一种新的基于线性EIV模型的鲁棒估计算法

提出了一种新的基于线性EIV模型的鲁棒估计算法——鲁棒扩充算法．该算法从结构化数据区域出发，逐渐扩充模型数据集，并不断更新模型参数的估计，直至找到所有模型数据．在每次迭代中，使用C-Step方法对集合进行调整，从而保证了算法的鲁棒性．同时，提出了关于粗差数据和结构化数据分布的结构化密度假设，结合Mean Shift算法，完成对算法的初始位置选取．仿真结果表明，该算法可以有效地处理含有多个结构和大量离群样本的混杂数据，与现有算法相比，具有更强的鲁棒性和更高的精度．

A Novel Robust Estimation Algorithm Based on Linear EIV Model

Robust estimation of multiple-structured data is a fundamental problem in computer vision. Based on the linear EIV model, a novel robust growing algorithm is proposed to estimate the inherent model parameters from the contaminated data. The algorithm starts from an initial subset in the area of structured data. Then it adds the model data points to the subset and updates the parameter estimate iteratively. At each iteration, the C-Step method originated from the MCD estimator is adopted to adjust the subset and to ensure the robustness of the algorithm by ejecting outliers. Based on the structured density assumption that the gross errors should be no denser than the structured data, otherwise the structured data would be indistinguishable from the gross errors; the mean shift algorithm is adopted to ensure a good initialization for the robust growing algorithm. Experiments show that the proposed algorithm can deal with contaminated data, which contain multiple structures and high percentage of gross errors efficiently, and has higher robustness and accuracy than the existing robust estimation algorithms.