ARIN® procedure for the normalization of multitemporal remote images through vegetative pseudo-invariant features

摘要

A method was developed to normalize multitemporal remote images based in vegetative pseudo-invariant features (VPIFs), as following: 1) defining the same parcel for each selected VPIF in each multitemporal image; 2) extracting the VIPF spectral bands data for each image; 3) calculating the correction factor (CF) for each image band to fit it to the same expected values, normally for each band the average of the series; 4) obtaining the normalized images by transforming each original image band through the corresponding CF linear functions. We have validated ARIN using a series of six GeoEye-1 satellite images taken over the same Southern of Spain scene, from early April to October. Citrus orchards (CIT), riparian trees (POP), olive orchards (OLI) and Mediterranean forest trees (MFO) were the VPIFs chosen, among others. The VPIFs spectral band correction factors (CFs) to implement the ARIN linear normalization procedure largely varied among spectral bands for any given image and among images for any given spectral band. For the ARIN normalized images, the range and standard deviation of any spectral bands and vegetation indices values were considerably reduced as compared to the original images, regardless the VPIF or the combination of VPIFs selected for normalization, which proves the method efficacy. Moreover, ARIN method was easier and efficient than the absolute calibration QUAC method, and somehow similarly efficient as the highly tunable FLAASH, in which solar position and weather calibration parameters are required. ARIN® software was developed to automatically achieve the previously described procedure.