基于改进随机蕨算法的GIS增强现实应用研究

摘要

随机蕨算法由于其简单高效，实时性好的特点，在计算机视觉领域得到了广泛的应用。但是其在构造二值特征过程中采用的基于像素点灰度值比较的方法对图像变形的适应性较差。本文基于方向梯度直方图对光照变化，图像变形鲁棒性较好的特点，利用统计方向梯度直方图取代像素灰度值进行比较，以此构造特征点二值特征。基于改进的随机蕨算法，本文设计了一种将增强现实技术（ AR）应用于GIS的框架模型。实验结果表明，经过改进的随机蕨算法对目标识别的正确率得到了有效提升，且算法的实时性较好，适用于GIS的增强现实应用。%Random ferns algorithm has been widely used in the computer vision field, because of its simplicity and rapid computation. However, the method of constructing local binary features by comparing the intensity domain isn’ t stable for image changes.In view of the robustness the oriented gradient histogram ( HOG) offer to illumination and image changes, this article improved the random ferns algorithm by comparing the local oriented gradient histogram instead of the intensity domain.And basing on the improved random ferns algorithm, this paper designed a scheme applying augmented reality technique on GIS.The experimental results showed that improved random ferns algorithm has obviously higher precision for object recognition as well as rapid computation, so it’s suitable for the appli-cation of augmented reality on GIS.