Math and Modeling for Camera Developers
summary
Derives the pinhole projection p' = f x / z and rewrites it with homogeneous coordinates so all transforms become one matrix, combines intrinsics and extrinsics into a projection matrix, and shows calibration by solving 12 unknowns with six known 3D→2D correspondences (and uses ≥8 points to estimate the essential matrix for pose). Covers stereo depth z = d·fx / disparity, feature matching with SIFT/SURF plus RANSAC, triangulation and bundle adjustment*, Brown–Conrady lens distortion, and Android Camera2 + IMU hooks for depth, AR placement and 3D reconstruction.
AI-written summary · about
chapters (10)
- 0:00Camera math and modeling intro
- 1:07Camera intrinsics basics
- 12:31Homogeneous coordinates
- 17:48Lens distortion adjustments
- 22:473D to 2D with intrinsics
- 31:20Camera calibration matrix
- 36:40Depth with stereo vision
- 41:143D reconstruction use cases
- 46:57Android Camera2 API
- 50:28Developer applications
related talks
39mNow smile (also in 3D)! Exploring AR, ML and Camera related APIs on Android
41mComputer Vision for Remote AR
45mUnderstanding the Mathematical Foundations of Spatial Computing
35mBuilding Camera Apps on Android Today
14mOddEyeCam: A Sensing Technique for Body-Centric Peephole Interaction using WFoV RGB and NFoV ...
31mBuilding Camera Features in a Snap
25mLibcamera: Making Complex Cameras Easy
14mDeepFisheye: Near-Surface Multi-Finger Tracking Technology Using Fisheye Camera