角点检测理论及算法研究-计算数学专业论文.docxVIP

华 华 中 科 技 大 学 硕 士 学 位 论 文 摘 要 提取图像的特征信息，是图像处理工作中的基础工作之一。在图像的各种特征 中，角点包含的信息数据量较小，即非常小的数据信息保存了图像的灰度变化的特 征信息，同时外界影响因素对角点特征提取算法的影响也很小。因此角点检测算法 是一项非常有意义并且有广阔应用前景的研究工作。 本文的主要工作之一是对目前使用较为广泛的Harr i s角点检测进行改进。提出 了基于高斯尺度和梯度方向的算法，以改善Harr i s角点检测算法尺度单一并且鲁棒 性不强的现状。算法设计的目的为在对算法的速度影响较小的前提下，有效的降低 角点的错误率，提高算法的鲁棒性。仿真实验表明，在噪声影响比较大的图像中， 该算法具有很好的抗噪性，并且检测出的角点更加稳定，定位更加准确。本文的另 一个重点工作是针对Tr aj kovi c角点检测算法出现错误角点率较高的情况进行改进, 算法通过对初选点局部区域灰度的变化分情况讨论，对初选点是否为角点进行二次 判断，降低了角点错误率，同时消除了Tr aj kovi c角点检测算法的异常斜边反应， 并通过仿真实验对改进的算法进行验证。 关键词：特征提取，Harris 角点检测，尺度空间，USAN 区域，Trajkovic 角点检测 I Abstract Extrac the feature information of image is one of the basis works in the image processing. Comapre with other image features which preserved the characteristics of the image,the corner point contain the smaller information data,less influence on the external factors, kept the feature information of the image with the least data. The study of the corner detection algorithm is therefore of great significance and have broad application prospects.One of the main work of this paper is the improvement of the Harris corner detection algorithm which is widely used at the present time. To solve the disadvantage of single scale and improve robust of the algorithm. The improvement approach of the Harris algorithm is based on Gaussian scale and gradient direction.the aim the improvement is to reducing the corner of the error rate effectively with less impact on the speed of the algorithm. Simulation results show that the improvement algorithm has good noise immunity in the noise affecting image,besides the corner detection is more stable and accurate positioning. Another focus of this research is Trajkovic corner detection,the result of this algorithm has a higher error proportion and to improvement the algorithm,this paper take the idea of local area to eliminate errors corners.The simulation results verified that the improvement of this paper have removed the situation of the hypotenuse abnormal response. Keywords: Extrac fe