Lenses, Field of View, Focus, and Distortion in Robot Cameras

Why Lens Choice Changes What Your Robot Can Perceive

A robot camera is not just a sensor; the lens defines what parts of the world are visible, how big objects appear, and how reliably you can measure geometry. Two cameras with the same resolution can behave very differently if one uses a wide-angle lens and the other uses a narrow lens. In robotics, lens choice is a decision about task performance: coverage versus precision, robustness versus sensitivity, and speed versus accuracy.

Key Terms You Will Use When Selecting a Lens

• Focal length (f): how strongly the lens magnifies the scene. Shorter focal length generally means wider view; longer focal length means narrower view and more magnification.
• Field of view (FOV): the angular extent of the scene captured (horizontal/vertical/diagonal). Wider FOV sees more but usually increases distortion and reduces pixels-per-degree.
• Perspective effects: how depth changes apparent size and shape. Wide lenses exaggerate perspective (near objects look much larger than far ones).
• Aperture (f-number): controls how much light enters and influences depth of field and motion blur.
• Focus distance: the distance at which the image is sharpest.
• Depth of field (DoF): range of distances that appear acceptably sharp.
• Distortion: deviation from an ideal pinhole model; affects straight lines and geometric measurements.
• Shutter type: global shutter exposes all pixels at once; rolling shutter exposes line-by-line, which can warp moving scenes.

Decision-Driven Lens Selection: Navigation vs Manipulation

Decision 1: What must be visible at once?

Navigation (corridor following, obstacle avoidance, lane/line tracking) often benefits from wide coverage to see nearby obstacles and maintain situational awareness. Manipulation (grasping, insertion, picking) often benefits from higher angular resolution on a smaller workspace area.

Decision 2: How far away are the important features?

Lens choice should be driven by the distance range where the robot must make decisions. For example, a warehouse robot may need to detect obstacles from 0.5–5 m, while a manipulator may need accurate geometry at 0.2–0.8 m. A lens that is sharp at infinity but weak at close range can fail manipulation even if it works for navigation.

Decision 3: How much measurement accuracy do you need?

If you need to measure positions (lane center, line offset, marker pose), you must consider: (1) distortion, (2) pixels-per-feature, and (3) stability under motion (shutter). Wide-angle lenses can be excellent for detection but can degrade metric accuracy unless calibrated and undistorted.

Focal Length, Field of View, and Perspective Trade-offs

How focal length changes what you see

With a fixed sensor size, shorter focal length increases FOV. This means more scene content, but each object occupies fewer pixels. For detection tasks, that may be fine; for precise localization (e.g., grasp points), fewer pixels can increase jitter in estimated positions.

Continue in our app.

• Listen to the audio with the screen off.
• Earn a certificate upon completion.
• Over 5000 courses for you to explore!

Or continue reading below...

Download the app

Perspective effects that matter in robotics

• Wide-angle perspective exaggeration: nearby objects appear much larger relative to far objects. This can help avoid collisions (near obstacles stand out) but can make distance intuition harder and can amplify errors when projecting points to the ground plane.
• Narrower FOV “compression”: depth differences look smaller. This can make far features easier to track (more pixels on them) but reduces peripheral awareness.

Wide-angle coverage vs distortion

Wide-angle lenses often introduce stronger radial distortion (barrel distortion), bending straight lines outward. This is not just a visual artifact: it shifts where edges and corners appear, which directly impacts line following, lane estimation, and marker pose.

Focus, Depth of Field, and Aperture: Getting Sharp Frames in Real Robots

Focus distance and what “sharp” means for algorithms

Many vision algorithms rely on high-frequency detail: edges, corners, texture. Defocus blur reduces edge contrast and spreads corners over multiple pixels, which can cause failures in feature detection, marker decoding, and pose estimation. A frame can look “okay” to a human but still be too soft for reliable corner localization.

Aperture and depth of field trade-offs

A wider aperture (smaller f-number, e.g., f/1.8) lets in more light, enabling shorter exposure times (less motion blur), but reduces depth of field, making focus more sensitive to distance changes. A smaller aperture (larger f-number, e.g., f/8) increases depth of field (more of the scene is sharp) but reduces light, often forcing longer exposure (more motion blur) or higher gain (more noise).

Practical step-by-step: Choosing focus and aperture for a robot

1. Define the working distance band: e.g., manipulation at 30–70 cm; navigation at 0.5–5 m.
2. Set focus to the most critical distance: for manipulation, focus near the typical grasp distance; for navigation, focus farther (or use a lens designed for good sharpness across distance).
3. Pick aperture based on motion and lighting: if the robot moves fast, prioritize shorter exposure (often requires wider aperture or more light). If geometry must be stable across depth, prioritize more DoF (smaller aperture) and compensate with illumination.
4. Validate with algorithm-relevant targets: test on edges/corners similar to your task (printed tag, checkerboard, lane tape) at near and far distances.
5. Lock settings when possible: auto-focus and auto-exposure can change frame-to-frame, causing apparent scale/brightness shifts that destabilize tracking.

How to Detect Focus Problems in Frames (Operational Checks)

Symptoms you can see in the image

• Edges look thick and transitions are gradual rather than crisp.
• Fine texture disappears (fabric, printed patterns, small text-like detail on tags).
• Corner detectors become unstable: the same corner “moves” between frames even when the scene is static.

