引導式測驗 · 中等 等級

相機校準測驗

測試內在參數和畸變參數、校準目標和不同視圖、重投影誤差、解析度處理和校準後驗證。

相關引導路徑相機校準
問題 1 的 8

相機校正估算的是什麼?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. 相機校正估算的是什麼?
    • The true dimensions of every object photographed
    • Camera intrinsics and lens-distortion parameters from known correspondences
    • The identities of objects in a scene
    • A guaranteed depth map from one image
  2. 為什麼要在不同的位置和方向拍攝目標?
    • Different views provide more informative constraints than near-duplicate images
    • The board pattern changes its physical dimensions
    • The camera automatically removes all distortion
    • Repeated identical views guarantee stable estimates
  3. 在相機校正中,重投影誤差比較的是什麼?
    • Observed target points with image points projected using the estimated camera parameters
    • Estimated focal length before and after optimization
    • Known 3D target coordinates directly with raw pixel coordinates, without projection
    • Detected target corners with their nearest neighbors before fitting a camera model
  4. 單憑較低的重投影誤差無法證明什麼?
    • That camera parameters were estimated
    • That the target corners were detected
    • That every downstream distance measurement is accurate in all scenes
    • That the fitting process used image points
  5. 何時應該重新檢查相機參數?
    • Never once a calibration file is saved
    • After changes to lens, focus, resolution, mounting, or image processing
    • Only after changing the file name
    • Whenever the scene contains a person
  6. Which set contains only calibration target families listed in OpenCV’s tutorial?
    • Chessboards, AprilTag boards, and projected checkerboard patterns
    • Circle grids, QR-code arrays, and laser-dot patterns
    • ChArUco boards, ArUco-only boards, and circular ring targets
    • Chessboards, ChArUco boards, and circle patterns
  7. How should calibration quality be validated?
    • Report only a mean residual without inspecting views
    • Use the calibration target as the only validation scene
    • Use held-out views and application-level measurements in addition to reprojection diagnostics
    • Tune until one training image looks undistorted
  8. What can monocular calibration alone not guarantee?
    • Metric depth for every point in a single image
    • An estimate of lens distortion
    • A camera matrix from known correspondences
    • A reprojection residual