Title: Detection of Underdeveloped Hazelnuts from Fully Developed Nuts by Impact Acoustics Authors
|Onaran, Ibrahim - BILKENT UNIVERSITY|
|Yardimici, Yasimin - MIDDLE EAST TECH UNIV|
|Cetin, Enis - BILKENT UNIVERSITY|
Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 15, 2006
Publication Date: December 1, 2006
Repository URL: http://naldc.nal.usda.gov/download/1631/PDF
Citation: Onaran, I., Pearson, T.C., Yardimici, Y., Cetin, E. 2006. Detection of underdeveloped hazelnuts from fully developed nuts by impact acoustics. Transactions of the ASABE. Vol. 49(6):1971-1976. Interpretive Summary: The acoustic emissions from inshell hazelnuts as they impact with a steel plate were analyzed for their ability to distinguish nuts with fully developed kernels from those with underdeveloped kernels. The analysis included examination of the acoustic signals in the time domain as well as the frequency domain. Classification accuracies as high as 97% were achieved by this simple and low cost method. The system has a potential to sort nuts at rates up to 40 per second. Nuts with underdeveloped kernels are of lower value and can be more likely to contain aflatoxin. Thus, this method should give hazelnut producers and exporters a means to produce a higher quality and safer product.
Technical Abstract: Shell-kernel weight ratio is a vital quality measurement of hazelnuts as it helps to identify nuts with underdeveloped kernels. Nuts containing underdeveloped kernels may also contain mycotoxin producing molds, which can cause cancer and is heavily regulated in international trade. A prototype system was set up to detect underdeveloped hazelnuts by dropping them onto a steel plate and processing the acoustic signal generated when kernels hit the plate. A feature vector obtained by a combination of line spectral frequencies and time-domain maxima describing both time and frequency nature of the impact sound was extracted from each impact sound signal and used by a support-vector machine for classification. A classification accuracy of 97% was achieved in experimental studies.