【作者】 徐秀英； 刘卓； 王博亮；

【Author】 Xu Xiu-ying, Liu Zhuo, Wang Bo-liang(Department of Computer Science, Xiamen University, Xiamen, 361005;College of Electricity Science and Engineering, National University ofDefense Technology,Changsha 410073;

【机构】 厦门大学计算机科学系； 国防科技大学电子科学与工程学院；

【摘要】 本文提出了一种基于三角函数的深度图像自适应局部表面拟合方法,设计并实现了迭代寻优算法获得最佳拟合系数和拟合函数表达式,并将它应用于深度图像的特征提取,即根据拟合结果计算深度图像的高斯曲率和平均曲率。理论分析和实验结果表明,与基于多项式的曲面拟合方法比较,本文提出的基于三角函数的曲面拟合方法效果更好,尤其是在边缘点处,使用较少的基函数便可以获得更加准确的拟合结果。迭代寻优算法稳健可靠。且在同等精度下,计算复杂度更小。利用本文的表面拟合算法提取深度图像的局部几何特征,可以进行边缘检测、图像分割等处理。更多还原

【Abstract】 The surface fining is one key step of range image process. In this paper, a new method based on trigonometric function was proposed to fit the surface of range images. Then an iterative optimism algorithm was designed to calculate the optimized fitting coefficients and fitting function expressions. The result of surface fitting was used to extract features from range images, such as calculating Gauss curvature and average curvature. Theoretical analysis and experimental results prove that, comparing with polynomial-based fitting method, the presented algorithm is robust and the current algorithm fits much better with few functions on the edge points than polynomial-based fitting method. Furthermore, to achieve same fitting accuracy it spent much less time than the polynomial fitting method. The presented method can be used in the edge detection, image segment and object recognition of the range image.更多还原