Recognition and Position Estimation of Pears in Complex Orchards Using Stereo Camera and Deep Learning Algorithm

Siyu Pan, Ayanori Yorozu, Akihisa Ohya, Tofeal Ahamed

2023

Abstract

Complex orchards present difficulties for fruit-picking robots due to shadows, overlapping fruits, and obstructing branches, resulting in errors during grasping. To improve the robustness of fruit-picking robots in the complex environment, this study compared the performance of different types of deep learning algorithms (Mask R-CNN, Faster R-CNN, and YOLACT) for pear recognition under different conditions (high and low light). Additionally, the ZED2 stereo camera with the algorithm of the highest precision for estimating the position of separating and aggregating pears. For pear recognition, the mAPs of Mask R-CNN were 95.22% and 99.45%, Faster R-CNN were 87.90% and 87.52%, YOLACT were 87.07% and 97.89% in the validation and test set. For position estimation, the mean error of separating pears was 0.017m, the standard deviation was 0.015m and the goodness of fit reached 0.896; The mean error of aggregating pears were 0.018m and the standard deviation was 0.021m and the goodness of fit reached 0.832. A pear recognition and positioning system was developed by ZED2 stereo camera with deep learning algorithm. It aimed to generate precise bounding boxes and recognize pears in a complex orchard within the range of 0.1 to 0.5m. The mean error of separating pears and less than 0.27m for aggregating pears. This demonstrated the system’s capability to accurately position and differentiate between individual pears and clusters in challenging orchard environments.

Download


Paper Citation


in Harvard Style

Pan S., Yorozu A., Ohya A. and Ahamed T. (2023). Recognition and Position Estimation of Pears in Complex Orchards Using Stereo Camera and Deep Learning Algorithm. In Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO; ISBN 978-989-758-670-5, SciTePress, pages 632-639. DOI: 10.5220/0012179200003543


in Bibtex Style

@conference{icinco23,
author={Siyu Pan and Ayanori Yorozu and Akihisa Ohya and Tofeal Ahamed},
title={Recognition and Position Estimation of Pears in Complex Orchards Using Stereo Camera and Deep Learning Algorithm},
booktitle={Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO},
year={2023},
pages={632-639},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012179200003543},
isbn={978-989-758-670-5},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO
TI - Recognition and Position Estimation of Pears in Complex Orchards Using Stereo Camera and Deep Learning Algorithm
SN - 978-989-758-670-5
AU - Pan S.
AU - Yorozu A.
AU - Ohya A.
AU - Ahamed T.
PY - 2023
SP - 632
EP - 639
DO - 10.5220/0012179200003543
PB - SciTePress