Abstract In three-dimensional visualization, sufficient memory and computing power can ensure real-time graphics rendering. However, due to equipment and algorithm performance limitations, it is difficult to graphically present big volume data efficiently and accurately, especially for high-dimensional and large volume geological data. In this paper we propose a real-time visualization method for logging data, which combines volume data compression and fast switching algorithm. First, we introduce an adaptive sampling method for large volume of data compression. Each block of the same size is sampled according to the dispersion and the sampling density grade, after which ray casting algorithm is used to render compressed volume data. Second, aiming at the graphic presentation delay caused by the exchange of large amounts of data in internal and external memory, a fast switching algorithm(FSA) based on subspace learning is presented. The attributes with strong correlation are put into the same group, from which feature subspace are learned and a mapping model between associated attributes is established according to base vector invariance. Once we need to switch from the currently displayed attribute to another for display, only a few coefficient values in the mapping model need to be changed, reducing the amount of data exchange. Our proposed method can greatly increase the compression ratio and reduce the computing time, ensuring real-time visualization for geological data.

中文翻译：

---

基于子空间学习的测井属性实时切换与可视化

摘要 在三维可视化中，充足的内存和计算能力可以保证图形的实时渲染。然而，由于设备和算法性能的限制，难以高效、准确地图形化呈现大体积数据，尤其是对于高维、大体积的地质数据。在本文中，我们提出了一种记录数据的实时可视化方法，它结合了体数据压缩和快速切换算法。首先，我们介绍了一种用于大量数据压缩的自适应采样方法。根据离散度和采样密度等级对每个相同大小的块进行采样，然后使用光线投射算法渲染压缩的体数据。第二，针对内存与外存大量数据交换导致的图形呈现延迟问题，提出了一种基于子空间学习的快速切换算法(FSA)。将相关性强的属性放入同一组，从中学习特征子空间，并根据基向量不变性建立关联属性之间的映射模型。一旦我们需要从当前显示的属性切换到另一个进行显示，只需要改变映射模型中的少数系数值，减少数据交换量。我们提出的方法可以大大提高压缩率并减少计算时间，确保地质数据的实时可视化。将相关性强的属性放入同一组，从中学习特征子空间，并根据基向量不变性建立关联属性之间的映射模型。一旦我们需要从当前显示的属性切换到另一个进行显示，只需要改变映射模型中的少数系数值，减少数据交换量。我们提出的方法可以大大提高压缩率并减少计算时间，确保地质数据的实时可视化。将相关性强的属性放入同一组，从中学习特征子空间，并根据基向量不变性建立关联属性之间的映射模型。一旦我们需要从当前显示的属性切换到另一个进行显示，只需要改变映射模型中的少数系数值，减少数据交换量。我们提出的方法可以大大提高压缩率并减少计算时间，确保地质数据的实时可视化。只需要改变映射模型中的少数系数值，减少了数据交换量。我们提出的方法可以大大提高压缩率并减少计算时间，确保地质数据的实时可视化。只需要改变映射模型中的少数系数值，减少了数据交换量。我们提出的方法可以大大提高压缩率并减少计算时间，确保地质数据的实时可视化。