Abstract: Abstract: This study aims to propose a universal method to detect the chlorophyll content of green plant leaves using different smart phones, thereby improving the prediction accuracy and practicability of RGB（Red, Green, Blue） images. A low-cost and portable auxiliary shoot device was also designed with a built-in active light source to reduce the influence of environment light and camera angle on the imaging quality. A color correction was used to reduce the color difference of the collected images. Three smart phones with different brands and prices were selected to carry out the experiments on sugarcane leaves. After that, some prediction models were established for the leaf chlorophyll content. A remote diagnose system under the WeChat applet was then developed to realize in situ, real-time and nondestructive detection of chlorophyll content in plant leaves. An auxiliary shoot device was designed in a shape of semi-sphere, with an inside coating of white diffuse reflective material, particularly for the even white light during image acquisition. A small shooting window was opened on the top, where eight white LEDs were distributed inside. The plant leaf was placed flat at the bottom base to capture a well-controlled leaf image in the enclosed chamber. A color checker with the 24 standard colors (X-Rite Colorchecker Classic Nano) was needed in the chamber in advance to calculate the color correction matrix using polynomial regression for each smart phone. An automatic image processing procedure was conducted to extract the RGB values of the color blocks and the leaves. Subsequently, a total of 39 color features were extracted from the leaves. A Pearson correlation analysis was implemented between the leaf color features and chlorophyll content. The results showed that most of the color features presented a strong correlation greater than 0.8 for the individual smart phone in the original and color-corrected images, indicating a relatively fewer improvement. But, the color correction of hybrid images with the chlorophyll content better improved than that of the original. The color feature (B-G-R)/(B+G) was a correlation of 0.84, indicating the highest improvement of 89% before correction. A Principal Component Analysis (PCA) was then adopted to reduce the dimension of the color features. The first five and fifteen principal components were used to build the Multivariate Linear Regression (MLR) and the Support Vector Regression (SVR) models for the individual and all smart phones. The color correction showed that there were only a few contributions to predict the chlorophyll content with the individual smart phone. More importantly, no color correction was needed if only one sensor was used to capture the RGB images. The iPhone 8P behaved the highest of R2 to reach about 0.76 and 0.79 for MLR and SVR models, respectively. But, an outstanding improvement was made for the hybrid image set with all smart phones. The SVR model performed better prediction accuracy and stability. The SVR model reached a mean R2 of 0.721 4 under the five-fold cross-validation, with an increase of 14.6%, while the mean RMSE was 0.328 8 mg/g, with a decrease of 13.3% before the color correction. Meanwhile, the lowest standard deviation of R2 was only 0.0042 in the general SVR model, compared with the rest, indicating the highest stability. The prediction was integrated with a remote server, where a WeChat applet was developed to provide a convenient way of leaf chlorophyll content prediction for the users with their own smart phones. This model can be further extended to various green plants.