Study On Non-destructive Detection Of Light Mechanical Damage Of Fruits Based On Hyperspectral Imaging

Since ancient times,China has built up a country with agriculture,a large agricultural production country and a large population.Grain,fruit and vegetables are the three main products of agricultural production.Fruit and vegetables are rich in nutrition and are indispensable in people’s daily life,but they contain more water,not easy to store,difficult to transport,and easily rotten and deteriorate in storage and transportation.Therefore,in the process of transportation and sale,we need to pretreat them and classify them according to their quality.The traditional classification and classification method based on human eye detection has the disadvantages of easy fatigue,high labor intensity,long labor time,lack of objectivity and so on,which cannot meet the increasingly rich and efficient market demand.In recent years,a hyper-spectral imaging technology combining spectral information and image information has been applied to nondestructive testing of agricultural products.This technology,because it can reflect more comprehensive information of agricultural products,has attracted more and more researchers’ attention in the field of non-destructive testing of agricultural products.Hyper-spectral imaging technology and appropriate spectral reconstruction algorithm can be used to analyze the quality of agricultural products from two aspects:spectral information and image information.The spectral information can reflect the composition and internal structure of agricultural products.The image information can reflect the appearance of agricultural products,surface pollution,defect,size and color,etc.,and can detect the quality of agricultural products by extracting the characteristic wavelengths of the hyper-spectral images of agricultural products.Because hyper-spectral images belong to massive and high-dimensional information,data redundancy often exists.In later stage,spectral reconstruction and spectral analysis need to be dimensionality reduction.Principal component analysis,band ratio algorithm and support vector machine method are commonly used in the processing of hyper-spectral image data.The advantage of the principal component analysis is that the loss of information is less;the disadvantage is that the amount ofinformation is large and the processing speed is slow.The band ratio algorithm can make the difference between the bands more obvious,reduce the influence of the uneven illumination,make up the shortage of the single band image information,and supervise the classification support to the quantity machine method,the maximum likelihood method and so on.A small number of bands can be detected in the interesting area,but the classification effect needs further study.In this study,the common apples and Yali pears in the market were selected as the testing object.After the injury of the artificial machinery,the hyper-spectral imaging technology was used and the three data processing algorithms were combined to detect the damage of the fruit,and the analysis methods of the three hyper-spectral images were analyzed and compared.Because of the different physical structure of different fruits and other chemical components,such as fruit water content and hardness,the same kind of dimensionality reduction data processing algorithm is different for the hyper-spectral image processing of different fruits.The experimental results show that the fast and effective method to detect Apple damage is the band ratio algorithm,and the principal component analysis is the best way to detect the damage of Yali pear.