Use of Variogram Parameters in Analysis of Hyperspectral Imaging Data Acquired from Dual-Stressed Crop Leaves

Abstract

Abstract:A detailed introduction to variogram analysis of reflectance data is provided, and variogram parameters (nugget, sill, and range values) were examined as possible indicators of abiotic (irrigation regime) and biotic (spider mite infestation) stressors. Reflectance data was acquired from 2 maize hybrids (Zea mays L.) at multiple time points in 2 data sets (229 hyperspectral images), and data from 160 individual spectral bands in the spectrum from 405 to 907 nm were analyzed. Based on 480 analyses of variance (160 spectral bands × 3 variogram parameters), it was seen that most of the combinations of spectral bands and variogram parameters were unsuitable as stress indicators mainly because of significant difference between the 2 data sets. However, several combinations of spectral bands and variogram parameters (especially nugget values) could be considered unique indicators of either abiotic or biotic stress. Furthermore, nugget values at 683 and 775 nm responded significantly to abiotic stress, and nugget values at 731 nm and range values at 715 nm responded significantly to biotic stress. Based on qualitative characterization of actual hyperspectral images, it was seen that even subtle changes in spatial patterns of reflectance values can elicit several-fold changes in variogram parameters despite non-significant changes in average and median reflectance values and in width of 95% confidence limits. Such scattered stress expression is in accordance with documented within-leaf variation in both mineral content and chlorophyll concentration and therefore supports the need for reflectance-based stress detection at a high spatial resolution (many hyperspectral reflectance profiles acquired from a single leaf) and may be used to explain or characterize within-leaf foraging patterns of herbivorous arthropods.

Using hyperspectral imaging to determine germination of native Australian plant seeds

“…The research objective is therefore to identify portions of the wavelength spectrum, in which seeds show a significant and measurable change in certain parts of the examined reflectance spectrum and associated that change in reflectance with certain traits, such as, germination (yes/no). A wide range of classification methods have been used as part of using reflectance data to characterize seeds and food products; these classification methods include support vector machine (SVM) [32], variogram analysis [33,34], partial least square analysis [35], and linear discriminant analysis (LDA) [36]. LDA is based on discriminant functions, which are linear combinations of features (in this case reflectance values in spectral bands) and with one function for each target class.…”

Use of variogram analysis to classify field peas with and without internal defects caused by weevil infestation

“…In which a, b, and c are fitted coefficients and D denotes the lag distance and F(D) is the semi-variance at lag distance, D. Although other regression fits are typically used (Nansen, 2012), an important advantage of the regression fit in Eq. (2) is a very high level of regression convergence.…”

“…Spatial structure analysis based on geostatistics (variogram analysis) is considered one of the most powerful and robust approaches to spatial data analysis (Isaaks and Srivastava, 1989), and recent studies have shown how variogram parameters derived from high-resolution reflectance data can be used to detect different traits in a range of target objects (Nansen, 2011(Nansen, , 2012Nansen et al, 2010a;Nansen et al, 2010b;Nansen et al, 2009;Nansen et al, 2010c). In the variogram analysis (PROC VARIOGRAM) of reflectance data at 782 nm, we used the following variogram settings: (1) lag distances = 1, and (2) number of lag intervals = 10.…”

“…Consequently, it is possible to characterize the spatial structure of reflectance values in a single spectral band (using variogram analysis) and determine to what extent a given quality trait of the target object is associated with a particular spatial data structure. While use of variogram analysis is well-described in geosciences, landscape ecology, oceanography, and other large-scale life science applications, it is only recently that this approach has been applied to high-resolution hyperspectral imaging data (Nansen, 2011(Nansen, , 2012Nansen et al, 2010a;Nansen et al, 2009). The assumption behind variogram-based analysis is that the spatial structure of reflectance values in individual spectral bands can be described by three fitted variogram parameters, and that these variogram parameters vary significantly among classes of target objects.…”

“…Linear discriminant analysis (Fisher, 1936) has been used successfully in many hyperspectral image based classifications of food products (Park et al, 2007;Gcdmez-Sanchis et al, 2008;Nansen et al, 2008;Gowen et al, 2009;Liu et al, 2010;Kalkan et al, 2011;Shahin and Symons, 2011). An alternative classification approach is possible, when hyperspectral imaging data are acquired, because each pixel is associated with a relative coordinate (x and y) within the image cube (Nansen, 2012). Consequently, it is possible to characterize the spatial structure of reflectance values in a single spectral band (using variogram analysis) and determine to what extent a given quality trait of the target object is associated with a particular spatial data structure.…”