Prediction of chlorophyll and anthocyanin contents in purple lettuce based on image processing

Abstract

Listen

Translate article

Abstract. The contents of chlorophyll and anthocyanin are important indexes for evaluating the growth status of purple lettuce. To quickly obtain the chlorophyll and anthocyanin contents of purple lettuce leaves, a predicting method based on image processing technology was proposed in this paper. A total of 60 leaf samples and 60 images were collected. 90 color features were obtained by calculating and combining each color component of RGB, HSV and L*a*b* color space. Then the obtained color features were correlated with chlorophyll and anthocyanin contents respectively. According to correlation coefficient results, 4 color features which correlated with chlorophyll content and 5 color features that were significantly correlated with anthocyanin content were selected separately. Multiple linear regression (MLR), support vector machine (SVM) and random forest algorithm (RF) were used to establish regression models to predict chlorophyll and anthocyanin contents. The model calibration and validation analysis showed that for chlorophyll, the SVM prediction model had the best accuracy among three models, which determination coefficients (R2) and the root mean square error (RMSE) were 0.5272 and 0.9100 mg/g respectively, and its prediction R2 and RMSE reached 0.3898 and 1.0406mg/g respectively; for anthocyanin, the RF prediction model had the best effect, which R2 and RMSE of calibration were 0.7505 and 0.0066 mg/g respectively, and its prediction R2 and RMSE reached 0.5709 and 0.0090mg/g respectively. The results showed that the SVM prediction model for chlorophyll content and the random forest prediction model for anthocyanin content could be used to guide the actual production of purple lettuce.