IMAGE FUSION FOR DATA WITH DIFFERENT SPECTRAL RANGES

摘要

Images with low spatial resolution or limited spectral range can decrease the accuracy on remote sensing analysis, such as target detection for small substances and classification in homogenous area. However, collecting image data with high quality on spatial and spectral information is not always available for every studies and projects because of the economical and physical limitations. Image fusion methods combine the information of two or more data, and produce an image with sufficient information. Image fusion method was developed to enhance the series of bands of multispectral image using a panchromatic image on a satellite system first, but now it is applied to the multi-sensor images with different platforms and even the images with different wavelength ranges. In this study, image fusion was performed based on spectral unmixing model to enhance spectral and spatial information of images. Nonnegative matrix factorization technique also adopted to estimate optimized high frequency for image fusion. The unmixing-based image fusion method was applied to airborne data taken in different time, and data from unmanned aerial system with two different sensor were also used for the image fusion in this study. In this study, images with different spectral range taken from different sensor/time were spatially and spectrally enhanced, and the fusion results can improve the result of remote sensing analysis with its high spectral and spatial information.