Research on segmentation algorithm of rice canopy image

Rice nitrogen status is an important index for rice growth evaluation，rice yield and quality estimation. It has important significance for the precise diagnosis of rice nitrogen status and effective management of rice production. Computer image processing technology is able to detect rice nutrition information accurately and nondestructively on a real time basis，which has widely applied in precision agriculture. The accuracy and speed of rice canopy segmentation directly affects the performance of model and efficiency of nutrition information detection. The image characteristics of RGB color channels of three typical rice canopy images in paddy field were analyzed in this research. A G-R color channel based OSTU method was proposed to segment rice canopy images from the edge，color and region of images. Experiments were conducted to compare the time required and segmentation effects with other methods. Results showed that in the 40 images，the success rate of proposed segmentation method reached 92.5% with the minimum time required