Identification and quantification of cyst nematode in sugar beet seeds by hyperspectral NIR imaging


  • Vermeulen, P. , Tossens, A. , Amand, O. , Dardenne, P. , Baeten, V. & Fernández Pierna, J.A. (2009). Identification and quantification of cyst nematode in sugar beet seeds by hyperspectral NIR imaging. Poster in: 14th International Conference on Near Infrared Spectroscopy (ICNIRS): Breaking the dawn, Bangkok - Thailand, 7-13 November 2009.
Type Poster
Year 2009
Title Identification and quantification of cyst nematode in sugar beet seeds by hyperspectral NIR imaging
Event name 14th International Conference on Near Infrared Spectroscopy (ICNIRS): Breaking the dawn
Event location Bangkok - Thailand
Label U15-1491
Recnumber 613
Event date 7-13 November 2009
Endnote Keywords cyst|nematode|sugar beet|hyperspectral NIR imaging|identification|quantification|
Abstract [Introduction] The damage caused by nematode on the sugar beet root lead to a yield reduction and is related to the cyst number. The current work, carried out in collaboration with SESVANDERHAVE Company, aims to assess by hyperspectral NIR imaging the presence of cyst nematode on sugar beet root. The objective of the study is to discriminate between cyst, root and soil support as well as to quantify the cyst nematode presence. [Materials and Methods] For this experiment, 30 plants of sugar beet with different level of resistance, were grown in a soil support spread in plastic plates: 20 plants were infested with cyst, 10 plants were not infested and were used as control. The number of cyst nematode was previously counted by optical microscopy at SESVANDERHAVE. Then, to cover the root area, 4 images by plant were acquired with the hyperspectral imaging system installed at CRA-W. The instrument used is a MatrixNIR® Chemical Imaging System (Malvern instruments Ltd) recording sequential images with an InGaAs array detector (240x320 pixels) active in the 900-1700 nm range, that means 76 800 spectra per image. The data treatment was carried out under Matlab 7.5.0 (R2007b). For the identification of cysts, 30 spectra have been selected for each type of structure that can be found in the plate (root, cyst and soil support). Spectra were preprocessed with 1st Derivative and a PCA was performed. For the quantification, five spectral libraries (cyst, root, soil support, plastic plate and teflon support) have been built by selecting pixels in the images of 4 plants. Those libraries were used for the building of discrimination equations in order to identify cysts from root, soil support particles and background (plastic and teflon support). SVM, Support Vector Machines, was used as classification method for the construction of these models. Three equations have been built: "background vs. soil support + root + cyst", "soil support vs. root + cyst" and "root vs. cyst". Then, these equations have been applied successively to all the pixels in the images of the 30 plants in order to build a mask, by isolating the cysts, and then calculate the number of pixels detected as cysts by surface unit. [Results and Discussion] Regarding the identification, the PCA performed on the preprocessed data shows that the cysts can be clearly discriminated from the root. Regarding the quantification, a correlation of 0.65 has been calculated between the number of cysts counted on the roots and the number of pixels recognized as cyst nematode by the NIR imaging. It has to be noted that cysts are counted on the total surface of the glass plate while NIR imaging analyses the total surface by acquisition of 4 images. To avoid some overlapping between images taken on the same plant, additional images will be acquired covering the total surface of the glass plate using a different optical configuration. This study showed the potential of the hyperspectral NIR imaging to discriminate the cysts from the root and the soil support in a sugar beet root as well as to quantify the number of cysts.
Fichier
Caption U15-1491-vermeulen-2009.pdf
Authors Vermeulen, P., Tossens, A., Amand, O., Dardenne, P., Baeten, V., Fernández Pierna, J.A.

Team