Monocular SLAM is the visual SLAM variant that uses a single camera stream to estimate pose and reconstruct map structure - it is lightweight and widely accessible, but must resolve scale ambiguity through motion and optimization.

What Is Monocular SLAM?

• Definition: SLAM using one RGB camera without direct depth measurements.
• Primary Challenge: Absolute scale is unobservable from single-view geometry alone.
• Initialization Need: Requires sufficient parallax to triangulate initial landmarks.
• Common Systems: ORB-SLAM family and direct monocular pipelines.

Why Monocular SLAM Matters

• Hardware Simplicity: Minimal sensor setup for low-cost deployment.
• Wide Availability: Works with commodity cameras on phones and robots.
• Research Importance: Strong baseline for learning-augmented SLAM.
• Portability: Easy integration into embedded platforms.
• Foundation Layer: Can be extended with inertial fusion to recover scale.

Monocular SLAM Strategies

Feature-Based Methods:

• Track sparse keypoints and build map landmarks.
• Robust and interpretable.

Direct Methods:

• Optimize photometric error over image intensities.
• Dense usage of image information.

Visual-Inertial Extensions:

• Add IMU to resolve scale and improve robustness.
• Common in mobile and drone systems.

How It Works

Step 1:

• Track visual correspondences and estimate relative camera motion.

Step 2:

• Triangulate landmarks, optimize local map, and apply loop closure for drift correction.

Monocular SLAM is the most accessible SLAM configuration that delivers real-time mapping from a single camera while trading off direct metric scale observability - with good initialization and optimization, it performs remarkably well in many settings.