Predicting Bruise Susceptibility Of Fruit Based On Optical Properties

Fruit damage is an important factor affecting its quality, which not only bring the appearance of defects, but will bring the loss of nutrients and bacterial infections deeper flaw. Fruit damage not only to reduce its quality, affecting consumer desire to buy, will bring great economic losses to growers or distributors. Fruit damage occurs in fruit production, harvesting, packing, transportation, storage and other chains. Prior knowledge of the characteristics and the possibility of fruit damage in all aspects of processing, to provide information to take appropriate protective measures for reducing fruit damage has important value. Bruise susceptibility is an important index to evaluate the possibility of bruise caused by the fruit’s resistance to external impact, which defined as the volume of damage resulting from unit energy absorbed, expressed in the units of mm3J-1. Destructive measurement methods have been investigated to estimate the bruise susceptibility of fruit in the past years, which are suitable for small samples of fruit, and it is difficult to meet the rapid detection of large samples. In order to overcome these problems, Visible and Short-Wave Near-Infrared (Vis/SWNIR) and Hyperspectral Scattering Imaging (HSI) were used to develop a fast and nondestructive contactless model which was used to predict the bruise susceptibility of fruit.The main contents in this paper are as follows:1. Study the prediction methods of bruise susceptibility of Apple base on Vis/SWNIR, and applied to the predict bruise susceptibility of’Golden Delicious’apples. Vis/SWNIR spectra between 400 and 1,100 nm were acquired for’Golden Delicious’apples. Successive projections algorithm (SPA) was used to select the main feature. In order to make full use of spectral information, the removed features was random add in the Sub-model which is constructed by the Partial least squares (PLS). Finally, Selective ensemble learning based on feature selection (SELFS) was used to develop prediction model for bruise susceptibility of apples. This research showed that the Vis/SWNIR technique could be used for determination the bruise susceptibility of apples, and SELFS model had stronger robustness and generalization ability.2. The objective of this research was to investigate HIS to predict the bruise susceptibility of apples. Because of HSI has a large amount of information, so the effective feature extraction method for targeted information has important significance. By comparing the optical properties parameters(μa,μs’), Modified lorentzian distribution (MLD), Generalized gaussian distribution (GGD) and Mean spectral (MS), the MS had a higher prediction precision for bruise susceptibility, and had obvious advantages to Vis/SWNIR.3. In order to further improve the prediction precision for bruise susceptibility apple, comprehensive Vis/SWNIR and HSI of the two kinds of sensor information fusion technology was used to establish bruise susceptibility model for’Golden Delicious’apples. The result showed that the model of multi-sensor information fusion had significantly improving the prediction accuracy than a single sensor. This showed that multi-sensor information fusion technology had improved the prediction ability of model.