Abstract:  In order to improve the practicability of water content detection in crop based on computer vision technology， the method of image processing on the mobile internet was proposed and studied， in which， mobile phone is responsible for plant leaf image acquisition and the results showing， and the server running the background algorithm including image analysis and water content detection. Combination algorithm of Homomorphic filtering\|MSR was adopted to eliminate the natural conditions affected by uneven light and color distortion， a variety of features on color and texture were extracted， and the water detection model was established by principal component analysis and regression method. The tested results with maize leaves showed that， the mean relative error was 3645 4%， the variance was 2121 4， and the detection system worked properly. This method had some advantages， such as convenient to obtain image， getting the results in real time， the acceptable detection error， and updating and extension of algorithm was transparent to the user， which was very useful for farmers and agricultural technical personnel.

Key words: moisture content detection, computer vision, the mobile internet, illumination enhancement, feature fusion