Detection of Sitotroga cerealella (Olivier) infestation of Wheat Kernels Using Hyperspectral Reflectance

Abstract

Hyperspectral reflectance data were used to detect internal infestations of Angoumois grain moth, Sitotroga ceralella (Olivier), in wheat kernels. Kernel reflectance was measured with a spectroradiometer over a wavelength range of 350–2500 nm. Kernel samples were selected randomly and scanned every 7 d after infestation to determine the ability of the hyperspectral reflectance data to discriminate between infested and uninfested kernels. Immature stages of S. ceralella inside wheat kernels can be detected through changes in moisture, starch, and chitin content of the kernel. By using the spectrally-derived moisture variable (Log[1/R972nm]-Log[1/R1032nm]) and starch variable (Log[1/R982nm]-Log[1/R1014nm]), it was possible to discriminate between infested and uninfested wheat kernels with 100% classification accuracy based on 90% confidence intervals. Significant differences in the spectral reflectance between the infested and uninfested kernels were due to changes in moisture and starch content in wheat kernels. Three of the four chitin variables showed slight discrimination between the infested and uninfested wheat kernels based on 90% confidence intervals with 63.9%, 68.8%, 66.7%, and 41.6% classification accuracy of the three variables (Log[1/R1130nm]-Log[1/R1670nm]), (Log[1/R1139nm ]-Log[1/R1320nm]), (Log[1/R1202nm]-Log[1/R1300nm]), and (Log[1/R2046nm]-Log[1/R2302nm]), respectively. Spectral reflectance changes as a function of wheat kernel position relative to the spectroradiometer sensor did not differ significantly (P > 0.10).

Using Spatial Structure Analysis of Hyperspectral Imaging Data and Fourier Transformed Infrared Analysis to Determine Bioactivity of Surface Pesticide Treatment

“…Standard analytical approaches used in analysis of reflectance data include (see [9] for review): discriminant analysis [10,11,26], principal component analysis [3,21,[26][27][28], multi-regression approaches, like partial least square (PLS) [12,29,30], use of spectral band ratios (indices) [3,[31][32][33], decision trees [34], artificial neural networks [17,28,34], and support vector machines [8,35]. One common denominator in all of these analytical approaches is that they do not incorporate-or take advantage of -the spatial information available in a HI data.…”

Evolutionary refinement approaches for band selection of hyperspectral images with applications to automatic monitoring of animal feed quality

“…Using HSI technology has the following advantages over traditional microscopy: it is non-destructive, generally does not require much physical preparation of the target objects, and can provide real-time results. Hyperspectral imaging technology has been evaluated as part of machine vision based quality control of a wide range of food products, including meat [28,37,40,42], fruits and vegetables [8,11,20,21,30,31,35,38,48], grain and flour [6,13,14,16,39,47], and animal feed [5,17,34,36].…”

Machine Vision Detection of Bonemeal in Animal Feed Samples

“…Mainly due to growing public concern about contaminants and defects in food and feed production, reflection-based technologies are being used to develop machine vision systems for detection of defects and contaminants in a wide range of food products, including meat, 11,12 fruits and vegetables, [13][14][15][16][17][18] grain and flour, [19][20][21][22] and animal feed. [23][24][25] Pierna et al 24 collected reflectance data in the 900 to 1700 nm wavelength range (spectral bands in 10 nm increments) from animal feed particles and bonemeal fragments, and validation was based upon placement of individual feed particles in a cross surrounded by vegetal feed particles.…”