Abstract
A procedure to achieve the semi-automatic relative image normalization of multitemporal remote images of an agricultural scene called ARIN was developed using the following procedures: 1) defining the same parcel of selected vegetative pseudo-invariant features (VPIFs) in each multitemporal image; 2) extracting data concerning the VPIF spectral bands from each image; 3) calculating the correction factors (CFs) for each image band to fit each image band to the average value of the image series; and 4) obtaining the normalized images by linear transformation of each original image band through the corresponding CF. ARIN software was developed to semi-automatically perform the ARIN procedure. We have validated ARIN using seven GeoEye-1 satellite images taken over the same location in Southern Spain from early April to October 2010 at an interval of approximately 3 to 4 weeks. The following three VPIFs were chosen: citrus orchards (CIT), olive orchards (OLI) and poplar groves (POP). In the ARIN-normalized images, the range, standard deviation (s. d.) and root mean square error (RMSE) of the spectral bands and vegetation indices were considerably reduced compared to the original images, regardless of the VPIF or the combination of VPIFs selected for normalization, which demonstrates the method’s efficacy. The correlation coefficients between the CFs among VPIFs for any spectral band (and all bands overall) were calculated to be at least 0.85 and were significant at P = 0.95, indicating that the normalization procedure was comparably performed regardless of the VPIF chosen. ARIN method was designed only for agricultural and forestry landscapes where VPIFs can be identified.
Highlights
Remote sensing observations are usually instantaneous and are affected by many factors, such as atmospheric conditions, sun angle, viewing angle, dynamic changes in the soil and plant– atmosphere system, and changes in the sensor calibration over time [1,2]
ARIN Procedure and Software The procedure developed for the relative normalization of multitemporal images consists of the following steps: 1) selecting one or several vegetative pseudo-invariant features (VPIFs); 2) defining the same parcel or parcels for each selected VPIF in each multitemporal image; 3) extracting the VPIF spectral band data for each image; 4) calculating the correction factors (CFs) for each image band to fit each band value to the average value of the image series; and 5) obtaining the normalized images by transforming each band through CF linear functions
The NDVI evolution varied significantly in vegetative features (VVFs), while it was relatively stable in VPIFs
Summary
Remote sensing observations are usually instantaneous and are affected by many factors, such as atmospheric conditions, sun angle, viewing angle, dynamic changes in the soil and plant– atmosphere system, and changes in the sensor calibration over time [1,2]. The goal of radiometric corrections is to remove or compensate for all of the above effects Exceptions to this procedure include corrections for actual changes in the ground target to retrieve surface reflectance (absolute correction) or to normalize the digital counts obtained under the different conditions and to establish them on a common scale (relative correction) [2]. Absolute radiometric corrections (ARC) make it possible to relate the digital counts in satellite image data to radiance at the surface of the Earth. This relation requires sensor calibration coefficients, an atmospheric correction algorithm and related input data among other corrections [2]. Absolute surface reflectance retrieval may not always be practical [2]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
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
