This study tackles the intricate challenge of pedestrian detection in low-resolution aerial videos, where conventional methods relying on facial features struggle due to image blurriness and human variability. Our innovative approach shifts the focus from appearance to movement patterns, analyzing short video sequences (approximately half a second). We address two primary tasks: video stabilization and identification of moving objects. By utilizing specialized features capturing both spatial and temporal information, we effectively classify these moving objects as pedestrians. Across diverse and challenging datasets, our method demonstrates superiority over traditional approaches, particularly in low-resolution scenarios. It excels by achieving higher detection accuracy and minimizing false positives. The proposed approach is an ideal solution for real-world applications, especially in situations where video quality is compromised, such as surveillance footage and drone video processing. The significance of our method lies in its ability to adapt to challenging environments, providing robust pedestrian detection in low-resolution aerial footage. With enhanced accuracy and reduced false positives, our approach offers practical advantages for applications requiring reliable performance in compromised video quality scenarios, including surveillance and drone video processing.