Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 351 - 368
Published: Jan. 1, 2024
Language: Английский
Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 351 - 368
Published: Jan. 1, 2024
Language: Английский
Artificial Intelligence in Agriculture, Journal Year: 2025, Volume and Issue: unknown
Published: Jan. 1, 2025
Language: Английский
Citations
0WSEAS TRANSACTIONS ON SIGNAL PROCESSING, Journal Year: 2025, Volume and Issue: 21, P. 31 - 40
Published: April 7, 2025
One of the most relevant, but at same time time-consuming and costly, aspects infrastructure system is monitoring road infrastructures, often subject to deterioration that compromises their use. Current systems consist individual reports or use human resources that, through equipped vehicles, have purpose carrying out a reconnaissance process, which characterized by errors uncertainties. In this context, aim work was experiment implement an experimental innovative Automated Integrated Sensing System (AISS) for infrastructures. This system, starting from Remote images Unmanned Aerial Vehicles (UAVs), uses Mask R-CNN neural network identify cracks. information, together with other included in database, then used Geographical Information (GIS) relative visualization. therefore proposes methodology implementation helps policy makers determining urgent interventions. fact, categorization severity degradation user-friendly visualization, allow us make decisions based on data.
Language: Английский
Citations
0Frontiers in Plant Science, Journal Year: 2024, Volume and Issue: 15
Published: June 18, 2024
The nonuniform distribution of fruit tree canopies in space poses a challenge for precision management. In recent years, with the development Structure from Motion (SFM) technology, unmanned aerial vehicle (UAV) remote sensing has been widely used to measure canopy features orchards balance efficiency and accuracy. A pipeline volume measurement based on UAV was developed, which RGB digital surface model (DSM) orthophotos were constructed captured images, then segmented using U-Net, OTSU, RANSAC methods, calculated. accuracy segmentation compared. results show that U-Net trained DSM achieves best task, mean intersection concatenation (MIoU) 84.75% pixel (MPA) 92.58%. However, estimation only achieved Root square error (RMSE) 0.410 m 3 , relative root (rRMSE) 6.40%, absolute percentage (MAPE) 4.74%. deep learning-based method higher both task task. For volumes up 7.50 OTSU achieve an RMSE 0.521 0.580 respectively. Therefore, case manually labeled datasets, use segment region can measurement. If it is difficult cover cost data labeling, ground partitioned yield more accurate than RANSAC.
Language: Английский
Citations
2Lecture notes in computer science, Journal Year: 2024, Volume and Issue: unknown, P. 351 - 368
Published: Jan. 1, 2024
Language: Английский
Citations
0