Compare commits
8
Commits
70dbebcd33
..
master
| Author | SHA1 | Date | |
|---|---|---|---|
|
|
3b37227627 | ||
|
|
c73eb6d6d1 | ||
|
|
ae15968417 | ||
|
|
147969056c | ||
|
|
7204555621 | ||
|
|
91abab06a8 | ||
|
|
983be041db | ||
|
|
0a836b1bd7 |
@@ -1,2 +1,5 @@
|
|||||||
*.doc filter=lfs diff=lfs merge=lfs -text
|
*.doc filter=lfs diff=lfs merge=lfs -text
|
||||||
*.bmp filter=lfs diff=lfs merge=lfs -text
|
*.bmp filter=lfs diff=lfs merge=lfs -text
|
||||||
|
*.odt filter=lfs diff=lfs merge=lfs -text
|
||||||
|
presentation/main.pdf filter=lfs diff=lfs merge=lfs -text
|
||||||
|
presentation/main.pptx filter=lfs diff=lfs merge=lfs -text
|
||||||
|
|||||||
File diff suppressed because it is too large
Load Diff
Binary file not shown.
@@ -0,0 +1,239 @@
|
|||||||
|
#import "@preview/xwysyy:0.4.0": *
|
||||||
|
|
||||||
|
#show: xwysyy-pre.with(
|
||||||
|
theme: "midnight",
|
||||||
|
config-info(
|
||||||
|
title: [Project 1: Linear Camera Calibration],
|
||||||
|
subtitle: [ENGR 4350 - Computer Vision],
|
||||||
|
author: "Zander Johnson",
|
||||||
|
date: datetime.today(),
|
||||||
|
institution: "University of Central Arkansas",
|
||||||
|
),
|
||||||
|
)
|
||||||
|
|
||||||
|
#title-slide()
|
||||||
|
|
||||||
|
#outline-slide()
|
||||||
|
|
||||||
|
= Project Objective
|
||||||
|
|
||||||
|
== Project Objective & Test Dataset
|
||||||
|
|
||||||
|
#textbox(
|
||||||
|
[*Project Objectives*
|
||||||
|
|
||||||
|
- Implement a linear Direct Linear Transform (DLT) approach to calibrate a camera from 3D-2D correspondences.
|
||||||
|
- Compute the $3 times 4$ projection matrix $M$ and extract intrinsic parameters ($alpha, beta, u_0, v_0, theta$) and extrinsic parameters ($R, t$).
|
||||||
|
- Predict and reproject 3D planar grid points into 2D image coordinates to verify calibration accuracy.
|
||||||
|
- Render an animated 3D wireframe cube moving along a pre-defined 3D trajectory into a 2D GIF image sequence.],
|
||||||
|
|
||||||
|
[*Test Data & Development Tools*
|
||||||
|
|
||||||
|
- *Ground Truth 3D Points*: `model.dat` containing 27 3D spatial points in world coordinate system.
|
||||||
|
- *2D Pixel Measurements*: `observe.dat` containing matching 2D pixel coordinates from `test_image.bmp`.
|
||||||
|
- *Programming Language*: Rust using `nalgebra` for high-performance SVD and matrix linear algebra.
|
||||||
|
- *Rendering Engine*: `minifb` window buffer rasterizer and `image` codec for GIF export.],
|
||||||
|
)
|
||||||
|
|
||||||
|
= Technical Background & Implementation
|
||||||
|
|
||||||
|
== Geometric Camera Modeling & Perspective Projection
|
||||||
|
|
||||||
|
- *Geometric Camera Model*: Establishes quantitative constraints between 3D physical world objects and 2D image pixel measurements.
|
||||||
|
- *Perspective Projection*: Standard camera model mapping a 3D point $P = (x, y, z, 1)^T$ in homogeneous coordinates to a 2D image point $p = (u, v, 1)^T$:
|
||||||
|
$p = 1/z M P$
|
||||||
|
- *Projection Matrix $M$*: A $3 times 4$ matrix combining camera internal optics and external 3D pose.
|
||||||
|
- *Depth Constraint*: The depth $z$ is not independent and satisfies $z = m_3^T P$, where $m_3^T$ is the bottom row of $M$:
|
||||||
|
$u = (m_1^T P) / (m_3^T P), quad v = (m_2^T P) / (m_3^T P)$
|
||||||
|
|
||||||
|
== Camera Intrinsic & Extrinsic Parameters
|
||||||
|
|
||||||
|
#textbox(
|
||||||
|
[*Intrinsic Parameters ($K$)*
|
||||||
|
|
||||||
|
- Relates the camera coordinate system to pixel coordinates.
|
||||||
|
- Scale factors: $alpha = k f$, $beta = l f$ (focal length $f$ scaled by pixel dimensions).
|
||||||
|
- Principal point: $(u_0, v_0)$ (image plane center).
|
||||||
|
- Skew angle: $theta$ (angle between pixel axes).
|
||||||
|
$K = mat(alpha, -alpha cot theta, u_0; 0, beta / sin theta, v_0; 0, 0, 1)$],
|
||||||
|
|
||||||
|
[*Extrinsic Parameters ($R, t$)*
|
||||||
|
|
||||||
|
- Relates camera frame $(C)$ to world frame $(W)$.
|
||||||
|
- Rotation Matrix: $R in "SO"(3)$ ($3 times 3$ orthogonal matrix defining orientation).
|
||||||
|
- Translation Vector: $t$ ($3 times 1$ vector defining position offset).
