In order to monitor the growth of citrus, a method to obtain the complete surface information of single citrus fruit by inspection robot was proposed. LabelMe was used as a tool to mark citrus fruits and segmented branches and trunks, and the model of Mask R-CNN was used to train and recognize them. The 3D information of citrus fruits and segmented branches and trunks can be extracted from the images with depth information taken by Kinect V2 camera. According to the growth of citrus fruit and the occurrence of diseases and insect pests, the surface of citrus fruit was divided into regions, and integrated system was used to describe the information acquisition rate. The CCD camera connected to the end of the manipulator was used to acquire the surface information of the citrus fruit. Three positions of the manipulator were proposed to obtain the complete surface information of the citrus fruit. A prototype inspection robot was built and simulated under the robot operating system (ROS). The process of citrus surface information acquisition in the laboratory environment was proposed. In the laboratory environment, the complete surface information of citrus was obtained by the inspection robot. The experimental results showed that the closer the ratio of horizontal to vertical stems of citrus fruit was to 1, the more complete the outer surface information obtained by photographing was. When the photographing mode of three photographing positions was used, the higher the success rate and work efficiency of avoiding obstacles can be ensured while the citrus surface integrity rate was obtained. The citrus surface information acquisition rate can reach 94.21%, and the average movement time of the manipulator was 86.57s. The research result provided a method for human to get the surface condition of citrus, monitor the growth of citrus and control the diseases and insect pests.