【摘要】 图像滤波和边缘检测是视觉导航系统中道路检测和障碍物检测等复杂视觉处理的关键步骤,其性能和处理时间直接影响了后续图像处理的性能及视觉系统的整体响应时间.为此,提出了一种基于FPGA的实时图像模板滤波及边缘检测方法,将LoG模板分解为两个一维模板的卷积和,从而降低了算法运算量.并充分利用FPGA的并行机制及片内丰富的RAM资源,采用分布式算法,用查找表代替乘法器进行乘法运算.实验表明该方法满足视觉导航系统中实时性需求,且又适用于其他基于模板运算的图像处理算法.更多还原

【Abstract】 Image filtering and edge detection are the key procedures in complex vision processing,such as road edge and obstacle detection in vision-based navigation systems. Its performance and processing time directly affect the performance of the subsequent image processing and the total response time for vision-based navigation systems. An implementation of real-time image filtering and edge detection based on FPGA is proposed, which decomposes the LoG template into two one-dimension templates to decrease the calculating amount. Taking full advantage of the parallel operating character and the abundance on-chip RAM resource of FPGA, the distributed arithmetic is introduced, which establishes a look-up-table to take the place of the multiplication. Experiment results show this method satisfies the real-time requirements, and can be applied to other template convolution based image processing algorithms.更多还原