Target detection by lidar is challenging due to the difficulty in obtaining the complex attitude of targets and capturing the real coincidence between target and facula. To address this problem, in this paper, a real-time mapping method of laser beam to complex targets based on GPU programming is proposed. By taking advantages of modern graphics hardware with respect to GPU programming technology and frame buffer object merit, the proposed approach takes each surface light source matrix as the observer, renders the current scene in the light source spatial coordinate system, and records the rendering results into the memory texture. To realize real-time mapping and rendering, the results observed by the light source in the world coordinates are restored and mapped to the model. Based on deep cache principle of Zbuffer and texture mapping principle, the model information (e.g., light source irradiance, vertex position and patch normal on the vertex of each triangular patch) can be correctly obtained with virtue OSG file reading-writing plug-in. Extensive experiments demonstrate the strong universality of the proposed algorithm. It is powerful in reading three-dimension files of various formats and is suitable for uniform or non-uniform surface light sources. It meets the quasi real-time computational requirements of two surface light sources with low requirements on system graphics hardware. Various model information could be acquired in quasi real-time, e.g., the components of the illuminated surface piece, the vertices of the illuminated triangular surface, the normal information and the irradiation intensity received by the vertex of the triangular patch. The algorithm is novel in providing reference and basis for laser illumination, recognition and detection.