A review on the recent developments in vision-based apple-harvesting robots for recognizing fruit and picking pose
Yanqiang Zhang, Na Li, Lijie Zhang, Jianfeng Lin, Xiao Gao, Guangyi Chen
Abstract
Fast and accurate recognition of fruit and picking pose is important for vision-based apple-harvesting robots in complex orchard environments. Despite significant improvement in these fields, the widespread use of harvesting robots in orchards is yet to be reported. This study was aimed at reviewing the state-of-the-art vision-based harvesting robots, focusing on recent development in detecting fruit and picking poses. Notably, this study briefly reviews vision-based harvesting robots in the aspects of picking success rate, time, and environment to summarize the bottlenecks and future developmental trends of robots. Furthermore, a detailed recognition algorithm review on the visual devices, recognition algorithm classification, and construction as well as improvement ideas of the mainstream recognition algorithms is provided. Additionally, this study deeply reviews the pose estimation algorithms to comprehensively explain the mechanism and classification of the algorithms and the representation of the fruit picking pose, thereby summarizing and analyzing the pose estimation for better use in agriculture. This study provides an in-depth analysis of the algorithms for recognizing fruit and its picking pose for vision-based harvesting robots, which will aid in further developing the apple harvesting robots.