Improved Detection and Location of Small Crop Organs by Fusing UAV Orthophoto Maps and Raw Images DOI Creative Commons
Huaiyang Liu, Huibin Li, Haozhou Wang

et al.

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(5), P. 906 - 906

Published: March 4, 2025

Extracting the quantity and geolocation data of small objects at organ level via large-scale aerial drone monitoring is both essential challenging for precision agriculture. The quality reconstructed digital orthophoto maps (DOMs) often suffers from seamline distortion ghost effects, making it difficult to meet requirements organ-level detection. While raw images do not exhibit these issues, they pose challenges in accurately obtaining detected objects. detection was improved this study through fusion with using EasyIDP tool, thereby establishing a mapping relationship data. Small object conducted by Slicing-Aided Hyper Inference (SAHI) framework YOLOv10n on accelerate inferencing speed farmland. As result, comparing directly DOM, accelerated accuracy improved. proposed SAHI-YOLOv10n achieved mean average (mAP) scores 0.825 0.864, respectively. It also processing latency 1.84 milliseconds 640×640 resolution frames application. Subsequently, novel crop canopy dataset (CCOD-Dataset) created interactive annotation SAHI-YOLOv10n, featuring 3986 410,910 annotated boxes. method demonstrated feasibility detecting three in-field farmlands, potentially benefiting future wide-range applications.

Language: Английский

Improved Detection and Location of Small Crop Organs by Fusing UAV Orthophoto Maps and Raw Images DOI Creative Commons
Huaiyang Liu, Huibin Li, Haozhou Wang

et al.

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(5), P. 906 - 906

Published: March 4, 2025

Extracting the quantity and geolocation data of small objects at organ level via large-scale aerial drone monitoring is both essential challenging for precision agriculture. The quality reconstructed digital orthophoto maps (DOMs) often suffers from seamline distortion ghost effects, making it difficult to meet requirements organ-level detection. While raw images do not exhibit these issues, they pose challenges in accurately obtaining detected objects. detection was improved this study through fusion with using EasyIDP tool, thereby establishing a mapping relationship data. Small object conducted by Slicing-Aided Hyper Inference (SAHI) framework YOLOv10n on accelerate inferencing speed farmland. As result, comparing directly DOM, accelerated accuracy improved. proposed SAHI-YOLOv10n achieved mean average (mAP) scores 0.825 0.864, respectively. It also processing latency 1.84 milliseconds 640×640 resolution frames application. Subsequently, novel crop canopy dataset (CCOD-Dataset) created interactive annotation SAHI-YOLOv10n, featuring 3986 410,910 annotated boxes. method demonstrated feasibility detecting three in-field farmlands, potentially benefiting future wide-range applications.

Language: Английский

Citations

0