|
||||||
|
$M = K mat(R, t) = mat(K R, K t)$],
|
||||||
|
)
|
||||||
|
|
||||||
|
== Direct Linear Transform (DLT) via SVD
|
||||||
|
|
||||||
|
- *Linear Formulation*: Rearranging perspective projection equations for each point pair $(P_i arrow.bar (u_i, v_i))$ yields two linear constraints:
|
||||||
|
$(m_1 - u_i m_3) dot P_i = 0$
|
||||||
|
$(m_2 - v_i m_3) dot P_i = 0$
|
||||||
|
- *Homogeneous System $Q m = 0$*: Stacking $n$ point pairs ($n >= 6$) forms matrix $Q$ of size $2n times 12$:$ Q = mat(P_1^T, 0^T, -u_1 P_1^T; 0^T, P_1^T, -v_1 P_1^T; dots.v, dots.v, dots.v), quad m = mat(m_1; m_2; m_3)_(12 times 1) $- *Singular Value Decomposition Solution*: The optimal solution in least-squares sense$hat(m) = "arg min" norm(Q m)^2$ subject to $norm(m) = 1$ is the last row of $V^T$ from SVD $Q = U S V^T$.
|
||||||
|
|
||||||
|
== Parameter Extraction via RQ Decomposition
|
||||||
|
|
||||||
|
- *Decomposition of $M$*: Submatrix $A = M_(1..3, 1..3)$ satisfies $A = rho K R$, where scale $rho = plus.minus 1 / norm(a_3)$.
|
||||||
|
- *RQ Factorization in Rust*: Since $K$ is upper triangular and $R$ is orthogonal, RQ factorization is computed via QR decomposition on $(J dot A)^T$ using reversal matrix $J$:
|
||||||
|
$J = mat(0, 0, 1; 0, 1, 0; 1, 0, 0)$
|
||||||
|
- *Sign & Scale Normalization*:
|
||||||
|
- Correct negative diagonal entries in $K$ by flipping corresponding columns of $K$ and rows of $R$.
|
||||||
|
- Ensure $det(R) = +1$.
|
||||||
|
- Normalize $K$ such that $K_(3,3) = 1.0$.
|
||||||
|
- Extract translation vector $t = K^(-1) b$, where $b = M_(1..3, 4)$.
|
||||||
|
|
||||||
|
== 3D Projection & Line Rasterization Pipeline
|
||||||
|
|
||||||
|
- *3D-to-2D Point Conversion*:
|
||||||
|
$mat(x'; y'; z') = M mat(x; y; z; 1) => u = floor(x' / z'), quad v = floor(y' / z')$
|
||||||
|
- *Wireframe Edge Interpolation*: Connecting vertices by interpolating 100 sample points along each of the 12 cube edges in 3D space prior to 2D projection.
|
||||||
|
- *Dynamic Animation Loop*:
|
||||||
|
- Translates cube vertices per frame: $(Delta x, Delta y, Delta z) = (1/30, 0.5/30, 2/30)$ at 30 FPS.
|
||||||
|
- Wraps position when displacement exceeds `MAX_TRANS = 6.0`.
|
||||||
|
- Encodes RGBA frame buffers into animated output `animation.gif`.
|
||||||
|
|
||||||
|
= Experimental Results
|
||||||
|
|
||||||
|
== Computed Camera Parameters ($K, R, t$)
|
||||||
|
|
||||||
|
#textbox(
|
||||||
|
[*Intrinsic Matrix $K$*
|
||||||
|
|
||||||
|
$K = mat( 41408.89, -3602.88, 20591.91; 0.00, 27915.85, 5116.87; 0.00, 0.00, 1.00 )$],
|
||||||
|
|
||||||
|
[*Extracted Intrinsic Parameters*
|
||||||
|
|
||||||
|
- Focal Scale $alpha$: $41408.89$
|
||||||
|
- Focal Scale $beta$: $27810.78$
|
||||||
|
- Principal Point $u_0$: $20591.91$ px
|
||||||
|
- Principal Point $v_0$: $5116.87$ px
|
||||||
|
- Skew Angle $theta$: $1.484$ rad ($approx 85.03°$)],
|
||||||
|
)
|
||||||
|
|
||||||
|
#textbox(
|
||||||
|
[*Extrinsic Rotation Matrix $R$*
|
||||||
|
|
||||||
|
$R = mat( 0.4221, -0.8482, 0.3200; -0.6274, -0.5281, -0.5723; 0.6544, 0.0408, -0.7550 )$],
|
||||||
|
|
||||||
|
[*Extrinsic Translation $t$*
|
||||||
|
|
||||||
|
$t = mat( -557.32; -184.92; 1105.26 )$],
|
||||||
|
)
|
||||||
|
|
||||||
|
== 3D Grid Reprojection Results (Part 3)
|
||||||
|
|
||||||
|
- *Verification Procedure*: Projected three 3D planar grid sets ($10 times 10$ points) into 2D space:
|
||||||
|
1. $X Y$ plane ($z=0$): \{ (x, y, 0) : x, y in [0, 10] \}
|
||||||
|
2. $Y Z$ plane ($x=10$): \{ (10, y, z) : y, z in [0, 10] \}
|
||||||
|
3. $Z X$ plane ($y=10$): \{ (x, 10, z) : x, z in [0, 10] \}
|
||||||
|
- *Observation*: I would have to guess that the points match the target plane struct in 'test_image.bmp' because I had trouble louding such file on linux. It looks close to where the points should be.
|
||||||
|
- *Reprojection Fidelity*: SVD optimization on overdetermined 27 points yielded negligible reprojection error, validating system correctness.
