3 Three-Dimensional Restoration Cloud Point Creation Method Using GPU Accelerated Computing

摘要

The present invention relates to a three-dimensional reconstructed cloud point generation method using a paper-based accelerated computing, and more particularly, when performing an accelerated computing task using a GPU core, 3D reconstructed cloud point generation according to a conventional CPU single processing method The present invention relates to a three-dimensional reconstructed cloud point generation method using GUI-accelerated computing that can maintain reliability by improving processing speed. The method of generating a three-dimensional reconstructed cloud point using the present inventors accelerated computing, An indoor space image acquisition step (S100) for acquiring an indoor space image by the image input unit 110; The image color data extracting unit 120 obtains the indoor spatial image obtained by the image input unit and extracts the image color data using the multi-core, which is a GUP acceleration function for the color information of the obtained indoor spatial image. Data extraction step (S200), A color consistency analysis step (S300) for the color consistency analysis unit 130 to analyze the extracted image color data according to a consistent color definition defined by a GPU core; The pixel node generation unit 140 generates a color node sequential node generated by the GPU core and connects the associations to each other (S400); A rare depth map generation step (S500) for generating a rare depth map by the rare depth map generator 150 using the generated GPU multicore node; A node uniform division step S600 for uniformly dividing the pixel nodes generated by the GPU multicore node (S600), The dense depth map generation unit 170 may include a dense depth map generation step S700 for generating a dense depth map using the GPU multicore.