Automatic target is a key technology in the field of target spray robot. In order to meet the requirements of precision spraying operation for large spacing crops, an automatic target method based on visual feedback was proposed. A hybrid vision structure, including single scene camera and single (or multi) eye-in-hand camera was adopted. The scene camera mounted in the front of robot body was used to pre-locate target crop and estimate robot’s moving speed, and algorithm about location of target crop’s centroid and estimating of robot’s moving speed was studied. The eye-in-hand camera mounted at the end of spray mechanical arm with nozzle was used to track and aim at target crop. A visual tracking method which used image moments as image features was presented, and some important issues, such as moment features selection, image Jacobian calculation and real-time estimation of target depth were studied, and using real-time estimation of target depth would obtain better tracking trajectory than commonly used fixed depth in Cartesian space and image space. Simulation results showed that the visual tracking method can fulfill the target task and had high control precision. In order to further verify the feasibility of the automatic target method, a simplified prototype containing one spray mechanical arm was built and target experiments were carried out in laboratory. The results showed that the position error of nozzle in X, Y and Z directions was less than or equal to 6.5mm with vehicle speed of 150~200mm/s. This study can provide reference for the development of target spray robot.