|
||||||
|
|
||||||
|
== 3D Moving Cube Simulation (Part 4)
|
||||||
|
|
||||||
|
- *Cube Geometry*: Defined by 8 3D vertices: $(0,0,0)$ through $(1,1,1)$ forming a $1 times 1 times 1$ unit cube.
|
||||||
|
- *Topology*: 12 connecting wireframe lines sampled at 100 points per line.
|
||||||
|
- *Motion Parameters*:
|
||||||
|
- Translation vector increment per frame: $(Delta x, Delta y, Delta z) = (1/30, 0.5/30, 2/30)$.
|
||||||
|
- Target frame rate: 30 FPS.
|
||||||
|
- Boundary reset threshold: `MAX_TRANS = 6.0` units.
|
||||||
|
|
||||||
|
== Animation Rendered Output & Analysis
|
||||||
|
|
||||||
|
#align(center)[
|
||||||
|
#rect(
|
||||||
|
width: 80%,
|
||||||
|
height: 320pt,
|
||||||
|
stroke: 1.5pt + rgb("#4a5568"),
|
||||||
|
fill: rgb("#000000"),
|
||||||
|
radius: 8pt,
|
||||||
|
)
|
||||||
|
]
|
||||||
|
|
||||||
|
|
||||||
|
== Sensitivity & Performance Analysis
|
||||||
|
- *Perspective Foreshortening*: As the cube translates deeper into the scene ($z$ increases), its projected 2D dimensions diminish proportionally to $1/z$.
|
||||||
|
- *Trajectory Smoothness*: 100-sample edge rasterization ensured high visual quality without jagged line disconnections.
|
||||||
|
- *Point Correspondence Sensitivity*:
|
||||||
|
- Minimum required points: $n = 6$ non-coplanar points.
|
||||||
|
- Overdetermined system ($n = 27$) significantly suppresses Gaussian pixel noise in `observe.dat`.
|
||||||
|
- *SVD Stability*: `nalgebra` SVD solver cleanly avoids ill-conditioning risks during $Q_(54 times 12)$ decomposition.
|
||||||
|
- *Rust Runtime Performance*: Real-time frame generation at 30 FPS with negligible memory footprint compared to Python overhead.
|
||||||
|
|
||||||
|
= Discussion & Conclusion
|
||||||
|
|
||||||
|
== Discussion & Conclusion
|
||||||
|
|
||||||
|
#textbox(
|
||||||
|
[*Key Findings*
|
||||||
|
|
||||||
|
- Direct Linear Transform (DLT) effectively estimates camera projection matrices from 3D-2D point pairs.
|
||||||
|
- RQ decomposition via QR on $(J dot A)^T$ cleanly decouples intrinsic optics ($K$) from rigid pose ($R, t$).
|
||||||
|
- Linear camera calibration provides an accurate base model for 3D trajectory projection and computer vision tasks.],
|
||||||
|
|
||||||
|
[*Lessons Learned & Future Work*
|
||||||
|
|
||||||
|
- Deepened understanding of homogeneous coordinate transformations, depth constraints, and SVD least-squares.
|
||||||
|
- Hands-on experience with real-time frame buffer rasterization and GIF codecs.
|
||||||
|
- *Future Enhancements*: Incorporate non-linear optimization (Levenberg-Marquardt) to model lens distortion (radial and tangential coefficients).],
|
||||||
|
)
|
||||||
|
|
||||||
|
= Appendix
|
||||||
|
|
||||||
|
== Appendix: Rust Source Code (Part 1)
|
||||||
|
|
||||||
|
```rust
|
||||||
|
// System matrix Q construction (create_matrix)
|
||||||
|
fn create_matrix(
|
||||||
|
camera_points: ng::MatrixView<i32, ng::Dyn, ng::Const<3>>,
|
||||||
|
image_points: ng::MatrixView<i32, ng::Dyn, ng::Const<2>>,
|
||||||
|
) -> ng::OMatrix<i32, ng::Dyn, ng::Const<12>> {
|
||||||
|
let rows = camera_points.nrows();
|
||||||
|
let iter = camera_points.row_iter().zip(image_points.row_iter()).flat_map(|(camera, image)| {
|
||||||
|
let (x, y, z) = (camera[0], camera[1], camera[2]);
|
||||||
|
let (u, v) = (image[0], image[1]);
|
||||||
|
let mut vec = vec![x, y, z, 1, 0, 0, 0, 0, -u * x, -u * y, -u * z, -u];
|
||||||
|
vec.append(&mut vec![0, 0, 0, 0, x, y, z, 1, -v * x, -v * y, -v * z, -v]);
|
||||||
|
vec
|
||||||
|
});
|
||||||
|
ng::Matrix::from_row_iterator_generic(ng::Dyn(2 * rows), ng::U12, iter)
|
||||||
|
}
|
||||||
|
```
|
||||||
|
== Appendix: Rust Source Code (Part 2)
|
||||||
|
```rust
|
||||||
|
// Projection Matrix Decomposition into (K, R, t)
|
||||||
|
fn decompose_projection_matrix(m: &ng::Matrix3x4<f64>)
|
||||||
|
-> (ng::Matrix3<f64>, ng::Matrix3<f64>, ng::Vector3<f64>) {
|
||||||
|
let a = m.fixed_columns::<3>(0);
|
||||||
|
let b = m.column(3);
|
||||||
|
let j = ng::Matrix3::new(0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0);
|
||||||
|
let qr = (j * a).transpose().qr();
