户外增强现实的图像分割算法改进应用

Improvement and Application of Image Segmentation Algorithm for Outdoor Augmented Reality

传统图像识别过程多发生在环境较为稳定的室内场景中,而当前增强现实的发展趋向于对户外场景的应用,比如在旅游行业中对建筑、雕像等目标的识别。针对ARToolKit增强现实系统在户外自然场景下的图像分割过程因受到实时变化的光照、遮挡等因素的影响而出现错误分割导致增强现实效果难以实现的问题,提出一种基于直方图均衡化的改进GrabCut算法。采用直方图均衡化对图像非线性拉伸,提高图像对比度;对GrabCut算法的边界项函数进行均值化处理,减少边缘像素的错误分割,提高目标图像的边缘完整性,实现图像的准确分割。实验结果证明该方法在复杂现实场景中拥有较高的实时性和稳定性,改善了ARToolKit增强现实系统在户外场景中的处理性能。

The traditional image recognition process mostly occurs in indoor scenes with relatively stable environment,and the current development of augmented reality tends to apply to outdoor scenes,such as the identification of buildings,statues and other objects in the tourism industry.Aiming at the problem that the image segmentation process of ARToolKit augmented reality sys⁃tem in the outdoor natural scene is affected by the real-time changing illumination and occlusion,which causes the augmented reali⁃ty effect to be difficult to achieve,an improved GrabCut algorithm based on histogram equalization is proposed.Histogram equaliza⁃tion is used to nonlinearly stretch the image to improve image contrast.The boundary function of GrabCut algorithm is averaged to re⁃duce the error segmentation of edge pixels,improve the edge integrity of the target image and achieve accurate image segmentation.The experimental results show that the proposed method has high real-time performance and stability in complex real-world scenari⁃os,and improves the processing performance of ARToolKit augmented reality system in outdoor scenes.