A Camera/Ultrasonic Sensors Based Trunk Localization System of Semi-Structured Orchards

Abstract

Semi-structured orchard environment has the characteristics of high complexity and strong uncertainty, which brings many challenges to the localization and navigation of agricultural automation equipment with limited sensors. This paper introduces a tree trunk localization system in a semi-structured orchard using camera/ultrasonic sensors fusion. In our method, the location of each tree trunk, which is relative to the robot, is obtained through the trunk tracking camera/ultrasonic detection system on the robot. And the location and direction of the robot are obtained through wheel odometer and IMU, respectively, rather than the traditional high-precision RTK-GPS. In order to reduce the error caused by odometer and IMU, several control points are introduced to revise the tree trunks’ location, of which positions are determinate. Experiments suggest that in a simulated environment and a real orchard, which both have a row-column spacing of $4.3m\times 3.4m$, the tree trunk localization error of our approach is within 0.3m.