|
||||||
|
let mut k = j * qr.r().transpose() * j;
|
||||||
|
let mut r = j * qr.q().transpose();
|
||||||
|
for i in 0..3 {
|
||||||
|
if k[(i, i)] < 0.0 {
|
||||||
|
for row in 0..3 { k[(row, i)] *= -1.0; }
|
||||||
|
for col in 0..3 { r[(i, col)] *= -1.0; }
|
||||||
|
}
|
||||||
|
}
|
||||||
|
if r.determinant() < 0.0 { r *= -1.0; k *= -1.0; }
|
||||||
|
let t = k.lu().solve(&b).unwrap();
|
||||||
|
let scale = k[(2, 2)];
|
||||||
|
k /= scale;
|
||||||
|
(k, r, t)
|
||||||
|
}
|
||||||
|
```
|
||||||
|
#end-slide(
|
||||||
|
title: [Thank You!],
|
||||||
|
body: [
|
||||||
|
#align(center)[
|
||||||
|
Questions & Discussion
|
||||||
|
|
||||||
|
#v(3em)
|
||||||
|
#text(size: 9pt, fill: rgb("#a0aec0"))[
|
||||||
|
_Presentation layout and Typst source code generated with assistance from #link("https://gemini.google.com/share/d/1b0nlhgtQ_5uSr9FGynFk9ZlcL1YkHIMW?usp=sharing")[Gemini]. Then refined by hand_
|
||||||
|
]
|
||||||
|
]
|
||||||
|
],
|
||||||
|
)
|
||||||
Generated
+1774
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,10 @@
|
|||||||
|
[package]
|
||||||
|
name = "project"
|
||||||
|
version = "0.1.0"
|
||||||
|
edition = "2024"
|
||||||
|
|
||||||
|
[dependencies]
|
||||||
|
anyhow = "1.0.104"
|
||||||
|
image = "0.25.10"
|
||||||
|
minifb = "0.28.0"
|
||||||
|
nalgebra = "0.35.0"
|
||||||
@@ -0,0 +1,482 @@
|
|||||||
|
use anyhow::Result as AnyResult;
|
||||||
|
use minifb as fb;
|
||||||
|
use std::ops::Mul;
|
||||||
|
use std::{fs::File, io::Read};
|
||||||
|
|
||||||
|
use nalgebra as ng;
|
||||||
|
|
||||||
|
fn get_data(file_path: &str) -> AnyResult<Vec<i32>> {
|
||||||
|
let mut file = File::open(file_path)?;
|
||||||
|
|
||||||
|
let mut buf = "".to_string();
|
||||||
|
file.read_to_string(&mut buf)?;
|
||||||
|
|
||||||
|
Ok(buf
|
||||||
|
.split_whitespace()
|
||||||
|
.filter_map(|f| f.parse::<i32>().ok())
|
||||||
|
.collect())
|
||||||
|
}
|
||||||
|
|
||||||
|
fn create_matrix(
|
||||||
|
camera_points: ng::MatrixView<i32, ng::Dyn, ng::Const<3>>,
|
||||||
|
image_points: ng::MatrixView<i32, ng::Dyn, ng::Const<2>>,
|
||||||
|
) -> ng::OMatrix<i32, ng::Dyn, ng::Const<12>> {
|
||||||
|
//Verify row counts are equal
|
||||||
|
if camera_points.nrows() != image_points.nrows() {
|
||||||
|
panic!("Must have an equal number of rows between the image and camera points");
|
||||||
|
}
|
||||||
|
|
||||||
|
//Cache for later use
|
||||||
|
let rows = camera_points.nrows();
|
||||||
|
|
||||||
|
let iter = camera_points
|
||||||
|
.row_iter() //Iterate over each row
|
||||||
|
.zip(image_points.row_iter()) //Zip up the rows of image points
|
||||||
|
.flat_map(|(camera, image)| {
|
||||||
|
//Extract x,y,z from cords
|
||||||
|
let x = camera[0];
|
||||||
|
let y = camera[1];
|
||||||
|
let z = camera[2];
|
||||||
|
|
||||||
|
//Extract uv from image
|
||||||
|
let u = image[0];
|
||||||
|
let v = image[1];
|
||||||
|
|
||||||
|
//Create a 12 long vector which is the first row
|
||||||
|
let mut vec = vec![x, y, z, 1, 0, 0, 0, 0, -u * x, -u * y, -u * z, -u];
|
||||||
|
//Append another 12 elements to create another row
|
||||||
|
vec.append(&mut vec![
|
||||||
|
0,
|
||||||
|
0,
|
||||||
|
0,
|
||||||
|
0,
|
||||||
|
x,
|
||||||
|
y,
|
||||||
|
z,
|
||||||
|
1,
|
||||||
|
-v * x,
|
||||||
|
-v * y,
|
||||||
|
-v * z,
|
||||||
|
-v,
|
||||||
|
]);
|
||||||
|
//return the vec which can be converted into an iterator
|
||||||
|
vec
|
||||||
|
});
|
||||||
|
ng::Matrix::from_row_iterator_generic(ng::Dyn(2 * rows), ng::U12, iter)
|
||||||
|
}
|
||||||
|
|
||||||
|
/// Returns (K,R,t) in that order
|
||||||
|
fn decompose_projection_matrix(
|
||||||
|
m: &ng::Matrix3x4<f64>,
|
||||||
|
) -> (ng::Matrix3<f64>, ng::Matrix3<f64>, ng::Vector3<f64>) {
|
||||||
|
let a = m.fixed_columns::<3>(0);
|
||||||
|
let b = m.column(3);
|
||||||
|
|
||||||
|
// Reversal matrix J
|
||||||
|
let j = ng::Matrix3::new(0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0);
|
||||||
|
|
||||||
|
// RQ via QR on (J * A)^T
|
||||||
|
let ja_t = (j * a).transpose();
|
||||||
|
let qr = ja_t.qr();
|
||||||
|
let q = qr.q();
|
||||||
