摘要
以"旅行家3号(Voyager-III)"移动机器人为研究对象,探讨全景视觉下目标足球的椭圆拟合与提取算法。通过对现有的随机Hough变换方法和非线性最小二乘方法的对比,并探讨两者在不含有噪声影响和有孤立点存在影响下的1/2和1/4椭圆拟合结果,对计算的复杂度和实时性性能作出评价。在非理想光照条件下,利用Markowitz投资组合模型进行足球目标的特征提取,可使目标与背景边缘的差异性得到增强,目标边缘清晰,为后续全景视觉下圆形畸变后的椭圆拟合与提取去除噪声。实验表明,该方法适用于足球机器人比赛应用实际,识别结果准确、可靠。
The ellipse fitting and extraction algorithm of a soccer based on panoramic image was studied by taking Voyager-III mobile robot as the study object.The available random Hough transformation（RHT） and non-linear least squares are compared,and their fitting results of 1/2 1/4 ellipses are researched when without noise effects or with isolated points,and their computation complexity and real-time performance were evaluated.The Markowitz portfolio model was proposed to realize the feature extraction of the target soccer under un-ideal illumination condition,which helps to enhance the differences between the boundaries of target and background,making the edge of the target more clear.These primary steps could help to remove the noises in the subsequent ellipse fitting and extraction for coping with the roundness distortion under the panoramic camera.The experiments demonstrate that,the fitting scheme is testified to be an accurate one with strong robustness,which adapts to the practical applications of the soccer robots’ matches.
出处
《中国惯性技术学报》
EI
CSCD
北大核心
2013年第2期214-220,共7页
Journal of Chinese Inertial Technology
基金
国家自然科学基金(61073041)
