stereo slam
**Stereo SLAM** is the **visual SLAM approach that uses two synchronized cameras with known baseline to estimate depth directly and preserve metric scale** - this reduces ambiguity and improves robustness compared with monocular setups.
**What Is Stereo SLAM?**
- **Definition**: SLAM pipeline using left-right image pairs plus temporal tracking.
- **Scale Advantage**: Known baseline enables direct depth and absolute scale recovery.
- **Map Quality**: Better initial landmark depth than monocular triangulation.
- **Runtime Components**: Stereo matching, visual odometry, mapping, and loop closure.
**Why Stereo SLAM Matters**
- **Metric Reliability**: Maintains physically meaningful distances without extra sensors.
- **Faster Initialization**: Immediate depth estimates reduce startup fragility.
- **Tracking Robustness**: More stable under pure rotational motions than monocular.
- **Navigation Utility**: Strong fit for mobile robots and autonomous platforms.
- **Operational Tradeoff**: Higher compute due to stereo matching stage.
**Stereo SLAM Pipeline**
**Depth Estimation**:
- Compute disparity between synchronized camera views.
- Convert disparity to depth using calibrated baseline.
**Temporal Tracking**:
- Track landmarks and estimate pose over time.
- Fuse stereo depth with temporal correspondences.
**Global Optimization**:
- Detect loop closures and optimize pose graph for consistency.
- Update map landmarks after global correction.
**How It Works**
**Step 1**:
- Generate depth from stereo pairs and initialize map with metric landmarks.
**Step 2**:
- Run temporal pose tracking and periodic global optimization to maintain consistency.
Stereo SLAM is **a strong metric-accurate localization and mapping solution that balances visual richness with reliable scale estimation** - it remains a practical default when dual-camera hardware is available.