中文摘要：

提出了一种改进的峰值检测算法,将累加器空间视为一幅二维图像,利用局部算子区增强期望的直线和噪声对应的峰值之间的差别;利用最小二乘法拟合增强后的累加器空间对应的分布直方图,得到具有更强鲁棒性的阈值来确定期望峰值所在的位置。实验结果表明,该算法同时具有较高的精度和较强的鲁棒性,对先验知识的需求很低,具有较好的通用性,为自主应用方式奠定了基础。

英文摘要：

This paper proposed an improved Hough transform, in which the accumulator space was treated as another 2-D image, the difference between peaks caused by expected curves and noises in magnified by a local operator. Then histogram was fitted by non-linear least square method, finally the threshold was got to determine the location of expected peaks with higher robustness. The experimental results show that higher accuracy and robustness can be achieved with low requirements about preliminary knowledge. The proposed method can be used in different application, especially for potential real autonomous application.