WCD-YOLOv11: A lightweight YOLOv11 model for the real-time image processing in UAV

Liangshuai Liu

doi:10.1016/j.aej.2025.10.045

WCD-YOLOv11: A lightweight YOLOv11 model for the real-time image processing in UAV

2025-11-14

0

OA

AI

DOI：10.1016/j.aej.2025.10.045

Original

Save

View PDF

Abstract

En 中文

With the rapid development of drone technology, drone imagery has become increasingly popular in real-time target detection, intelligent monitoring, and other applications. However, existing detection models still face challenges in accuracy and efficiency, particularly with small objects, complex backgrounds, and resource constraints. To address these issues, we propose a lightweight YOLOv11 improvement model, WCD-YOLOv11. In this model, we integrate several modules, including Wavelet Transform Depthwise Convolution (WTDWC), Cross-Stage Partial Network (CSPNet), and Partial Convolution (P-Conv). WTDWC optimizes downsampling, reducing computational load and improving small object detection; CSPNet enhances multi-scale object detection, improving stability; and P-Conv improves performance by focusing on valid pixels in occluded and incomplete objects. These modules address challenges in small object detection, misdetection in complex backgrounds, and real-time processing in limited environments. Experiments on VisDrone and UAVDT show WCD-YOLOv11 outperforms peers—achieving 0.875 mAP@0.5, 0.505 AP small, 111.8 FPS on VisDrone, and 0.86 mAP@0.5, 0.42 AP small on UAVDT. The results demonstrate that WCD-YOLOv11 improves detection accuracy, inference speed, and computational efficiency, highlighting its potential for real-time drone image processing.

Keywords:

Lightweight YOLO model

Real-time target detection

Drone imagery

Wavelet Transform Depthwise Convolution

Cross-Stage Partial Network (CSPNet)

Partial Convolution (P-Conv)

WCD-YOLOv11: A lightweight YOLOv11 model for the real-time image processing in UAV

Abstract

AI Summary

Journal

Organization