Core components of visual SLAM
SLAM system structure.
tracking estimates per-frame pose, mapping builds and refines the 3D map, loop closure detects revisits and corrects drift.
confusing tracking with mapping or omitting loop closure.
What's really being asked
The question checks conceptual understanding of how a moving camera localizes itself while building a map at the same time, and why drift makes loop closure necessary. Strong answers separate the responsibilities cleanly.
The full answer
Tracking is the front end: it extracts and matches features between consecutive frames, estimates the camera pose for each frame typically by minimizing reprojection error, and runs fast enough for real time. Mapping is the back end: it triangulates new 3D points, maintains keyframes, and refines poses and landmarks together using local bundle adjustment to keep the map accurate. Loop closure detects when the camera revisits a known location, usually via appearance-based place recognition like bag-of-words, then performs pose-graph or global bundle adjustment to redistribute accumulated drift, making the trajectory globally consistent.
The mistakes people make
Saying tracking builds the map, conflating the front and back ends. Or omitting loop closure, which leaves the map drifting unboundedly over long runs. Or describing it as pure odometry with no global optimization.
What usually comes next
What is the difference between SLAM and visual odometry, the answer being loop closure and a globally consistent map. How does bundle adjustment work. Why is drift inevitable without loop closure. How do you detect loops robustly and avoid false positives.
A concrete example
A robot drives a loop around a room. Tracking estimates each frame's pose in real time and mapping builds a sparse point cloud of landmarks, but small per-frame errors accumulate as drift, so the start and end of the trajectory do not align. When the robot returns to the entrance, loop closure recognizes the scene through appearance-based place recognition, adds a constraint linking the current pose to the earlier one, and runs global pose-graph optimization that snaps the trajectory closed and redistributes the error, correcting the whole map so it is globally consistent.
Interview question
What primarily distinguishes full visual SLAM from plain visual odometry?
- a.SLAM uses RGB cameras while odometry requires LiDAR
- b.SLAM adds loop closure and global optimization for a consistent mapCorrect
- c.Odometry estimates 3D points while SLAM does not build a map
- d.SLAM runs offline only, odometry runs in real time
Why? this is the answer
Odometry tracks incremental motion but drifts; SLAM adds loop closure and global optimization to maintain a consistent map. Sensor type and offline-only claims are incorrect generalizations.
Just read this? Test yourself on what you have been reading.
Read the original → thinkautonomous.ai
- #slam
- #tracking
- #mapping
- #loop-closure
- #robotics
Put your scrolling time to good use
Learn one idea, try a quiz and save useful cards for revision. Tezvyn makes it easy to learn and stay current in your tech field, a few minutes at a time.
The iPhone app is on the way
We are building it. Until it lands, nothing here is held back from you: every interview card, your saved cards, streaks and the job board all work in Safari, plus hundreds of free practice quizzes of thirty questions each. Sign in and it all carries over to the app the day it arrives.
Want it as an icon? Tap Share at the bottom of Safari, then Add to Home Screen. It opens full screen and the cards you have read stay available offline.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles