forked from raulmur/ORB_SLAM
-
Notifications
You must be signed in to change notification settings - Fork 0
/
README.txt
176 lines (104 loc) · 7.01 KB
/
README.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
ORB-SLAM is a versatile and accurate Monocular SLAM solution able to compute in real-time the camera trajectory and a sparse 3D reconstruction of the scene in a wide variety of environments, ranging from small hand-held sequences to a car driven around several city blocks. It is able to close large loops and perform global relocalisation in real-time and from wide baselines.
Related Publications:
[1] Raúl Mur-Artal, J. M. M. Montiel and Juan D. Tardós. ORB-SLAM: A Versatile and Accurate Monocular SLAM System. IEEE Transactions on Robotics (accepted), 2015.
Link to pdf: http://webdiis.unizar.es/~raulmur/MurMontielTardosTRO15.pdf
1) License
ORB-SLAM is released under a GPLv3 license. Please note that we provide along ORB-SLAM a modified version of g2o and DBoW2 which are both BSD.
For a closed-source version of ORB-SLAM for commercial purposes, please contact the authors.
If you use ORB-SLAM in an academic work, please cite:
@article{murSubTro2015,
title={{ORB-SLAM}: a Versatile and Accurate Monocular {SLAM} System},
author={Mur-Artal, Ra\'ul, Montiel, J. M. M. and Tard\'os, Juan D.},
journal={IEEE Transaction on Robotics (to appear). arXiv preprint arXiv:1502.00956},
year={2015}
}
2) Prerequisites (dependencies)
Boost
We use the Boost library to launch the different threads of our SLAM system.
sudo apt-get install libboost-all-dev
ROS
We use ROS to receive images from the camera or from a recorded sequence (rosbag), and for visualization (rviz, image_view).
We have tested ORB-SLAM in Ubuntu 12.04 with ROS Fuerte, Groovy and Hydro.
If you do not have already installed ROS in your computer, we recommend you to install the Full-Desktop version of ROS Fuerte.
<http://wiki.ros.org/fuerte/Installation/Ubuntu>.
g2o dependencies (Eigen, BLAS, LAPACK, CHOLMOD)
We use g2o to perform several optimizations. We include a modified copy of the library including only the components we need
and also some changes that are listed in Thirdparty/g2o/Changes.txt.
In order to compile g2o you will need to have installed CHOLMOD, BLAS, LAPACK and Eigen3.
sudo apt-get install libsuitesparse-dev
sudo apt-get install libblas-dev
sudo apt-get install liblapack-dev
sudo apt-get install libeigen3-dev
DBoW2
We make use of some components of the DBoW2 library for place recognition and feature matching. We include a modified copy of the library
including only the components we need and also some modifications that are listed in Thirdparty/DBoW2/LICENSE.txt.
It only depends on OpenCV, but it should be included in the ROS distribution.
3) Installation
1. Make sure you have installed ROS and all library dependencies (boost, eigen3, cholmod, blas, lapack).
2. In your ROS package path (check your environment variable ROS_PACKAGE_PATH) clone this repository:
git clone https://github.com/raulmur/ORB_SLAM.git ORB_SLAM
3. Build g2o. Go into Thirdparty/g2o/ and execute:
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make
Tip: To achieve the best performance in your computer, set your favorite compilation flags in line 97 and 98 of Thirdparty/g2o/CMakeLists.txt
(by default -03 -march=native)
4. Build DBoW2. Go into Thirdparty/DBoW2/ and execute:
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make
Tip: Set your favorite compilation flags in line 4 and 5 of Thirdparty/DBoW2/CMakeLists.txt
(by default -03 -march=native)
5. Build ORB_SLAM. In the ORB-SLAM root execute:
mkdir build
cd build
cmake .. -DROS_BUILD_TYPE=Release
make
Tip: Set your favorite compilation flags in line 12 and 13 of Thirdparty/DBoW2/CMakeLists.txt
(by default -03 -march=native)
4) Usage
See section 5 to run the Example Sequence.
1. Launch ORB-SLAM from the terminal (roscore should have been already executed):
rosrun ORB_SLAM ORB_SLAM PATH_TO_VOCABULARY PATH_TO_SETTINGS_FILE
You have to provide the path to the ORB vocabulary and to the settings file. The paths must be absolute or relative to the ORB_SLAM directory.
We already provide the vocabulary file we use in ORB_SLAM/Data/ORBvoc.txt.tar.gz. Uncompress the file, as it will be loaded much faster.
2. The last processed frame is published to the topic /ORB_SLAM/Frame. You can visualize it using image_view:
rosrun image_view image_view image:=/ORB_SLAM/Frame _autosize:=true
3. The map is published to the topic /ORB_SLAM/Map, the current camera pose and global world coordinate origin are sent through /tf in frames /ORB_SLAM/Camera and /ORB_SLAM/World respectively. Run rviz to visualize the map:
- for ROS Fuerte:
rosrun rviz rviz -d Data/rviz.vcg
- for ROS Groovy or newer versions:
rosrun rviz rviz -d Data/rviz.rviz
4. ORB_SLAM will receive the images from the topic /camera/image_raw. You can now play your rosbag or start your camera node.
Currently we do not support reading image files from disk. If you have a sequence with individual image files,
you will need to generate a bag from them. We provide a tool to do that: https://github.com/raulmur/BagFromImages
Tip: Use a roslaunch to launch ORB_SLAM, image_view and rviz from just one instruction. We provide an example:
- for ROS Fuerte:
roslaunch ExampleFuerte.launch
- for ROS Groovy or newer versions:
roslaunch ExampleGroovyOrNewer.launch
5) Example Sequence
We provide the settings and the rosbag of an example sequence in our lab. In this sequence you will see a loop closure and two relocalisation from a big viewpoint change.
1. Download the rosbag file: http://webdiis.unizar.es/~raulmur/orbslam/downloads/Example.bag.tar.gz
Uncompress the file.
2. Launch ORB_SLAM with the settings for the example sequence. You should have already uncompressed the vocabulary file (/Data/ORBvoc.txt.tar.gz)
- for ROS Fuerte:
roslaunch ExampleFuerte.launch
- for ROS Groovy or newer versions:
roslaunch ExampleGroovyOrNewer.launch
3. Once the ORB vocabulary has been loaded, start playing the bag
rosbag play --pause Example.bag
and press space to start
6) The Settings File
ORB_SLAM reads the camera calibration and setting parameters from a YAML file. We provide an example in Data/Settings.yaml, where you will find all parameters and their description. We use the camera calibration model of OpenCV.
Please make sure you write and call your own settings file for your camera (copy the example file and modify the calibration)
7) Failure Modes
You should expect to achieve good results in sequences similar to those in which we show results in our paper [1], in terms of camera movement and texture in the environment. In general our Monocular SLAM solution is expected to have a bad time in the following situations:
- Pure rotations in exploration
- Low texture environments
- Many (or big) moving objects, especially if they move slowly.
The system is able to initialize from planar and non-planar scenes. In the case of planar scenes, depending on the camera movement relative to the plane, it is possible that the system refuses to initialize, see the paper [1] for details.
8) Need Help?
If you have any trouble installing or running ORB-SLAM, contact the authors.