Apple stem/calyx and defect discrimination using DT-CWT and LS-SVM

This paper proposed a method for apple stem/calyx and defects discrimination by integrating the Dual Tree Complex Wavelet Transform (DT-CWT) and Least Squares Support Vector Machines (LS-SVM) method. The DT-CWT was used to decompose the apple images, and the feature vectors were generated by computing mean and standard deviation from the coefficients of individual wavelet subbands and the LS-SVM was used for classification. 85 apple images were tested, in which there were 25 stem and calyx images respectively and 35 defect images. Moreover, the influence of the DT-CWT decomposition levels on the classification rate was analyzed. The result showed that with 3-level DT-CWT the best classification result could be obtained, and an overall detection rate of 97.1% was achieved.