Excessive chemical spraying can lead to waste, damage rice plants, and reduce yield. To resolve this problem, the precise spraying of interplant areas through targeting methods was investigated in this study. The touch-sensing data for identification and positioning was extracted by the trigger rule of binocular diffuse reflection sensors. Wavelet transform decomposition is used to extract time-domain and frequency-domain features from identification data. These combined features build a rice plant identification model based on support vector machine. The model could determine whether the light sensing system was triggered by rice plants or weed. Positioning data was divided into several regions based on the light sensing system. Then, maximum voltage was extracted from the positioning regions. The relative position of the rice plants was determined by fitting previous coordinates and maximum values. Finally, the nozzle position was corrected by the lateral movement of robot arms to achieve the precise targeted spraying. Field performance tests showed that the targeted intermittent spraying had 12.78% and 17.66% less spraying coverage rates in the non-targeted root and interline areas, respectively, and 21.9% more spraying coverage rates in the targeted interplant areas. The overall spraying dosages were 7.81 and 9.41 L for the targeted intermittent and non-targeted continuous spraying, respectively; the spraying dosage saved was 18%. The targeting application system proposed in this paper can achieve a significant reduction in chemical dosage. This research results provide a practical method for reduced herbicide application in rice paddies.