Abstract:The application system of real-time image recognition and variable spraying was designed based on the virtual image real-time controller CVS—1456. The original images, which contained pea seedlings, soil background, weed of cephalanoplos segetum, etc, were collected in normal sunlight. The color models of original images were analyzed and real-time weed recognition was realized based on R—B color features by using LabVIEW software and IMAQ Vision toolbox. The Canny algorithm was employed to detect weed edges, and three characteristic parameters of target weed, namely area, density, centroidal position, were extracted to provide positioning evidences for variable spraying. The random tests verified the accuracy and reliability of the purposed cephalanoplos segetum recognition method from complex background images based on R—B color features, in which the average right recognition rate was 83.5%, mean square deviation 0.066.