关键词: 体特征表达, 高度场八叉树, 投影变换, 光线投射

Abstract: Volumetric features representation plays a vital role for user understanding and recognizing the virtual environment. The current algorithm is inefficient due to its large storage and inconvenient acceleration in GPU, and it is difficult to satisfy the real-time requirements of visualization. Aiming at this problem, an efficient volumetric features representation algorithm using heightfields-octree is proposed. The algorithm can not only solve the problem that the heightfields can only represent 2.5 dimensional scene, and cannot express the true 3 dimensional scene, but also provide a new feasible way for volumetric features representation. The heightfields representation of 3D scene is generated by octree structure, which extends the traditional heightfields of z to x, y and z three directions. Firstly, a preprocessing method of triangular is put forward, ensuring model accuracy and data integrity. Secondly, an algorithm of heightfields judgment and rasterization on projection transformation is proposed, converting geometric primitives into heightfields of two-dimensional space. Finally, the ray casting algorithm based on heightfields octree is realized. The experimental results show that the algorithm can dramatically reduce data storage capacity, and higher ray casting efficiency, and better expression of 3 dimensional scene.

Key words: volumetric features representation, heightfields-octree, projection transformation, ray casting