ELM在图像边缘抽取抗噪阈值优选中的应用研究

Application Research on ELM in Optimization of Anti-noise Threshold in Image Edge Extraction

为解决噪声敏感性导致的图像边缘不连续、抽取速度慢和识别效果差等难题,提出了一种基于ELM的图像几何特征识别算法。解剖了图像几何特征识别存在的问题,讨论了最优阈值可最大限度地区分图像的前景和背景颜色,基于ELM模型研究了计算最优阈值的算法,借助噪声阈值限制路径代价函数以缩小搜索范围,从而提高算法执行速度。仿真实验验证了所提出算法在图像处理中可取得了良好的图像边缘识别效果。研究结果表明,基于ELM的图像几何特征识别算法是合理、可行和有效的。

In order to solve the puzzles such as discontinuous edge,slow extraction speed and poor recognition effect in image recognition caused by noise sensitivity,the paper proposed an algorithm of geometric feature recognition based on ELM.In this paper,it dissected the puzzles of image geometric feature recognition,and discussed the optimal threshold to distinguish foreground and background color to the maximum extent.Based on the ELM model,studied the algorithm of calculating the optimal threshold,and in which,the path cost function is limited by the noise threshold to narrow the search range,so as to improve the execution speed of the algorithm.Simulation results show that the proposed algorithm can achieve good image recognition performance in image processing.The results show that the geometric feature recognition algorithm based on ELM is reasonable,feasible and effective.