3D Gaussian Splatting-Based Localization Algorithm for Occluded Tea Shoots
Haoran Wu, Jingui Zhang, Yongjun Yang, Daoxiong Gong · 2025
In automated tea picking, accurately localizing tea shoot picking points is crucial for improving picking quality. However, existing localization methods struggle to handle the occlusion of tea shoots in the complex and dynamic environment of tea gardens. To address this issue, this paper proposes a 3D Gaussian Splatting-based localization algorithm for tea shoot picking points. The algorithm enhances the 3D reconstruction details of tea shoots by integrating the Mip-3DGS algorithm. By computing the 3D coordinates of picking points from multiple viewpoints, it effectively utilizes rich 3D information to achieve precise localization of both unobstructed and occluded tea shoot picking points. Experimental results show that the proposed method achieves a localization success rate of 90.04% with an average absolute error of 0.88 mm, meeting the precision requirements for automated tea picking in complex tea garden environments.