Torna alla Home Page Consiglio per la ricerca in agricoltura e l'analisi dell'economia agraria Search in the site...  

European Charter for Researchers      HR EXCELLENCE IN RESEARCH
CREA - Via Po, 14 - 00198 ROMA
P.IVA: 08183101008 - C.F.: 97231970589
Tel: +39 06 478361 - Fax: +39 06 47836320 -
Posta Elettronica Certificata:


freccina You are here: Home->Publications->Datasheet

Publication datasheet
Hyperspectral Imaging Characterization Of Agricultural Topsoil Copper Concentration
Antonucci, F.; Menesatti, P.; Canali, E.; Giorgi, S.; Maienza, A.; Stazi, S. R.
ITA, ita
CIGR Workshop on Image Analysis in Agriculture. 26-27 August 2010, Budapest, Hungary
Kind of publication:
Abstract in Italian:
Abstract in English:
Soil characterization in agriculture represents an important tool to plan its productivity and the quality of the resulted products. It is conventionally performed adopting physical-chemical analyses that classify soil samples delivered to specialized laboratories. The objective of this research is to develop a hyperspectral imaging system to estimate the concentration of copper in topsoil, as an alternative to the standard chemical analyses. Hyperspectral Imaging is a technique of high technological and methodological complexity but with elevated informative content. Sensors collect hyperspectral information as sets of images. The images were acquired within two range of the electromagnetic spectrum: visible-near infrared (VIS-NIR) and near infrared (NIR). These were consequently combined to yield a three dimensional hyperspectral cube for processing and analysis. To carry out and compare the chemical analyses soil samples were primarily air-dried, removing and discarding after sifting the fraction > 2 mm. Metal soil samples were prepared by adding 20 ml of copper sulphate solution (CuSO4•5H2O) to the test soil to concentrations ranging from 1 to 1000 mg of copper per kg of soil step 50 mg. The twenty amended soil samples obtained were oven-dried at 65°C for 48 h to produce samples with the same moisture. Samples were placed into a Duraplan® borosilicate optical-glass Petri dish, 3 for each concentration. Petri dishes were randomly scanned using VIS-NIR and NIR spectrophotometers and spectral data were acquired. Partial Least Squares regressions (PLS) were performed with the software Matlab 7.5 on all the collected spectral arrays, in order to test the ability of the proposed model to quantify the different concentrations of copper content. PLS is a soft-modeling method to built predictive models when the factors are many and highly collinear. The procedure emphasizes the prediction of the responses not necessarily trying to understand the relationship between variables. Results indicate that the correlation between predicted values and the observed chemical values is highly significant with a coefficient of determination (R˛) in the test of 0.95 for the VIS-NIR range and of 0.82 for the NIR. This method investigates whether the system can help in explaining heavy metal interactions with other soil properties. This could lead to significant benefits for soil remediation, survey and precision farming, representing a rapid and non-destructive alternative to classical soil analysis for the detection and characterization of copper topsoil concentration.

Logo mySQL Logo PHP