引導式測驗 · 中等 等級

RGB-D 相機與深度感測器測驗

查看 RGB-D 對齊、立體和主動深度感測、相機內在特性、無效像素、同步和 3D 映射檢查。

問題 1 的 8

A color image and depth image come from separate sensors. What links their pixels correctly?

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. A color image and depth image come from separate sensors. What links their pixels correctly?
    • Matching image resolutions alone
    • A shared filename and frame index alone
    • Calibrated intrinsics and an extrinsic transform between the sensors
    • Only a color white-balance correction
  2. How does a stereo depth camera primarily estimate distance?
    • From emitted-light travel time
    • From disparity between calibrated views
    • From distortion of a projected pattern
    • From one RGB view without paired geometry
  3. Why can far-away stereo depth be sensitive to small matching errors?
    • Disparity is small at distance, so an error has a larger depth effect
    • Depth error must shrink when disparity becomes small
    • A small disparity error changes only image color
    • Stereo depth is independent of baseline
  4. A depth pixel is flagged invalid on shiny packaging. Which action best preserves measurement honesty?
    • Set missing values to zero and call them measured
    • Interpolate every invalid pixel and label it direct sensor output
    • Discard the validity mask after color alignment
    • Keep it marked missing or uncertain instead of treating it as measured depth
  5. What does the term RGB-D tell you about a camera system?
    • It provides color plus depth data, without specifying one sensing mechanism
    • It necessarily uses time-of-flight hardware
    • It necessarily uses two stereo imagers
    • It derives true range from RGB color alone
  6. Why can unsynchronized color and depth streams misalign a moving hand?
    • Their physical baseline changes with hand motion
    • Synchronization is irrelevant once intrinsics are known
    • All depth pixels become invalid whenever a hand moves
    • They may describe different poses of the hand at different capture times
  7. What did KinectFusion demonstrate beyond one depth snapshot?
    • That one depth image is already a complete 3D map
    • Tracking and fusing moving depth-camera frames into an indoor surface map
    • That RGB alone supplies metric depth without calibration
    • That moving-camera tracking can be omitted during fusion
  8. Which check is most useful before using an RGB-D cloud for physical measurement?
    • Inspect only whether colored rendering looks smooth
    • Fill every depth hole and call the values observed
    • Compare dimensions and invalid-depth coverage against trusted references
    • Compare file size rather than physical dimensions