Speaker
Description
Typical farming tasks and farm environments are technically challenging for automated robots to replace labor. Successfully implementing automation in agriculture is challenging because there are many technical variables that vary from country to country, region to region, and crop to crop, and it is difficult to make it economically viable.
It is difficult to develop agricultural automation technology for agricultural products that are mainly produced in Korea because there are few cases of related research and development technologies. In particular, in the case of ginseng, there are many difficulties in developing related technologies because farming work has been mostly dependent on manpower and there are few cases of related technology development because it is mainly cultivated in Korea.
A labor-intensive task for ginseng farmers is the sorting and sorting of harvested ginseng, but the complexity and variety of ginseng shapes make it difficult to automate.
In this paper, we conducted a study to implement a robotic system that sorts and stacks ginseng (Panax ginseng) placed on a work table by a robot. We developed a deep learning-based artificial intelligence to determine the gripping site for ginseng in a robotic system that performs the task of picking up randomly placed ginseng on a work table and stacking them in a specified position.
We propose a deep learning technique for predicting phage location and orientation based on individual background segmentation of ginseng. The appearance of ginseng, which consists of rhizome, main root, secondary root, and root hair, varies greatly among individuals depending on the shape and size of the secondary root and root hair. Therefore, we designed a pipeline that predefines the centers of the rhizome and main root of ginseng as gripping points and simultaneously estimates the grippable orientation information in two-dimensional images from the mask regions of the rhizome and main root output from the instance segmentation deep learning model. Through the experiment of estimating gripping points and orientation information for harvested ginseng, we confirmed that the proposed AI has sufficient performance to be utilized in a robotic system for ginseng sorting.
This work was supported by Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry (IPET) through (High Value-added Food Technology Development Program), funded by Ministry of Agriculture, Food and Rural Affairs (MAFRA) (321049-05).
References
[1]ALLEN, W., 2022. What is a pick and place robot? [WWW Document]. URL https://6river.com/what-is-a-pick-and-place-robot/
[2]K. He, G. Gkioxari, P. Dollár, and R. Girshick, “Mask R-CNN,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 42, no. 2, pp. 386–397, 2020, doi: 10.1109/TPAMI.2018.2844175.
| Keywords | Deep Learning, ginseng, AI, Robot, automation |
|---|
