve the model to realize the pixel extraction of the region of interest. The test results of the embodiments show that the present invention can achieve significantly better extraction results than the traditional C-V model with fewer iterations.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an active contour model-based automatic region-of-interest extraction method for a hyperspectral image that can distinguish object types and has a fast calculation speed. Background technique [0002] At present, hyperspectral remote sensing technology is developing towards higher spatial resolution, higher spectral resolution and higher temporal resolution, which makes the data volume of hyperspectral images increase exponentially. Take the AVIRIS (AirborneVisible / Infraed Imaging Spectrometer) type hyperspectral image as an example, it has 224 continuous bands, each band image contains 512×614 pixels, each pixel occupies 16bit, and its storage space More than 140M bytes. Therefore, high-efficiency coding of hyperspectral remote sensing images is one of the methods to alleviate the contradiction between information acquisition and transmission of hyperspectral image data. [000...

AI Technical Summary

Problems solved by technology

[0004] Existing methods for extracting regions of interest from hyperspectral images mostly use typical methods such as the maximum displacement method, some important bit-plane displacement methods, and one-by-one bit-plane displacement methods, but these methods do not consider the geological features of hyperspectral images. (such as vegetation, water bodies, rock mines, soil, urban artificial targets, etc.), it is possible to divide pixels belonging to different ground objects into the same area of ​​interest, and the calculation speed is slow, which affects the further practical application of hyperspectral images

Images