Design of a System for Obtaining the Diameter of Tree Trunks Using Depth Vision

In order to improve the autonomous navigation and operation efficiency of orchard robots, this study proposes an improved YOLOv5 algorithm to identify the trunk of the fruit tree, and combines with the depth camera for accurate positioning. By combining the SENet attention mechanism module with the residual module in the network, we obtain the improved SE-Res module, which can enhance the extraction of useful feature information and compress useless feature information. After experimental verification, the accuracy of the improved YOLOv5 model is increased by 2.38 percentage points, the recall rate is increased by 0.84 percentage points, the frame rate is reduced by 0.99 frames/s, and the mAP is increased by 0.05 percentage points. Experimental results show that the method can accurately identify and locate fruit trees in the autonomous navigation and fertilization of orchard robots, so as to improve operation efficiency and ensure operation quality, and realize the intellectualization and rationalization of fruit tree fertilization in the orchard environment.