WATERCORE FEATURES FOR SORTING RED DELICIOUS APPLES: ASTATISTICAL APPROACH
1999
Watercore is a serious internal defect found in most apple cultivars. X-ray imaging has shown potential for
nondestructive detection of watercore, however, the major challenge is to decide which features should be used for fruit
classification. A solution to this problem is feature selection based on their rank ordering. This article describes a
stepwise feature selection procedure. The performance of the selected features was tested using the Bayes classifier.
Spatial features (area, intensity) and (cosine, wavelet) transform coefficients were evaluated for their discriminating
power. The spatial features performed better than the transform features. A linear Bayesian classifier with three input
features (fruit area in the segmented image, mean intensity of fruit in the original image, and 10th harmonic of the
discrete cosine transform) achieved an accuracy of 79%.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
20
Citations
NaN
KQI