ENHANCED AERIAL OBJECT DETECTION USING SUPERPIXEL-GUIDED MULTISCALE CNNS FROM UAV IMAGERY

Abstract

Enhanced aerial object detection is crucial for various applications, but existing methods face challenges in efficiency and accuracy. This paper presents an innovative approach leveraging superpixel-guided multiscale convolutional neural networks (CNNs) applied to UAV-captured imagery. The proposed method enhances feature extraction efficiency and computational performance through superpixel segmentation. Multiscale CNNs extract features at various levels of detail, enabling accurate detection of objects such as buildings, roads, vehicles, and vegetation. The system showcases resilience to lighting, terrain, and object scale variations, promising superior segmentation accuracy. Integration of superpixel guidance enhances adaptability and efficiency, highlighting its potential for urban planning, environmental monitoring, and infrastructure management.