Abstract: In the traditional method of marine radar echo simulation, large numbers of polygons are obtained using the intersection of scanning fans and Electronic Navigation Chart（ENC）. A method of marine radar echo simulation based on Digital Elevation Model（DEM） image enhancement is proposed in this paper, so as to improve the poor real-time performance and lowly realistic echo image of traditional ones. The method establishes a heterogeneous system of Central Processing Unit（CPU） combined with Graphics Processing Unit（GPU）. The real-time echo is produced via cluster and connecting steps for reducing the number of polygons, after which the echo image is enhanced via the structure of GPU. In order to eliminate the impact of meridianal parts, DEM of the same harbor is used to modify the echo image by splitting harbor map into multiple segments. The process of simulation is according to principle of the producing of radar echo, and the sector-banded texture blending algorithm is used to replace the traditional ones based on the pixel-scan model. The simulation results show that the efficiency of image enhancement using GPU increases in large scenes, thus the overall speedup achieves 240 times than the traditional ones using CPU.

Key words: image enhancement, meridianal parts, radar echo, Digital Elevation Model（DEM）, Graphics Processing Unit（GPU）

摘要： 针对传统的航海雷达回波模拟利用扫描扇形与电子海图求交，得到的回波多边形数量巨大，大场景的回波模拟无法实时处理、真实感差等问题，提出了利用DEM图像增强的回波模拟方法。该算法建立了一种CPU结合GPU的异构系统，首先在CPU中通过聚合、连通步骤减少多边形数量实时生成回波，然后在GPU中叠加港口DEM高程信息对回波实现图像增强，纬度上分段处理消除纬度渐长率的影响。该算法实现了回波图像从扫描像素点到扇形带状回波的转化，仿真过程符合真实的雷达回波形成原理。在较大仿真场景下，回波图像增强经过GPU处理后并行效率提高，相对CPU的加速比可至240倍。

关键词: 图像增强, 纬度渐长率, 雷达回波, 数字高程模型, 图像处理单元