Skip to main content
ARS Home » Midwest Area » East Lansing, Michigan » Sugarbeet and Bean Research » Research » Publications at this Location » Publication #312898

Title: Detection of defect in pickling cucumbers using hyperspectral imaging

Author
item Lu, Renfu

Submitted to: Michigan State University Cucumber Reporting Session
Publication Type: Experiment Station
Publication Acceptance Date: 12/7/2014
Publication Date: 12/9/2014
Citation: Lu, R. 2014. Detection of defect in pickling cucumbers using hyperspectral imaging. Michigan State University Cucumber Reporting Session. p. 22-30.

Interpretive Summary:

Technical Abstract: Pickling cucumbers are susceptible to damage due to adverse growth condition, improper harvest timing, and inappropriate harvesting and postharvest handling operations. There are typically five to 10 percentages of harvested pickling cucumbers that are not suitable for pickling and hence should be removed before brining. Currently, both manual and machine vision methods are being used to sort and grade pickling cucumbers for surface color, size, shape and/or surface blemishes. Internal defect in cucumbers may occur in the forms of mechanical damage, soft or hollow tissue due to abnormal growth, pest infestation, etc. An automated inspection technology is thus needed for rapid assessment of internal quality of pickling cucumbers. In this research, a laboratory online hyperspectral imaging system, coupled with a light emitting diode (LED) configuration, was used to simultaneously acquire reflectance (400-700 nm) and transmittance (700-1,000 nm) images for normal and defective pickling cucumbers under the two online speeds of 81.3 mm/s and 165 mm/s at an imaging rate of 2 ms per scan. Three hundred ‘Journey’ pickling cucumbers, hand harvested from a commercial field, were imaged before and after they were subjected to mechanical damage. Defective cucumbers were artificially created by applying two different levels of rolling load to the fruit. Two waveband selection methods, i.e., minimum redundancy-maximum relevance (MRMR) and principal component analysis (PCA), were used to select the optimal wavebands from the acquired hyperspectral images for discriminating defective cucumbers from normal ones. The optimum two-waveband ratios selected by MRMR and PCA were then used for classification of normal and defective cucumbers. It was found that the two-waveband ratio of 887/837 nm selected by MRMR with Mahalanobis discriminant analysis outperformed PCA in differentiating defective cucumbers from normal ones, with the overall accuracy of greater than 94% for the two conveyor speeds. The overall classification accuracies were lowered to about 80% when the cucumbers were classified into three grades of normal and minor and severe defect. The waveband ratio method is effective in differentiating normal and mechanically damaged cucumbers, and it could be implemented for online sorting and grading of cucumbers for internal defect.