| Popis: |
This paper presents a method to improve the robustness of SLAM system and to make full use of scene information for higher accuracy. Due to the strong illumination robustness of Hue in HSV space, we use Gaussian mixture model to improve matching performance during front matching. And we found that the BOW(Bag of Words) similarity distribution is gaussian, so that we divide frames into scenes and build up the scene distribution model. Based on this, the accuracy of SLAM is higher and lose little real-time performance. |
