Ref: C0457 Investigation of cobweb disease and green mold development and investigation of champignon caps treated with prochloraz-manganese using hyperspectral imaging Viktória Parrag, József Felföldi, Dániel Szöllősi and Ferenc Firtha, Corvinus University of Budapest, Department of Physics and Control, Somloi út 14-16, H-1118 Budapest András Geösel, Corvinus University of Budapest, Department of Vegetable and Mushroom Growing, Ménesi street 44, H-1118 Budapest Abstract White button mushroom (Agaricus bisporus) is a well-known edible mushroom, it is the most widely cultivated and harvested mushroom all over the world. Champignon can be attacked by different pests and diseases. Cobweb disease is a very common fungal infection, it caused serious losses in Europe, in the USA and in Australia. There are several fungus species causing cobweb disease, the most important is Dactylium dendroides. Different fungicides could be applied against this spoilage, prochlorazmanganese is a frequently used compound in chemical control. Hyperspectral imaging (HSI) became very popular in food quality measurements, because it is a rapid and non-destructive remote sensing technology and it can provide both spatial and spectral information from the object. This can be very useful if the distribution of a component or other feature is not homogeneous in the sample. The aim of our study was to detect fungal infections (cobweb disease and green mold) and fungicide treatment (prochloraz-manganese) on the surface of the mushroom caps. Samples were divided into 4 groups: treated with fungicide (1.), infected with dactylium (2.), infected with trichoderma (3.) and control (treated with distilled water) (4.). The caps were photographed and hyperspectral images were recorded in the wavelength range of 900-1700 nm (placed in the same determined direction on a sample holder). The area of the treatment was marked. The average spectra of the selected areas were saved, than a normalization algorithm and a Savitzky-Golay smoothing was carried out on the spectra to reduce the disturbing effects (e.g. the curvature of the mushroom cap). The classification of the infected and control samples with linear discriminant analysis was successful. Keywords: HSI, cobweb, dactylium, trichoderma Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 1/6
1 Introduction 1.1 Microscopic fungi cause serious problem for the mushroom industry For the food industry and consumers the quality of the mushroom is an essential question. The quality is perceptible as the appearance of the fruiting body (size, colour, maturity stage, development stage), which is affected by different effects and parameters like microbial diseases, mechanical damage, water content or natural browning (Aguirre, Frias, Barry-Ryan, & Grogan, 2008). The button mushroom can be attacked by different pests and diseases. In general, several fungi including cobweb, Trichoderma green mold, Verticillium dry bubble, Neurospora pink mold and Penicillium blue green mold cause diseases by growing in the substrate or on the surface of Agaricus bisporus. (Fletcher & Gaze, 2008) In recent years, cobweb diseases have been widespread and caused serious losses in Europe, the USA and Australia. (Fletcher & Gaze, 2008) The most typical symptom of this infection is the cobweb-like mycelial growth on the surface of the fruiting body or the casing soil. On the cap two types of spots appear: the first type is dark brown with indefinite contour, the second type is light brown with round contour (Győrfi, 2010). Trichoderma is also a well-known, aggressive fungal infection. The signs of the infection can appear in the compost and on the casing soil: its web is initially white, after a few days it turns into green. On the fruiting body little spots appear. The symptoms of trichoderma and dactylium infections are very similar (Győrfi, 2010). 1.2 Novel methods in food quality measurement In the last decade new methods became popular in the food industry, these techniques are rapid and non-destructive. Abbott (1999) summarized the present used quality measuring methods, including optical techniques like near infrared spectroscopy (NIRS) and hyperspectral imaging (HSI). Computer vision systems have beed developed for automated inspection and grading of mushrooms, for the measurement of the developmental stage and for the identification of bacterial discoloration.(brosnan & Sun, 2004) 1.3 Applications of HSI for the investigation of white button mushroom Several studies investigated white button mushroom using hyperspectral imaging. Gowen et al. (2008a) measured the quality of the mushroom slices in terms of moisture content, colour and texture. Taghizadeh, Gowen, and O Donnell (2009) predicted moisture content of the caps with Partial Least Squares Regression (PLSR) models. Gowen et al. (2008b) detected mechanical damage with Principal Component Analysis (PCA), than Gowen, Taghizadeh, and O Donnell (2009) used hyperspectral imaging and Linear Discriminant Analysis (LDA) for the early detection of freeze damage in white button mushrooms. Gaston and co-workers (2011) investigated the potential of Vis-NIR HSI to detect microbial spoilage, namely Pseudomonas tolaasii (brown blotch) and to discriminate it from mechanical damage. The detection and identification of microscopic fungi with hyperspectral imaging technique are relevant questions, since these infections cause great losses for the mushroom industry. 2 Materials and methods Mushrooms were purchased on a local market. Samples were divided into 4 groups: treated with fungicide (1.), infected with dactylium (2.), infected with trichoderma (3.) and control (treated with distilled water) (4.). The samples were stored under controlled conditions for 5 days (at appr. 6 C, the relative humidity was around 90%). On the samples we marked the area of the treatment or the infection. 50 μl of the fungicide, infectious solution or distilled water was uniformly distributed on the surface of every mushroom cap. Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 2/6
