The Research of SPAD in Rice Leaves Based on Machine Learning

The purpose of this study was to explore the relationship between the leaf image parameters and SPAD values under RGB color space to provide theoretical basis for rapid and accurate acquisition of plant SPAD values and theoretical guidance for scientific fertilization. Based on Android, the image of rice leaves was acquired, and the color feature parameters of rice leaves were collected interactively by using JNI and OpenCV. BP neural network algorithm was applied to train the sample data of rice color characteristic parameters, and R, G and B were systematically acquired as input variables to predict the SPAD values of rice. The results showed that the RGB of the area which were circled manually were obtained using Android. And the root mean square error of which were 1.086, 1.413 and 0.8383, respectively; meanwhile, compared with other models, the BP neural network prediction model proposed in this study had the lowest root-mean-square error, which was 1.6679. The study can accurately obtain the RGB and SPAD values without contacting the leaf organs, providing a new universal method for the rapid and non-destructive measurement of SPAD values.