Skip to main content

Table 1 Comparison of visual SLAM algorithms combined with deep learning to remove dynamic targets

From: Dynamic visual SLAM and MEC technologies for B5G: a comprehensive review

Algorithm

Camera type

Method

Application scenarios

DS-SLAM

D

SegNet +Mobile consistency testing

I

[35]

D

PSP Net+Mobile consistency testing

I

DynaSLAM

S,M,D

Mask R-CNN+Multi-view geometry

I/O

PSP Net-SLAM

D

PSP Net+Optical flow method

I

[46]

M

YOLO v3

I

Dynamic-SLAM

M

SSD+Selective tracking algorithm

I/O

Detect-SLAM

D

SSD+Movement probability propagation

I

SOF-SLAM

D

SegNet+For polar geometric constraints

I/O

DSO-Dynamic

M

Mask R-CNN+DSO

O

  1. M represents a monocular camera, S represents a binocular camera, D represents an RGB-D camera, I represents an indoor environment, and O represents an outdoor environment