一种基于特征点提取与随机树的增强现实系统

AN AUGMENTED REALITY SYSTEM BASED ON FEATURE-POINT EXTRACTION AND RANDOM TREE

在增强现实领域,实时性是一个重要的问题。利用传统的方法训练与识别往往需要大量的时间。提出并实现了一种基于图像特征点快速提取与随机树分类的增强现实系统。在对一幅或很少的几幅包含标志物的原始图像进行训练之后,该系统能快速高效地识别摄像头新采集到的图像中的标志物,并计算标志物的空间位置坐标,有效地将真实场景与虚拟物体进行合成。该系统大大减少了识别与合成所花的时间,并在较大的视角和尺度变化下仍体现出良好的效果。

Real-time performance is an important issue in augmented reality field.To train and recognise using traditional means usually need much time.We proposed and implemented an augmented reality system which is based on fast extraction of image feature point and random tree classification.The system is able to quickly and efficiently recognise markers in the image newly captured by video camera after to be trained by one or several primitive images with the markers,and can calculate the coordinates of marker＇ s spatial position,effectively integrate the real scenes with virtual objects.The system drastically diminishes the time cost in recognition and integration,and performs fairly good effect in a greater visual angle and scale variation.