Apple Nir Spectral Classification Method

Min Li,Jin Cao,Linju Lu
DOI: https://doi.org/10.2991/icecee-15.2015.38
2015-01-01
Abstract:This paper proposed an apple near infrared spectral classification method. Red Fuji apples from Shandong and Shaanxi province, "Huaniu" apples from Gansu province were used as experimental materials. NIR data of three kinds of apples after preprocessing by wavelet soft threshold, was removed the noise and redundancy. Then the method of Principal Component Analysis (PCA) was used to reduce the dimension, and the Fisher Decision Analysis (FDA) was used for further feature extraction. Finally the K-Nearest Neighbor (KNN) classification was run, and K = 4. Through the experimental comparison, the method can achieve good feature extraction and classification of apples. The correct identification rate reaches above 96%. This method can realize different kinds of apples nondestructively, rapidly and accurately, which provides a new idea for near infrared spectral analysis technology.
What problem does this paper attempt to address?