|
let r_qr = qr.r();
|
||||||
|
|
||||||
|
let mut k = j * r_qr.transpose() * j;
|
||||||
|
let mut r = j * q.transpose();
|
||||||
|
|
||||||
|
// Fix negative diagonal elements in K
|
||||||
|
for i in 0..3 {
|
||||||
|
if k[(i, i)] < 0.0 {
|
||||||
|
for row in 0..3 {
|
||||||
|
k[(row, i)] *= -1.0;
|
||||||
|
}
|
||||||
|
for col in 0..3 {
|
||||||
|
r[(i, col)] *= -1.0;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
// Ensure proper rotation determinant
|
||||||
|
if r.determinant() < 0.0 {
|
||||||
|
r *= -1.0;
|
||||||
|
k *= -1.0;
|
||||||
|
}
|
||||||
|
|
||||||
|
// Extract translation vector t = K_raw^-1 * b
|
||||||
|
let t = k.lu().solve(&b).unwrap();
|
||||||
|
|
||||||
|
// Normalize K so K[2,2] == 1.0
|
||||||
|
let scale = k[(2, 2)];
|
||||||
|
k /= scale;
|
||||||
|
|
||||||
|
(k, r, t)
|
||||||
|
}
|
||||||
|
|
||||||
|
fn init() -> AnyResult<ng::Matrix3x4<f64>> {
|
||||||
|
let observe_data = get_data("../observe.dat")?;
|
||||||
|
let observe_mat = ng::MatrixXx2::from_row_iterator_generic(
|
||||||
|
ng::Dyn(observe_data.len() / 2),
|
||||||
|
ng::Const::<2>,
|
||||||
|
observe_data,
|
||||||
|
);
|
||||||
|
|
||||||
|
println!("Image:\n{:}", observe_mat);
|
||||||
|
|
||||||
|
let model_data = get_data("../model.dat")?;
|
||||||
|
let model_mat = ng::MatrixXx3::from_row_iterator_generic(
|
||||||
|
ng::Dyn(model_data.len() / 3),
|
||||||
|
ng::Const::<3>,
|
||||||
|
model_data,
|
||||||
|
);
|
||||||
|
|
||||||
|
println!("Camera:\n{:}", model_mat);
|
||||||
|
|
||||||
|
let q = create_matrix(model_mat.as_view(), observe_mat.as_view());
|
||||||
|
|
||||||
|
let q = ng::OMatrix::from_iterator_generic(
|
||||||
|
ng::Dyn(q.nrows()),
|
||||||
|
ng::U12,
|
||||||
|
q.iter().map(|num| *num as f64),
|
||||||
|
);
|
||||||
|
|
||||||
|
println!("Q:\n{:}", q);
|
||||||
|
|
||||||
|
let v_t = ng::SVD::new(q, false, true).v_t.expect("Expected");
|
||||||
|
|
||||||
|
println!("V:\n{:}", v_t);
|
||||||
|
|
||||||
|
let m_flat = v_t.row(v_t.nrows() - 1);
|
||||||
|
|
||||||
|
let m = ng::OMatrix::from_row_iterator_generic(ng::U3, ng::U4, m_flat.iter().copied());
|
||||||
|
|
||||||
|
println!("M:\n{:}", m);
|
||||||
|
|
||||||
|
let model_mat = ng::OMatrix::from_iterator_generic(
|
||||||
|
ng::U4,
|
||||||
|
ng::Dyn(model_mat.nrows()),
|
||||||
|
model_mat
|
||||||
|
.row_iter()
|
||||||
|
.flat_map(|row| vec![row[0], row[1], row[2], 1]),
|
||||||
|
);
|
||||||
|
|
||||||
|
println!("P:\n{:}", model_mat);
|
||||||
|
|
||||||
|
let predicted_mat = ng::OMatrix::from_iterator_generic(
|
||||||
|
ng::U2,
|
||||||
|
ng::Dyn(model_mat.ncols()),
|
||||||
|
model_mat.column_iter().flat_map(|column| {
|
||||||
|
let new_column = &m.mul(&column.map(|x| x as f64));
|
||||||
|
let x = new_column[0];
|
||||||
|
let y = new_column[1];
|
||||||
|
let z = new_column[2];
|
||||||
|
|
||||||
|
vec![(x / z) as i32, (y / z) as i32]
|
||||||
|
}),
|
||||||
|
);
|
||||||
|
|
||||||
|
println!("Predict:\n{:}", predicted_mat);
|
||||||
|
|
||||||
|
let (k, r, t) = decompose_projection_matrix(&m);
|
||||||
|
|
||||||
|
println!("K:\n{:}\nR:\n{:}\nt:\n{:}", k, r, t);
|
||||||
|
|
||||||
|
let alpha = k[0];
|
||||||
|
let theta = 1.0_f64.atan2(k[3] / -alpha);
|
||||||
|
let beta = k[4] * theta.sin();
|
||||||
|
let u_0 = k[6];
|
||||||
|
let v_0 = k[7];
|
||||||
|
|
||||||
|
println!(
|
||||||
|
"alpha: {:}, theta: {:}, beta: {:}, u_0: {:}, v_0: {:}",
|
||||||
|
alpha, theta, beta, u_0, v_0
|
||||||
|
);
|
||||||
|
|
||||||
|
Ok(m)
|
||||||
|
}
|
||||||
|
|
||||||
|
//Returns (u,v)
|
||||||
|
fn convert_3d_to_2d(m: &ng::Matrix3x4<f64>, point: &Point3d) -> Point2d {
|
||||||
|
let dim_3 = ng::Matrix4x1::new(point.x, point.y, point.z, 1.0);
|
||||||
|
|
||||||
|
let dim_2 = m * dim_3;
|
||||||
|
|
||||||
|
Point2d::new((dim_2[0] / dim_2[2]) as i32, (dim_2[1] / dim_2[2]) as i32)
|
||||||
|
}
|
||||||
|
|
||||||
|
const WINDOW_HEIGHT: usize = 480;
|
||||||
|
const WINDOW_WIDTH: usize = 640;
|
||||||
|
|
||||||
|
fn main() -> AnyResult<()> {
|
||||||
|
//Handles all of the startup init code! Instead of keeping it in the way of the actual image drawing.
