Apple Classification Based on Information Fusion of Internal and External Qualities

Abstract

Apple classification plays an important role in improving the sales of apples. Based on both the internal and external qualities of an apple, in this paper, we propose to classify apples by DS theory-based information fusion. Soluble solid content is selected for apple internal quality detection. Making near-infrared spectroscopy nondestructive testing, principal component analysis -Martensitic distance method and multiple Scattering correction are used to preprocess the spectral data collected. Partial least squares prediction model is established with genetic algorithm selecting the wavelength characteristics. The color, shape, diameter and defect of apple are taken as the important indexes of external quality detection, and the sample images are analyzed and studied. The RGB color model and HSI color model commonly used in image processing are introduced. Selecting the median filtering algorithm for image denoising, the prediction model of support vector machine is established. In order to effectively avoid the classification error caused by the traditional hard classification using threshold and to make the detection result more accurate, the analysis of uncertain factors was introduced in the aspect of apple classification, and DS evidence theory was used to fuse the prediction results of internal and external quality.