【Objective】In order to improve the efficiency and utilization of conventional unmanned aerial vehicle (UAV) spraying in fertilizer and pesticide applications, an variable rate UAV spraying system was developed based on image recognition.【Method】Median filter was applied to the images for denoising. K_means clustering algorithm was then used to segment the UAV images to extract 22 texture features and the color of non-crop region. Support vector machine (SVM) classifier was designed for classification. According to the 17 selected characteristic parameters, the non-crop region was recognized through the SVM classifier with Radial basis function (RBF) as the kernel function. Finally, precision spraying was achieved with controllable nozzles based on the recognition results.【Result】The recognition accuracy reached up to 76.56%. In undisturbed wind farm, the reduction rate reached 32.7% with the threshold P of 10%.【Conclusion】This research can serve as reference guides for application of precise spraying control technology in agricultural aviation.