Estimation of moisture content in solanaceae seedlings using hyperspectral
Background: This research was performed to develop moisture content model for solanaceae seedlings, such as chili pepper and tomato, based on hyperspectral imagery.
Methods: After exposing to high temperature, the reflectance of chili (n=45) and tomato (n=45) seedlings were calculated using the hyperspectral imagery, and the moisture content of all seedlings was measured. Then the predicting models for estimating moisture content were developed with PLS Regression analysis by using the two factors
Results: As a result, the chilli model showed 0.68 of R2, 1.43% of RMSE and 1.61% of RE, which indicate accuracy and precision respectively. The tomato model showed 0.74 of R2 2.77% of RMSE and 3.09% of RE. Combining all samples (n=90), the solanaceae model showed 0.67 of R2 2.53% of RMSE and 1.61% of RE. Finally the full-cross validation showed 0.59 of R2, 2.83% RMSE and 3.17% of RE.
Discussions: Accuracy and precision of the tomato model was slightly better than those of the chilli model. This might be affected by widely distributed variation in tomato seedlings. The solanaceae model, which is combined with chilli and tomato, showed lower accuracy than each single model, however, the precision was higher than the tomato model. This is because that tomato samples distribution was located in chilli distribution and this affected the performance better. Finally, it is considered that both models can be significantly used to estimate moisture content, as gradients of trend line are almost same and intersected.
Conclusion: It is considered that the accuracy and precision of the estimated models possibly can be improved, if samples are under a lot of stress. The improved models will be utilized as the basis for developing low-priced sensors.