Abstract: Over or under-fertilization has normally resulted in environmental pollution or the reduction of fruit production, particularly in orchards. In this study, a precise target fertilization device with variable rates was developed for fruit trees in an orchard. The volume of the groove was automatically adjusted for the fertilizer discharging wheel, according to the diameter of the canopy of fruit trees, and the amount of fertilizer required by the target. The rotation speed of the fertilizer discharging wheel was continuously altered with the speed of the vehicle in the process of fertilization. The external groove wheel structure was adopted in the device to tailor the groove volume. A Lidar sensor was used to detect the canopy position of fruit trees in real time, thereby determining the starting and stopping position of fertilization. A hall sensor was used to detect the speed of the fertilization vehicle, and STM32F407VET6 microcontroller was used as the core to design the controller. Three kinds of granular fertilizers were selected as test materials, including urea, compound fertilizer, and organic compound fertilizer. The amount of fertilizer discharging was also calibrated under various speeds of fertilizer discharging wheels with different groove openings. The fertilizer discharging quantity of a single groove showed a negative linear relationship between the rotation speed of fertilizer discharging wheel, and the determination coefficient R2 was greater than or equal to 0.93. Some mathematical relationships were established between the target fertilizer amount of a single fruit tree and the rotation speed of fertilizer discharging wheel, the speed of fertilization vehicle, the volume of the groove of fertilizer discharging wheel, and the diameter of the canopy of the fruit tree. The control rules of rotation speed were also established for the fertilizer discharging wheel. An experiment of variable fertilization was performed on the citrus trees in a greenhouse, when the target fertilization amount of a single fruit tree was changed from 50 to 2 500 g. The results showed that the relative error of the maximum fertilization was 5.17% in a single fruit tree. Moreover, the average relative error was less than 3%, and the maximum coefficient of variation was 1.47%. Furthermore, the different granular fertilizers were accurately applied under the changing speed of the fertilization vehicle. The relative error of the maximum fertilization was 4.83% in a single citrus tree, the average relative error was less than 4.5%, and the maximum coefficient of variation was 6.96%, and the automatic fertilization was completed within the diameter of the canopy of the fruit tree. Consequently, the device can be expected to apply the fertilizer on the specific target, according to the diameter of the fruit tree canopy, suitable for a small or large amount of quantitative fertilization for the different kinds of granular fertilizer. Precise variable fertilization can contribute to different sizes of fruit trees.