|
||||||
|
let m = init()?;
|
||||||
|
let mut window = fb::Window::new(
|
||||||
|
"Display Buffer",
|
||||||
|
WINDOW_WIDTH,
|
||||||
|
WINDOW_HEIGHT,
|
||||||
|
fb::WindowOptions::default(),
|
||||||
|
)?;
|
||||||
|
|
||||||
|
let mut points_3d = Vec::new();
|
||||||
|
|
||||||
|
for i in 0..(10 * 10) {
|
||||||
|
points_3d.push(Point3d::new((i / 10) as f64, (i % 10) as f64, 0.0));
|
||||||
|
}
|
||||||
|
|
||||||
|
for i in 0..(10 * 10) {
|
||||||
|
points_3d.push(Point3d::new((i % 10) as f64, 10.0, (i / 10) as f64));
|
||||||
|
}
|
||||||
|
|
||||||
|
for i in 0..(10 * 10) {
|
||||||
|
points_3d.push(Point3d::new(10.0, (i % 10) as f64, (i / 10) as f64));
|
||||||
|
}
|
||||||
|
|
||||||
|
let points = points_3d
|
||||||
|
.into_iter()
|
||||||
|
.map(|point| convert_3d_to_2d(&m, &point))
|
||||||
|
.collect::<Vec<Point2d>>();
|
||||||
|
|
||||||
|
let mut buffer = vec![0 as u32; WINDOW_WIDTH * WINDOW_HEIGHT];
|
||||||
|
|
||||||
|
update_buffer_with_pixels(&mut buffer, points);
|
||||||
|
|
||||||
|
//Clone this, so we can clone this back into buffer later, for each frame. So we can keep the BG pixels.
|
||||||
|
let background_buffer = buffer.clone();
|
||||||
|
|
||||||
|
window.update_with_buffer(&buffer, WINDOW_WIDTH, WINDOW_HEIGHT)?;
|
||||||
|
window.set_target_fps(30);
|
||||||
|
|
||||||
|
let trb = Point3d::new(1.0, 1.0, 1.0);
|
||||||
|
let trf = Point3d::new(0.0, 1.0, 1.0);
|
||||||
|
let tlb = Point3d::new(1.0, 1.0, 0.0);
|
||||||
|
let tlf = Point3d::new(0.0, 1.0, 0.0);
|
||||||
|
let blf = Point3d::new(0.0, 0.0, 0.0);
|
||||||
|
let blb = Point3d::new(1.0, 0.0, 0.0);
|
||||||
|
let brf = Point3d::new(0.0, 0.0, 1.0);
|
||||||
|
let brb = Point3d::new(1.0, 0.0, 1.0);
|
||||||
|
|
||||||
|
let mut cube = Cube::new(trb, trf, tlb, tlf, blf, blb, brf, brb);
|
||||||
|
|
||||||
|
let mut gif_buffer: Vec<Vec<u8>> = Vec::new();
|
||||||
|
|
||||||
|
let mut x_trans = 0.0;
|
||||||
|
let mut y_trans = 0.0;
|
||||||
|
let mut z_trans = 0.0;
|
||||||
|
|
||||||
|
const X_TRANS_RATE: f64 = 1.0 / 30.0;
|
||||||
|
const Y_TRANS_RATE: f64 = 0.5 / 30.0;
|
||||||
|
const Z_TRANS_RATE: f64 = 2.0 / 30.0;
|
||||||
|
|
||||||
|
const MAX_TRANS: f64 = 6.0;
|
||||||
|
|
||||||
|
while window.is_open() {
|
||||||
|
//Reload the background for each frame
|
||||||
|
let mut buffer = background_buffer.clone();
|
||||||
|
|
||||||
|
cube.translate(Point3d::new(X_TRANS_RATE, Y_TRANS_RATE, Z_TRANS_RATE));
|
||||||
|
|
||||||
|
update_buffer_with_pixels(&mut buffer, cube.render(&m));
|
||||||
|
|
||||||
|
x_trans += X_TRANS_RATE;
|
||||||
|
y_trans += Y_TRANS_RATE;
|
||||||
|
z_trans += Z_TRANS_RATE;
|
||||||
|
|
||||||
|
if x_trans > MAX_TRANS {
|
||||||
|
cube.translate(Point3d::new(-x_trans, 0.0, 0.0));
|
||||||
|
x_trans = 0.0;
|
||||||
|
}
|
||||||
|
if y_trans > MAX_TRANS {
|
||||||
|
cube.translate(Point3d::new(0.0, -y_trans, 0.0));
|
||||||
|
y_trans = 0.0;
|
||||||
|
}
|
||||||
|
if z_trans > MAX_TRANS {
|
||||||
|
cube.translate(Point3d::new(0.0, 0.0, -z_trans));
|
||||||
|
z_trans = 0.0;
|
||||||
|
}
|
||||||
|
|
||||||
|
//Required to close, and update frame;
|
||||||
|
if window.is_key_down(fb::Key::Q) {
|
||||||
|
break;
|
||||||
|
}
|
||||||
|
window.update_with_buffer(&buffer, WINDOW_WIDTH, WINDOW_HEIGHT)?;
|
||||||
|
gif_buffer.push(
|
||||||
|
buffer
|
||||||
|
.iter()
|
||||||
|
.flat_map(|pixel| pixel.to_be_bytes())
|
||||||
|
.collect(),
|
||||||
|
);
|
||||||
|
}
|
||||||
|
|
||||||
|
export_gif(WINDOW_WIDTH, WINDOW_HEIGHT, 30, gif_buffer)?;
|
||||||
|
|
||||||
|
Ok(())
|
||||||
|
}
|
||||||
|
|
||||||
|
///A Cube defined by 8 3d points.
