Data Association in ORBSLAM2
本文最后更新于：May 7, 2023 pm
[TOC]
Overview
Map Points & KeyFrames
 Each keyframe \(K_i\) stores
ID
1
2
3static long unsigned int nNextId;
long unsigned int mnId;
const long unsigned int mnFrameId; camera pose
 camera intrinsics
KeyPoints related
1
2
3
4
5
6
7
8
9
10
11
12
13// Number of KeyPoints
const int N;
// KeyPoints, stereo coordinate and descriptors (all associated by an index)
const std::vector<cv::KeyPoint> mvKeys;
const std::vector<cv::KeyPoint> mvKeysUn;
const std::vector<float> mvuRight; // negative value for monocular points
const std::vector<float> mvDepth; // negative value for monocular points
const cv::Mat mDescriptors;
// BoW
DBoW2::BowVector mBowVec;
DBoW2::FeatureVector mFeatVec;MapPoints associated to keypoints
1
2
3
4
5
6std::vector<MapPoint*> mvpMapPoints;
void KeyFrame::AddMapPoint(MapPoint *pMP, const size_t &idx) {
unique_lock<mutex> lock(mMutexFeatures);
mvpMapPoints[idx] = pMP;
}
 Each map point \(P_i\) stores
ID
1
2
3
4
5long unsigned int mnId;
static long unsigned int nNextId;
long int mnFirstKFid;
long int mnFirstFrame;Reference KeyFrame (TODO: Anchor Frame?)
1
KeyFrame* mpRefKF;
 3D position
cv::Mat mWorldPos;
 Mean viewing direction
cv::Mat mNormalVector;
 Best descriptor to fast matching
cv::Mat mDescriptor;
The maximum and minimum distances
1
2
3// Scale invariance distances
float mfMinDistance;
float mfMaxDistance;Keyframes observing the point and associated index in keyframe
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15std::map<KeyFrame*,size_t> mObservations;
int nObs;
void MapPoint::AddObservation(KeyFrame* pKF, size_t idx)
{
unique_lock<mutex> lock(mMutexFeatures);
if(mObservations.count(pKF))
return;
mObservations[pKF]=idx;
if(pKF>mvuRight[idx]>=0)
nObs+=2;
else
nObs++;
}Tracking counters
1
2int mnVisible;
int mnFound;
exigent culling mechanism
MapPoints
1
void LocalMapping::MapPointCulling();
KeyFrame
1
void LocalMapping::KeyFrameCulling();
Covisible Graph (KeyFrame Connections)
Covisibility information between keyframes is very useful in several tasks of our system, and is represented as an undirected weighted graph.
 Node (KeyFrame)
 Edge (Weight): 关键帧之间共视的 路标点数，至少15
1 

Spanning Tree
The system builds incrementally a spanning tree from the initial keyframe, which provides a connected subgraph of the covisibility graph with minimal number of edges. When a new keyframe is inserted, it is included in the tree linked to the keyframe which shares most point observations, and when a keyframe is erased by the culling policy, the system updates the links affected by that keyframe.
 Parent Node:
KeyFrame* mpParent;
, the pair with highest covisibility weight  Child Nodes:
std::set<KeyFrame*> mspChildrens;
1 

 TrackLocalMap 里 UpdateLocalKeyFrames更新局部地图中的关键帧
 闭环矫正时 优化 Essential Graph
Loop Edges
std::set<KeyFrame*> mspLoopEdges;
1 

Essential Graph
The Essential Graph contains the spanning tree, the subset of edges from the covisibility graph with high covisibility (\(\theta_{min} = 100\)), and the loop closure edges, resulting in a strong network of cameras.
 Node (KeyFrame)
 Edge
 Spanning tree edges
 Covisibility graph edges (weight > 100):
pKF>GetCovisiblesByWeight(100)
 Loop edges