鲁棒模型拟合中的进展与挑战.pdfVIP

鲁棒模型拟合中的进展与挑战 王菡子 厦门大学 模式分析与机器智能研究中心 Multi-structure fitting Contamination in data is unavoidable: Sensor noise; Faulty feature extraction; Segmentation errors; Multiple structures, etc. What do we need to estimate for multi-structure data:  The number of model instances  The scale of inlier noise of Figure 1. Multi-structure data each model instance.  The parameters of each model instance.