3D Semantic VSLAM of Indoor Environment Based on Mask Scoring RCNN

Algorithm 1

3D mapping method for inverse depth filtering.

(1)	Input: map point data x
(2)	Output: point cloud map y
(3)	The search range of scene depth measurement is defined as
(4)	The scene depth value limits the number of layers of matching cost in stereo matching
(5)	Isolated outlier culling while then The inverse depth distribution after transformation is
(6)	Inverse deep fusion
(7)	Isolated outlier secondary culling
(8)	Get point cloud map