一种基于图像分割技术和深度学习的高融合面部迷彩设计方法

A high fusion facial camouflage design based on image segmentation technique and deep learning

摘要:

针对传统面部迷彩设计主观性强、背景相似度差等问题，提出一种基于图像分割技术和深度学习的高融合面部迷彩设计方法。利用OpenCV视觉算法库和基于MediaPipe机器学习框架的FaceMesh深度学习模型，构建高融合面部迷彩计算机辅助设计系统。该系统能够准确检测识别面部轮廓和特征点，提取背景特征并构建符合背景特性的迷彩斑块库，实现迷彩斑块在面部自动调用填充，自动生成与背景高度融合且符合面部特性的面部迷彩设计方案。通过建立相似度指标评价体系，对采用上述方法设计的面部迷彩的伪装效果进行了实验验证。结果表明，该方法能够有效提高面部迷彩设计的科学性及伪装效果，为单兵在战场上快速精准地实施面部伪装提供了可靠且有效的方案。

Abstract:

Aiming at the problems of strong subjectivity and poor background similarity in traditional facial camouflage design, a high fusion facial camouflage design method based on image segmentation technology and deep learning was proposed. Using the OpenCV visual algorithm library andFaceMesh deep learning model based on MediaPipe machine learning framework, a high fusion facial camouflage computer-aided design system was constructed. The system can accurately detect and identify facial contours and feature points, extract background features and build a camouflage patch library that conforms to the background characteristics, so as to realize the camouflage patches in the face to be automatically called and filled, and to automatically generate facial camouflage design solutions that are highly fused with the background and conform to the facial characteristics. The camouflage effect of the facial camouflage designed by the above method was experimentally verified by establishing a similarity index evaluation system. The results show that this method can effectively improve the scientific nature of facial camouflage design and camouflage effect, and provide a reliable and effective solution for individual soldier to implement facial camouflage quickly and accurately on the battlefield.