In recent years, the development of UAV technology has provided new perspectives for solving some problems of urban traffic planning. The traffic data collected by the UAV platform are used to identify and analyze vehicles and pedestrians on the ground. Due to the large number of small targets in the UAV, the traditional multi-target detection algorithms have low performance and are prone to leakage and misdetection. In this paper, we propose the ALF-YOLOv8s algorithm to solve these problems. The ALF-YOLOv8s algorithm uses Alterable Kernel Convolution (AKConv) and Efficient Multi-scale Attention (EMA) to improve the feature extraction capability of the algorithm and Large Separable Kernel Attention (LSKA) into SPPF to enhance the semantic fusion between different feature layers The experimental results show that the ALF-YOLOv8s algorithm can detect targets more accurately and comprehensively than the traditional YOLOv8s algorithm.
Vehicle and Pedestrian Detection for UAV Platform Based on Deep Learning
Lect. Notes Electrical Eng.
International Conference on Computer Science and its Applications ; 2024 ; Pattaya, Thailand December 18, 2024 - December 20, 2024
2025-05-15
7 pages
Article/Chapter (Book)
Electronic Resource
English
Research on Deep Learning-Based Vehicle and Pedestrian Object Detection Algorithms
DOAJ | 2024
|