摘要
支持向量机应用于文本分类、手写数字识别、基因表达等许多领域,由于Harris角点检测算子对噪声点非常敏感,本文在文献[3]的基础上提出Harris算子和支持向量机相结合的方法来进行角点检测。首先利用Harris角点检测算法对两幅以上的无噪声图像提取角点,然后将提取的角点作为支持向量机的训练样本,构造支持向量机,最后利用训练好的支持向量机实现对未知角点的检测。这种方法对角点的检测较为准确,符合实时性的要求。
Harris corner detection is very sensitive to noise, this paper propose a corner detection algorithm based on support vector machine. Harris operator will be used in more than two images of a certain region for detecting the corner, detected corners as a support vector machine training samples, constructing the support vector machine. The support vector machine will be used to detect the non-noise images and the salt-and-pepper noise images. As a result, the support vector machine can be more accuratly detect the corner and non-corner. The results show that this algorithm is more accurate detection of the corner, in line with the requirements of real-time.
出处
《电子测试》
2011年第1期42-45,共4页
Electronic Test
关键词
支持向量机
特征点提取
核函数
统计学习理论
机器学习
support vector machine
feature extraction
kernel function
statistical learning theory
machine learning