Practical step-by-step: A simple focus/blur diagnostic you can automate

You can quantify sharpness using a focus measure such as the variance of the Laplacian. Higher values generally indicate sharper images.

# Pseudocode (language-agnostic) for blur detection using Laplacian variance
img_gray = to_grayscale(frame)
lap = laplacian(img_gray)
sharpness = variance(lap)
if sharpness < threshold:
    flag_frame_as_blurry()

• How to set the threshold: collect sharp frames and intentionally defocused frames in your real lighting; choose a threshold that separates them.
• How to use it: reject blurry frames for pose estimation, or trigger a warning to adjust focus/lighting/exposure.

Focus vs motion blur: don’t confuse them

Defocus blur is usually uniform across the image (depending on depth), while motion blur often has a direction (streaking) and increases with robot speed or vibration. Both reduce sharpness metrics, but motion blur correlates strongly with exposure time and motion.

Rolling vs Global Shutter: Interaction with Robot Motion

What changes with rolling shutter

With a rolling shutter, different rows (or columns) are captured at slightly different times. If the robot or objects move during readout, straight lines can appear slanted, and shapes can warp. This can break geometric assumptions used in line fitting, homography estimation, and pose estimation.

When rolling shutter becomes a problem

• Fast yaw rotation of a mobile robot: vertical structures lean; lane lines can shift laterally in the image.
• Vibration from motors: introduces wavy distortions frame-to-frame, harming feature tracking.
• High-speed manipulators: tool motion can warp the perceived shape of the gripper or target.

Global shutter advantages

Global shutter exposes all pixels at once, preserving geometry under motion. This is especially valuable for accurate pose estimation, visual odometry, and any measurement where straightness and rigidity matter.

Practical step-by-step: Mitigating rolling shutter artifacts if you cannot change the camera

1. Reduce exposure time: shorter exposure reduces motion blur and can reduce the apparent severity of rolling artifacts (though readout timing still exists).
2. Increase illumination: enables shorter exposure without raising gain too much.
3. Stabilize the mount: reduce vibration with mechanical damping and rigid brackets.
4. Limit angular velocity during critical measurements: slow down turns when reading markers or aligning to lines.
5. Prefer features near the image center: rolling artifacts often worsen toward edges depending on motion direction and lens distortion.

Lens Distortion: Radial and Tangential, and Why Calibration Matters

Radial distortion (barrel and pincushion)

Radial distortion increases with distance from the image center. In barrel distortion (common in wide lenses), straight lines bow outward. In pincushion distortion, lines bow inward. For robots, this shifts where edges and corners appear, causing systematic errors in measurements.

Tangential distortion (decentering)

Tangential distortion happens when the lens is not perfectly aligned with the sensor. It causes asymmetrical warping, where distortion differs by direction. This can be subtle but can significantly affect precise pose estimation and mapping.

Why calibration matters for accurate geometry

Many robotics computations assume an ideal camera model. Distortion violates that assumption. Calibration estimates intrinsic parameters (including distortion coefficients) so you can undistort images or correctly project rays into 3D. Without calibration, errors are not random; they are systematic and can bias control decisions.

How Distortion Influences Measurements (Lines, Lanes, and Marker Pose)

Lane/line position errors

Line following often relies on detecting a line and computing its offset from the image center or its angle. Distortion can curve a physically straight line, causing:

• Biased lateral offset: the detected line appears shifted, especially near image edges.
• Incorrect heading estimate: curvature can be mistaken for line angle change.
• Inconsistent measurements across the frame: the same line yields different fits depending on where it appears.

Marker pose errors (e.g., fiducials)

Pose estimation from a planar marker depends on accurate corner positions. Distortion moves corners nonlinearly, which can cause:

• Translation bias: marker appears closer/farther or shifted sideways.
• Rotation bias: marker appears tilted when it is not.
• Jitter: small detection noise is amplified when corners are near distorted regions.

Practical step-by-step: A workflow to make measurements robust to distortion

1. Calibrate the camera using a known target (checkerboard or similar) across the full field of view you will use.
2. Verify calibration quality: check reprojection error and visually confirm that straight lines become straight after undistortion.
3. Undistort before measurement for tasks like line fitting and marker pose, or use a model that accounts for distortion during estimation.
4. Define a “trusted region”: if wide-angle distortion is strong at the edges, restrict critical measurements to the central area or ensure your calibration is accurate at the periphery.
5. Recalibrate when hardware changes: lens refocus, zoom changes, impacts, or temperature shifts can alter intrinsics and distortion.

Putting It Together: Lens Selection Checklist for Real Robot Builds

For navigation (coverage-first, stable under motion)

• FOV: wide to moderate, enough to see obstacles and path boundaries.
• Shutter: prefer global shutter if turning fast or doing visual odometry; otherwise mitigate rolling shutter with short exposure and good mounting.
• Distortion: expect more with wide lenses; plan for calibration and undistortion in the pipeline.
• Focus/DoF: set for mid-to-far distances; ensure sufficient DoF for near obstacles if needed.

For manipulation (precision-first, corner accuracy)

• FOV: moderate to narrow to increase pixels on the target and reduce distortion impact.
• Focus: tuned to the working distance; avoid autofocus hunting.
• Aperture/DoF: enough DoF to keep both gripper and target sharp; add light rather than opening aperture too much if DoF becomes too shallow.
• Distortion: still calibrate; even mild distortion can bias pose at close range.