Fruit Stalk Recognition and Picking Point Localization of New Plums Based on Improved DeepLabv3+

Xiaokang Chen, Genggeng Dong, Xiangpeng Fan (), Yan Xu (), Tongshe Liu, Jianping Zhou and Hong Jiang
Additional contact information
Xiaokang Chen: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China
Genggeng Dong: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China
Xiangpeng Fan: Agricultural Information Institute, Chinese Academy of Agricultural Sciences, Beijing 100081, China
Yan Xu: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China
Tongshe Liu: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China
Jianping Zhou: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China
Hong Jiang: College of Intelligent Manufacturing and Modern Industry, Xinjiang University, Urumqi 830017, China

Agriculture, 2024, vol. 14, issue 12, 1-16

Abstract: Among the challenges posed by real orchard environments, where the slender new plum fruit stalks exhibit varying postures and share similar coloration with surrounding leaves and branches, the significant obscuration caused by leaves leads to inaccurate segmentation of the fruit stalks, thereby complicating the precise localization of picking points and other related issues. This paper proposes a method for new plum fruit stalk recognition and picking point localization based on the improved DeepLabv3+ model. Firstly, it employs the lightweight MobileNetv2 as the backbone feature extraction network. Secondly, the Convolutional Block Attention Module (CBAM) is integrated into the decoder to enhance the model’s ability to extract key features of the fruit stalks. Moreover, dense atrous spatial pyramid pooling (DenseASPP) is utilized to replace the original ASPP module, thereby reducing segmentation leakage. Finally, a picking point localization method is designed based on a refinement algorithm and an endpoint detection algorithm to meet the specific picking demands of new plum, identifying the endpoints along the skeletal line of the fruit stalks as the optimal picking points. The experimental results demonstrate that the mean intersection over union (MIoU) and mean pixel accuracy (MPA) of the enhanced DeepLabv3+ model are 86.13% and 92.92%, respectively, with a model size of only 59.6 MB. In comparison to PSPNet, U-Net, and the original DeepLabv3+ model, the MIoU improves by 13.78, 0.34, and 1.31 percentage points, while the MPA shows enhancements of 15.35, 1.72, and 1.38 percentage points, respectively. Notably, with the endpoint of the fruit stalk’s skeletal structure designated as the picking point for new plums, the localization success rate reaches 88.8%, thereby meeting the requirements for precise segmentation and picking point localization in actual orchard environments. Furthermore, this advancement offers substantial technical support for the research and development of new plum harvesting robots.

Keywords: deep learning; semantic segmentation; attention mechanism; picking point location (search for similar items in EconPapers)
JEL-codes: Q1 Q10 Q11 Q12 Q13 Q14 Q15 Q16 Q17 Q18 (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/2077-0472/14/12/2120/pdf (application/pdf)
https://www.mdpi.com/2077-0472/14/12/2120/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jagris:v:14:y:2024:i:12:p:2120-:d:1527500

Access Statistics for this article

Agriculture is currently edited by Ms. Leda Xuan

More articles in Agriculture from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().