|
||||||
|
#[derive(Debug)]
|
||||||
|
struct Cube {
|
||||||
|
pub trb: Point3d,
|
||||||
|
pub trf: Point3d,
|
||||||
|
pub tlb: Point3d,
|
||||||
|
pub tlf: Point3d,
|
||||||
|
|
||||||
|
pub blf: Point3d,
|
||||||
|
pub blb: Point3d,
|
||||||
|
pub brf: Point3d,
|
||||||
|
pub brb: Point3d,
|
||||||
|
}
|
||||||
|
|
||||||
|
impl Cube {
|
||||||
|
pub fn new(
|
||||||
|
trb: Point3d,
|
||||||
|
trf: Point3d,
|
||||||
|
tlb: Point3d,
|
||||||
|
tlf: Point3d,
|
||||||
|
blf: Point3d,
|
||||||
|
blb: Point3d,
|
||||||
|
brf: Point3d,
|
||||||
|
brb: Point3d,
|
||||||
|
) -> Self {
|
||||||
|
Cube {
|
||||||
|
trb,
|
||||||
|
trf,
|
||||||
|
tlb,
|
||||||
|
tlf,
|
||||||
|
blf,
|
||||||
|
blb,
|
||||||
|
brf,
|
||||||
|
brb,
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
pub fn render(&self, m: &ng::Matrix3x4<f64>) -> Vec<Point2d> {
|
||||||
|
let mut points = Vec::new();
|
||||||
|
|
||||||
|
let trb = self.trb.project(m);
|
||||||
|
let trf = self.trf.project(m);
|
||||||
|
let tlb = self.tlb.project(m);
|
||||||
|
let tlf = self.tlf.project(m);
|
||||||
|
|
||||||
|
let blf = self.blf.project(m);
|
||||||
|
let blb = self.blb.project(m);
|
||||||
|
let brf = self.brf.project(m);
|
||||||
|
let brb = self.brb.project(m);
|
||||||
|
|
||||||
|
points.append(&mut generate_line_between_verticies(&trf, &trb)); // Line 1
|
||||||
|
points.append(&mut generate_line_between_verticies(&trb, &tlb)); // Line 2
|
||||||
|
points.append(&mut generate_line_between_verticies(&tlb, &tlf)); // Line 3
|
||||||
|
points.append(&mut generate_line_between_verticies(&tlf, &trf)); // Line 4
|
||||||
|
points.append(&mut generate_line_between_verticies(&blf, &brf)); // Line 5
|
||||||
|
points.append(&mut generate_line_between_verticies(&brf, &brb)); // Line 6
|
||||||
|
points.append(&mut generate_line_between_verticies(&brb, &blb)); // Line 7
|
||||||
|
points.append(&mut generate_line_between_verticies(&blb, &blf)); // Line 8
|
||||||
|
points.append(&mut generate_line_between_verticies(&blf, &tlf)); // Line 9
|
||||||
|
points.append(&mut generate_line_between_verticies(&blb, &tlb)); // Line 10
|
||||||
|
points.append(&mut generate_line_between_verticies(&brb, &trb)); // Line 11
|
||||||
|
points.append(&mut generate_line_between_verticies(&brf, &trf)); // Line 12
|
||||||
|
|
||||||
|
points
|
||||||
|
}
|
||||||
|
|
||||||
|
///translates the cube by Point amount
|
||||||
|
pub fn translate(&mut self, translate_by: Point3d) {
|
||||||
|
self.trb.translate(&translate_by);
|
||||||
|
self.trf.translate(&translate_by);
|
||||||
|
self.tlb.translate(&translate_by);
|
||||||
|
self.tlf.translate(&translate_by);
|
||||||
|
|
||||||
|
self.blf.translate(&translate_by);
|
||||||
|
self.blb.translate(&translate_by);
|
||||||
|
self.brf.translate(&translate_by);
|
||||||
|
self.brb.translate(&translate_by);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
#[derive(Debug)]
|
||||||
|
struct Point3d {
|
||||||
|
pub x: f64,
|
||||||
|
pub y: f64,
|
||||||
|
pub z: f64,
|
||||||
|
}
|
||||||
|
|
||||||
|
impl Point3d {
|
||||||
|
pub fn new(x: f64, y: f64, z: f64) -> Self {
|
||||||
|
Point3d { x, y, z }
|
||||||
|
}
|
||||||
|
|
||||||
|
pub fn project(&self, m: &ng::Matrix3x4<f64>) -> Point2d {
|
||||||
|
convert_3d_to_2d(m, self)
|
||||||
|
}
|
||||||
|
|
||||||
|
pub fn translate(&mut self, translate_by: &Point3d) {
|
||||||
|
self.x += translate_by.x;
|
||||||
|
self.y += translate_by.y;
|
||||||
|
self.z += translate_by.z;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
#[derive(Debug)]
|
||||||
|
struct Point2d {
|
||||||
|
pub x: i32,
|
||||||
|
pub y: i32,
|
||||||
|
}
|
||||||
|
|
||||||
|
impl Point2d {
|
||||||
|
pub fn new(x: i32, y: i32) -> Self {
|
||||||
|
Self { x, y }
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
fn update_buffer_with_pixels(buf: &mut Vec<u32>, points: Vec<Point2d>) {
|
||||||
|
points.into_iter().for_each(|point| {
|
||||||
