Analiza metod YOLO w detekcji obiektów z obrazów rejestrowanych przez UAV oraz ich zastosowanie w pomiarze prędkości pieszych / Analysis of YOLO Methods in Detecting Objects from Images Recorded by UAVs and Their Application in Measuring Pedestrian Speed
Analiza metod YOLO w detekcji obiektów z obrazów rejestrowanych przez UAV oraz ich zastosowanie w pomiarze prędkości pieszych / Analysis of YOLO Methods in Detecting Objects from Images Recorded by UAVs and Their Application in Measuring Pedestrian Speed
Data
2025
Autorzy
Lichograj, Piotr
Chodyka, Marta
Szałkowski, Michał
Vladimir, Golovko
Tytuł czasopisma
ISSN czasopisma
Tytuł tomu
Wydawca
Akademia Bialska im. Jana Pawła II
Streszczenie
W ostatnich latach bezzałogowe statki powietrzne (UAV) stały się kluczowym narzędziem
pozyskiwania danych wizyjnych do analiz ruchu w środowisku miejskim. Perspektywa
lotnicza umożliwia szerokie zastosowania w monitoringu pieszych, jednak jednocześnie
generuje liczne wyzwania związane z niewielkim rozmiarem obiektów, zmienną wysokością lotu oraz niestabilnością detekcji. W rozdziale przedstawiono kompleksowy
przegląd architektur YOLO w kontekście ich wykorzystania w obrazach z UAV oraz omówiono
metody estymacji prędkości obiektów na materiale wideo. W części badawczej
zaprezentowano prototypowy system detekcji i estymacji prędkości pieszych z wykorzystaniem
modelu YOLOv8, wytrenowanego na specjalnie przygotowanym zbiorze danych
UAV. Oceniono skuteczność detekcji, stabilność śledzenia oraz wiarygodność pomiaru
prędkości. Wyniki potwierdzają, że nowsze generacje modeli YOLO, po odpowiednim
dostrojeniu do danych lotniczych, umożliwiają dokładną analizę ruchu pieszych, a temat
ten pozostaje obszarem o dużym potencjale badawczym.
In recent years, unmanned aerial vehicles (UAVs) have become essential tools for acquiring visual data to support motion analysis in urban environments. The aerial perspective enables effective pedestrian monitoring but poses significant challenges, including small object size, variable flight altitude, and detection instability. This chapter provides a comprehensive review of YOLO architectures and their applicability to UAV imagery, along with an overview of velocity estimation methods in video sequences. The experimental section introduces a prototype system for pedestrian detection and speed estimation using a YOLOv8 model trained on a custom UAV dataset. The system’s detection accuracy, tracking stability, and speed estimation reliability were evaluated. The findings demonstrate that modern YOLO models, when properly fine-tuned on UAVspecific data, enable accurate pedestrian motion analysis, while the topic itself remains an underexplored research area with substantial potential for future studies.
In recent years, unmanned aerial vehicles (UAVs) have become essential tools for acquiring visual data to support motion analysis in urban environments. The aerial perspective enables effective pedestrian monitoring but poses significant challenges, including small object size, variable flight altitude, and detection instability. This chapter provides a comprehensive review of YOLO architectures and their applicability to UAV imagery, along with an overview of velocity estimation methods in video sequences. The experimental section introduces a prototype system for pedestrian detection and speed estimation using a YOLOv8 model trained on a custom UAV dataset. The system’s detection accuracy, tracking stability, and speed estimation reliability were evaluated. The findings demonstrate that modern YOLO models, when properly fine-tuned on UAVspecific data, enable accurate pedestrian motion analysis, while the topic itself remains an underexplored research area with substantial potential for future studies.
In recent years, unmanned aerial vehicles (UAVs) have become essential tools for acquiring visual data to support motion analysis in urban environments. The aerial perspective enables effective pedestrian monitoring but poses significant challenges, including small object size, variable flight altitude, and detection instability. This chapter provides a comprehensive review of YOLO architectures and their applicability to UAV imagery, along with an overview of velocity estimation methods in video sequences. The experimental section introduces a prototype system for pedestrian detection and speed estimation using a YOLOv8 model trained on a custom UAV dataset. The system’s detection accuracy, tracking stability, and speed estimation reliability were evaluated. The findings demonstrate that modern YOLO models, when properly fine-tuned on UAVspecific data, enable accurate pedestrian motion analysis, while the topic itself remains an underexplored research area with substantial potential for future studies.
In recent years, unmanned aerial vehicles (UAVs) have become essential tools for acquiring visual data to support motion analysis in urban environments. The aerial perspective enables effective pedestrian monitoring but poses significant challenges, including small object size, variable flight altitude, and detection instability. This chapter provides a comprehensive review of YOLO architectures and their applicability to UAV imagery, along with an overview of velocity estimation methods in video sequences. The experimental section introduces a prototype system for pedestrian detection and speed estimation using a YOLOv8 model trained on a custom UAV dataset. The system’s detection accuracy, tracking stability, and speed estimation reliability were evaluated. The findings demonstrate that modern YOLO models, when properly fine-tuned on UAVspecific data, enable accurate pedestrian motion analysis, while the topic itself remains an underexplored research area with substantial potential for future studies.
Opis
Słowa kluczowe
UAV,
detekcja obiektów,
YOLOv8,
analiza ruchu pieszych,
estymacja prędkości,
pedestrian motion analysis,
speed estimation,
deep learning,
object tracking,
object detection