Kalman filtering approach to multispectral/hyperspectral image classification

Abstract

Linear unmixing is a widely used remote sensing image processing technique for subpixel classification and detection where a scene pixel is generally modeled by a linear mixture of spectral signatures of materials present within the pixel. An approach, called linear unmixing Kalman filtering (LUKF), is presented, which incorporates the concept of linear unmixing into Kalman filtering so as to achieve signature abundance estimation, subpixel detection and classification for remotely sensed images. In this case, the linear mixture model used in linear unmixing is implemented as the measurement equation in Kalman filtering. The state equation which is required for Kalman filtering but absent in linear unmixing is then used to model the signature abundance. By utilizing these two equations the proposed LUKF not only can detect abrupt changes in various signature abundances within pixels, but also can detect and classify desired target signatures. The performance of effectiveness and robustness of the LUKF is demonstrated through simulated data and real scene images, Satellite Pour I'Observation de la Terra (SPOT) and Hyperspectral Digital Imagery Collection (HYDICE) data.