The RGB images of the mushroom caps were recorded by a digital camera (Canon EOS 450D). The hyperspectral images were recorded using a pushbroom HSI instrument (Headwall Photonics: Specim spectrograph, Xeneth InGaAs 14 bit sensor having 256*320 resolution) within the wavelength range of 900-1700 nm. The setup of optics finally resulted 5 nm spectral and 0.475 mm spatial resolution. Figure 1: Selection of the regions of interest on the hyperspectral images with CuBrowser The image processing system and the sensor were controlled by Argus hyperspectral software (Firtha et al., 2012). Before all measurement series, the signal of dark and bright standards were measured to calculate reflectance (spectral reflection factor) of samples. On the hyperspectral images of the samples the areas of the treatment (regions of interest) were selected manually using CuBrowser MATLAB software (Firtha & Éder, 2012) and the average spectra of these areas were saved (Figure 1.). To reduce the disturbing effects of non-homogeneous illumination (caused by the curvature of the mushroom caps), a simple normalization algorithm (the average of the intensities on the whole spectrum was substracted from the single intensity values) and a Savitzky-Golay smoothing was carried out on spectra. The classification was carried out using principal component analysis (PCA) and linear discriminant analysis (LDA). The statistical analysis was carried out using RStudio version 0.97.336 software. 3 Results and Discussion The intensity values in the regions of 980-960 nm and 1655-1700 nm and at 1272 nm were excluded from the analysis due to the high signal to noise ratio. The standard deviation of the results by the fungicide treated samples was very high, therefore we analysed the results of the infected and control groups. Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 3/6
0,15 0,1 0,05 Intensity 0 900 1000 1100 1200 1300 1400 1500 1600 1700-0,05-0,1-0,15-0,2-0,25 Wavelength, nm Figure 2: Average spectra of the normalized intensity values of the groups (control: red, dactylium: black, trichoderma: green) on the first day of the experiment The average spectra of the normalized intensity values of the groups on the first day of the experiment is shown on Figure 2. The effect of the normalization is visible, the baseline fluctuation (caused by the non-homogeneous illumination) is removed. Figure 3: Score plot of the principal component analysis trichoderma: green) (control: red, dactylium: black, Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 4/6
On the plot of the principal component analysis (PCA) (Figure 3.) the centers of the three groups are observable, although there are major overlappings between the groups. To the first principal component (PC) had the highest impact the intensity at 1459 nm, to the second PC the intensity at 1383 nm and to the third PC the intensity at 980 nm. Figure 4: Plot of the linear discriminant analysis of the infected and control samples (control: red, dactylium: black, trichoderma: green) The result of the linear discriminant analysis shows successful classification of the groups (Figure 4.). The discrimination between control and infected samples was spectacularly successful. The dactylium and trichoderma infected groups could be separated also using LDA, however the efficiency of the classification is much lower. 4 Conclusions The results of the LDA indicate, that fungal infections of button mushroom can be detected using hyperspectral imaging, even in early stages, when the symptoms are not perceptible on the surface of the tissue. Previous studies demonstrated the capability of HSI to detect mechanical or freeze damage, microbial spoilage (brown blotch) and cobweb disease. In this study the measurement was carried out on the first days of the infection, when the existence of the spoilage is not perceptible. 5 Acknowledgements This work was supported by TÁMOP 4.2.1./B-09/1/KMR-2010-0005. 6 References Abbott, J. a. (1999). Quality measurement of fruits and vegetables. Postharvest Biology and Technology, 15(3), 207 225. Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 5/6
Aguirre, L., Frias, J. M., Barry-Ryan, C., & Grogan, H. (2008). Assessing the effect of product variability on the management of the quality of mushrooms (Agaricus bisporus). Postharvest Biology and Technology, 49(2), 247 254. Brosnan, T., & Sun, D.-W. (2004). Improving quality inspection of food products by computer vision a review. Journal of Food Engineering, 61(1), 3 16. Firtha, F., Jasper, A., Friedrich, L., Felföldi, J. (2012): Hyperspectral qualification of aged beef sirloin. CIGR-AgEng, International Conference on agricultural engineering, Valencia, SPC-03: IV International workshop on Computer Image Analysis in agriculture. P1876, fulltext: http://cigr.ageng2012.org/images/fotosg/tabla_137_ C1876.pdf Firtha, F., Éder, G. (2012) CuBrowser MATLAB software (ftp://fizika2.uni-corvinus.hu/ffirtha/ Argus-CuBrowser.pdf) Fletcher, J. T., & Gaze, R. H. (2008). Mushroom Pest and Disease Control: A Color Handbook (p. 192). Elsevier. Gaston, E., Frias, J. M., Cullen, P., & Donnell, C. O. (2011). Hyperspectral Imaging for the Detection of Microbial Spoilage of Mushrooms. Oral Presentation MCF1004 at the 11th International Conference of Engineering and Food. Athens, Greece, May, 2011. Gowen, A., O Donnell, C. P., Taghizadeh, M., Gaston, E., O Gorman, a., Cullen, P. J., Downey, G. (2008a). Hyperspectral imaging for the investigation of quality deterioration in sliced mushrooms (Agaricus bisporus) during storage. Sensing and Instrumentation for Food Quality and Safety, 2(3), 133 143. Gowen, A., O'Donnell, C., Taghizadeh, M., Cullen, P. J., Frias, J. M., Downey, G. (2008b): Hyperspectral imaging combined with principal component analysis for bruise damage detection on white mushrooms (Agaricus bisporus). Journal of Chemometrics 22, 259 267. Gowen, A. a., Taghizadeh, M., & O Donnell, C. P. (2009). Identification of mushrooms subjected to freeze damage using hyperspectral imaging. Journal of Food Engineering, 93(1), 7 12. Győrfi, J.(2010): Gombabiológia, gombatermesztés (Mushroom biology, Mushroom cultivation). Mezőgazda Kiadó, Budapest (Chapter 10) Taghizadeh, M., Gowen, A., & O Donnell, C. P. (2009). Prediction of white button mushroom (Agaricus bisporus) moisture content using hyperspectral imaging. Sensing and Instrumentation for Food Quality and Safety, 3(4), 219 226. Proceedings International Conference of Agricultural Engineering, Zurich, 06-10.07.2014 www.eurageng.eu 6/6