多核的鲁棒LS-SVM在图像边缘检测中的应用研究
Edge Detection Based on Robust Least Squares Support Vector Machines with Multiple Kernel
-
摘要: 针对传统最小二乘支持向量机易产生过拟合,在曲面拟合边缘检测中推广性能差的问题,提出了一种改进的多核的鲁棒最小二乘支持向量机图像边缘检测技术.并且利用粒子群算法对实验中的参数进行寻优,得到最优参数.通过与已有的Canny算法、BP神经网络算法以及使用单一核函数的标准LS-SVM相比较,验证了多核的鲁棒LS-SVM算法的有效性.实验结果表明:该算法提取的边缘比较精细、伪边缘较少,是一种有效的图像分析与处理的方法.Abstract: Based on the powerful nonlinear mapping ability of kernel learning,a novel method for edge extraction was proposed to overcome the over-fitting of original LS-SVM and improve the robustness of original LS-SVM.Particle Swarm Optimization(PSO) algorithm was used to the optimize the parameters.Compared with Canny method,BP neural network and standard LS-SVM,it is shown that the proposed method is effective and has better performance than other algorithms under the same condition,it is a very practical image processing algorithm.