A review of visual perception technology for intelligent fruit harvesting robots

Abstract

With the development of smart agriculture, fruit picking robots have attracted widespread attention as one of the key technologies to improve agricultural productivity. Visual perception technology plays a crucial role in fruit picking robots, involving precise fruit identification, localization, and grasping operations. This paper reviews the research progress in the visual perception technology for fruit picking robots, focusing on key technologies such as camera types used in picking robots, object detection techniques, picking point recognition and localization, active vision, and visual servoing. First, the paper introduces the application characteristics and selection criteria of different camera types in the fruit picking process. Then, it analyzes how object detection techniques help robots accurately recognize fruits and achieve efficient fruit classification. Next, it discusses the picking point recognition and localization technologies, including vision-based 3D reconstruction and depth sensing methods. Subsequently, it elaborates on the adaptability of active vision technology in dynamic environments and how visual servoing technology achieves precise localization. Additionally, the review explores robot mobility perception technologies, focusing on V-SLAM, mobile path planning, and task scheduling. These technologies enhance harvesting efficiency across the entire orchard and facilitate better collaboration among multiple robots. Finally, the paper summarizes the challenges in current research and the future development trends, aiming to provide references for the optimization and promotion of fruit picking robot technology.