Journal of Agricultural Machinery Engineering ISSN:2799-8673(Print) 2799-8819(Online) 농업기계공학

This study aimed to investigate the potential of hyperspectral imaging for non-destructively prediction of tensile strength of the tomato fruits, and the impact on the prediction models according to the ROI (Region of interests) selection methods in hyperspectral imaging. A total of 180 tomatoes grown in a hydroponic greenhouse was used as samples. Tomatoes images were obtained by using hyperspectral camera (400-1,000 nm), and then the tensile strength of tomatoes was measured by using UTM (Universal testing machine). Two ROI selection methods were considered in this study: first, whole area of tomatoes was selected as ROI, and second, 15 points in whole ROI were selected randomly. The spectrum extracted by two methods was calculated as average value. And then, the averaged spectrum and the measured tensile strength were used as input data to develop the tomatoes tensile strength prediction models based on PLSR (Partial least square regression). As a result, there were no significant difference between whole ROI (R2=0.691, RMSE=2.967 N) and random ROI (R2=0.690, RMSE=2.971 N) methods. Therefore, we confirmed that hyperspectral imaing and PLSR model could be used to non-destructively predict the tensile strength of tomatoes. However, it will be necessary to carry out more experiments to achieve a more comprehensive prediction in the future.