Improving terrain tile data selection efficiency, real-time loading of visible tile data and building GPU-based continuous Level of Details (LOD) are the key technologies for large scale terrain rendering on GPU. In this article, in order to reduce terrain tile data selection time, we build double layers tile quad tree for massive terrain data and organize tile data by designing Z-order space filling curve. According to the visible region coordinates obtained by GPU offscreen render to texture, we realize real-time loading of visible tile data from CPU to GPU. Map visible tile quad tree into twodimensional texture on GPU making full use of the characteristics of GPU multi-channel parallel processing. In order to execute tile error metric computation and LOD selection on GPU, we design GPU-based cuboid bounding unsaturated error metric, reduce CPU computational burden and enhance the terrain rendering performance. Experiments show that our algorithm can improve the utilization rate of GPU in the terrain rendering and achieve a good visual effect and a high frame rate.

This paper highlighted the technology of gray transformation to target the image captured using the camera, which converted a colored image into grayscale, filtered out the noise using the median, and extracted the image edge by using Canny operator. Because the extracting edges were not clear, it could not obtain the target pixel point accurately. For improving the accuracy and real-time feature extraction, corner feature points detection algorithm via randomized Hough transform based on spatial moment was proposed. To obtain accurate coordinates of the corners, the image was processed using matlab2011 software in the experiment. In order to achieve camera calibration rapidly and effectively, the linear calibration algorithm was improved. Using the least square method, the calculation process was simplified, and the calibration error was reduced. The experimental results show that the proposed algorithm was simple and effective, which improved the calibration accuracy, and verified its validity and feasibility.