Novel Line Representation with Consistent Second Order Statistics for SLAM Applications

抄録

Line is the most important feature in the feature based mapping. Although it has been widely used for, but we insist that there is no proper line model that is good for SLAM applications. To validate our claim, we firstly derive five necessary conditions that a good model should satisfy. Then we evaluate the previous line models and show that none of them satisfy all the necessary conditions. Finally, we propose a new line representation that fulfills the fore-mentioned conditions.