Abstract: Hyperspectral imaging technology was used to detect total flavonoid content distribution map in loquat (Eriobotrya japonica) leaves at different growth stages (Young leaf,Mature leaf,Old leaf).Firstly,120 loquat leaves were used to collect hyperspectral image data cube and determine total flavonoid concentrations.Secondly,the optimal model (r=0.8996, RMSE=1.72 mg/g) for predication of total flavonoid was built using Least-Squares Support Vector Machine method.Finally,the calibration model was used to predict the total flavonoid content of each pixel in the hyperspectral image.Distribution maps of total flavonoid content in three leaves (Young leaf,Mature leaf,Old leaf) were calculated.Our results indicated that the level of total flavonoid for Young leaf,Mature leaf and Old leaf were in an increasing order.Higher level of total flavonoid can be noticed in the leaf mesophyll regions near the primary and secondary veins,while the lower levels occurred at the margin area of the leaves.This research provided a method to determine major constituent of food and agricultural products.

Key words: hyperspectral imaging technology, loquat leaf, total flavonoid content, support vector machine, distribution