| Apples are prone to different degrees of mechanical bruise during the picking and transportation of apples,such as improper handling of workers and bumps in the road.Apple’s mechanical damage caused nutrient loss and huge economic losses.The time after bruised and bruise degree are common evaluation indexes of apple mechanical bruise,which has important practical significance for the system of apple quality classification.Based on hyperspectral imaging technology,this study collects data from bruise apple and analyzes it to obtain the following results:Extract the area of mechanical bruise to the apple.Based on the RELIEF algorithm,five sensitive wavelengths of 468.04 nm,549.96 nm,621.64 nm,683.08 nm and 974.92 nm are obtained according to their characteristic weights.Subsequently,a pixel-level apple bruise recognition logistic regression classification model is established at the selected wavelengths,and the recognition rate was 99.12%.Morphological manipulation results in a smoother bruise area.The results show that the method can effectively detect the bruise site.Estimate the time after mechanical bruise to the apple.Re-sampling the bruise area of apple,obtaining the characteristic spectrum of the bruise area of apple.Establishing an artificial neural network model about its spectrum and estimate the time after bruised unknown apple,the accuracy up to 91%.Apple mechanical bruise degree discrimination.Base on the existing mechanical bruise time estimation model,the degree of apple mechanical bruise is migration learned,and the artificial neural network model for the degree of mechanical bruise of apple is obtained.The model can determine the damage of apple bruise by up to 88%. |
