收藏切换
Structure-space-guided Method for the Segmentation of Mango Leaves and Lesions in Complex In-field Scenes
收藏切换
PDF
Danyang WANG1, 2, Weihong LIANG1, 2, Lu YE1, 2, Hailiang LI1, 2, Qian ZHENG1, 2, Guixiu HUANG3
Chinese Journal of Tropical Crops | 2026, 47(2) : 498 - 507
Less
收藏切换
Chinese Journal of Tropical Crops | 2026, 47(2): 498-507
Plant Protection & Bio-safety
Structure-space-guided Method for the Segmentation of Mango Leaves and Lesions in Complex In-field Scenes
Full
Danyang WANG1, 2, Weihong LIANG1, 2, Lu YE1, 2, Hailiang LI1, 2, Qian ZHENG1, 2, Guixiu HUANG3
Affiliations
  • 1.Institute of Scientific and Technical Information, Chinese Academy of Tropical Agricultural Sciences/Key Laboratory of Applied Research on Tropical Crop Information Technology of Hainan Province, Haikou, Hainan 571101, China
  • 2.Hainan Tang Huajun Academician Workstation, Haikou, Hainan 571101, China
  • 3.Environment and Plant Protection Institute, Chinese Academy of Tropical Agricultural Sciences, Haikou, Hainan 571101, China
Published: 2026-02-25 doi: 10.3969/j.issn.1000-2561.2026.02.020
Outline
收藏切换

To address mango leaf disease recognition in complex in-field scenes, where background interference and simultaneous infections across multiple leaves lead to feature recognition difficulties and multi-task segmentation problems, this paper proposes G-Mask2Former, a structure-space-guided model that achieves collaborative segmentation of leaf instances and lesion semantics. The method jointly learns leaf and lesion segmentation within a shared feature space, and leverages a leaf-structure prior to impose spatial constraints on the lesion predictions. First, based on Mask2Former, we use multi-scale pixel representations and mask attention to extract effective features of mango leaf diseases in complex scenes; by constructing pseudo-instances of lesions, we achieve leaf instance segmentation and lesion semantic segmentation within a single model, thereby simplifying model deployment and inference. Second, through a leaf structure-space soft gating module, the model is guided to focus on leaf regions, reducing interference from complex environments. Finally, we conduct experiments on a mango leaf disease dataset collected in complex in-field scenes and perform comparative analyses against other models. The proposed method achieves superior overall performance over classical baselines for segmenting mango leaves and diseases in complex in-field scenes. On the test set, the average precision (AP) for leaf instance segmentation reaches 87.37%, while the mean Intersection-over-Union (mIoU) for lesion semantic segmentation reaches 82.83%. The proposed method would provide an effective solution for mango leaf disease segmentation in complex environments and facilitates quantitative disease assessment.

structure-space guidance  /  mango  /  leaf disease  /  instance segmentation  /  semantic segmentation
Danyang WANG, Weihong LIANG, Lu YE, Hailiang LI, Qian ZHENG, Guixiu HUANG. Structure-space-guided Method for the Segmentation of Mango Leaves and Lesions in Complex In-field Scenes[J]. Chinese Journal of Tropical Crops, 2026 , 47 (2) : 498 -507 . DOI: 10.3969/j.issn.1000-2561.2026.02.020
Year 2026 volume 47 Issue 2
PDF
55
16
Cite this Article
BibTeX
Article Info
doi: 10.3969/j.issn.1000-2561.2026.02.020
  • Receive Date:2025-10-10
  • Online Date:2026-06-26
  • Published:2026-02-25
Article Data
Affiliations
History
  • Received:2025-10-10
  • Accepted:2025-11-20
Funding
Affiliations
    1.Institute of Scientific and Technical Information, Chinese Academy of Tropical Agricultural Sciences/Key Laboratory of Applied Research on Tropical Crop Information Technology of Hainan Province, Haikou, Hainan 571101, China
    2.Hainan Tang Huajun Academician Workstation, Haikou, Hainan 571101, China
    3.Environment and Plant Protection Institute, Chinese Academy of Tropical Agricultural Sciences, Haikou, Hainan 571101, China
References
Share
https://castjournals.cast.org.cn/joweb/rdzwxb/EN/10.3969/j.issn.1000-2561.2026.02.020
Share to
QR

Scan QR to access full text

Cite this article
BibTeX
Citations
表12种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏
  • BibTeX
  • EndNote
  • RefWorks
  • TxT