|
let index = (point.y as usize) * WINDOW_WIDTH + (point.x as usize);
|
||||||
|
buf[index] = u32::MAX
|
||||||
|
});
|
||||||
|
}
|
||||||
|
|
||||||
|
fn generate_line_between_verticies(point_1: &Point2d, point_2: &Point2d) -> Vec<Point2d> {
|
||||||
|
let mut points = Vec::new();
|
||||||
|
|
||||||
|
let x_diff = point_2.x - point_1.x;
|
||||||
|
let y_diff = point_2.y - point_1.y;
|
||||||
|
|
||||||
|
let x_change = x_diff as f64 / 100.0;
|
||||||
|
let y_change = y_diff as f64 / 100.0;
|
||||||
|
|
||||||
|
for i in 0..100 {
|
||||||
|
points.push(Point2d::new(
|
||||||
|
point_1.x + (i as f64 * x_change) as i32,
|
||||||
|
point_1.y + (i as f64 * y_change) as i32,
|
||||||
|
));
|
||||||
|
}
|
||||||
|
|
||||||
|
points
|
||||||
|
}
|
||||||
|
|
||||||
|
use image::codecs::gif::{GifEncoder, Repeat};
|
||||||
|
use image::{Delay, Frame, RgbaImage};
|
||||||
|
|
||||||
|
fn export_gif(
|
||||||
|
width: usize,
|
||||||
|
height: usize,
|
||||||
|
fps: u32,
|
||||||
|
frames: Vec<Vec<u8>>, // Each inner Vec is width * height * 4 bytes (RGBA)
|
||||||
|
) -> AnyResult<()> {
|
||||||
|
let file = File::create("animation.gif")?;
|
||||||
|
let mut encoder = GifEncoder::new(file);
|
||||||
|
|
||||||
|
// Loop forever
|
||||||
|
encoder.set_repeat(Repeat::Infinite)?;
|
||||||
|
|
||||||
|
for raw_buffer in frames {
|
||||||
|
// Construct the image from your raw pixel slice
|
||||||
|
let img = RgbaImage::from_raw(width as u32, height as u32, raw_buffer).unwrap();
|
||||||
|
|
||||||
|
// Calculate frame timing (e.g., 1000 / 30 = 33.3ms)
|
||||||
|
let delay = Delay::from_numer_denom_ms(1000, fps);
|
||||||
|
let frame = Frame::from_parts(img, 0, 0, delay);
|
||||||
|
|
||||||
|
encoder.encode_frame(frame)?;
|
||||||
|
}
|
||||||
|
|
||||||
|
Ok(())
|
||||||
|
}
|
||||||
+100
@@ -0,0 +1,100 @@
|
|||||||
|
**ENGR4350 Computer Vision**
|
||||||
|
|
||||||
|
**Project 1. Linear Approach to Camera Calibration**
|
||||||
|
|
||||||
|
**Due Data: September 20, 2026**
|
||||||
|
|
||||||
|
**Objective:** To use a linear approach to calibrate a camera and to use
|
||||||
|
the camera parameters to predict the 2D image-domain location of a set
|
||||||
|
of 3D points or a 3D moving object.
|
||||||
|
|
||||||
|
Procedure:
|
||||||
|
|
||||||
|
1. Study the lecture note by showing the use of projection matrix that
|
||||||
|
maps objects from 3-D to 2-D, the definition of intrinsic and
|
||||||
|
extrinsic parameters of a camera, and the linear approach to
|
||||||
|
geometric camera calibration.
|
||||||
|
2. Given an image *test_image.bmp* captured by a camera which is posed
|
||||||
|
toward a 3D chess board, we manually selected 27 points whose 2-D
|
||||||
|
image coordinates are saved in the file *observe.dat*, and whose 3-D
|
||||||
|
coordinates in a world coordinate are stored in the file
|
||||||
|
*model.dat*. Please use the linear approach to calibrate the camera
|
||||||
|
system by computing the projection matrix (M) from which you are
|
||||||
|
required to compute the intrinsic and extrinsic parameters,
|
||||||
|
including , u0, v0, α, β, and the rotation matrix (R), and the
|
||||||
|
shift vector (t).
|
||||||
|
3. In order to verify the correctness of the camera calibration, please
|
||||||
|
use the projection matrix to map the three set of 3D points, whose
|
||||||
|
3-D coordinates are {x,y,0\|x,y=0,...,10}, {x,10,z\|x,z=0,...,10},
|
||||||
|
{10,y,z\|y,z=0,...,10}, respectively, into the 2-D image space.
|
||||||
|
Discuss your results.
|
||||||
|
4. Create a video file that shows a 3D object moving in the 3D scene
|
||||||
|
along a specific pre-defined path. For example, a 3D cube (1x1x1)
|
||||||
|
can be displayed by nine lines connecting seven vertexes (see the
|
||||||
|
example below). Each line can be drawn with around 100 samples. (You
|
||||||
|
can see a video example from Slide 13 of the Lecture 8 handout in
|
||||||
|
the slideshow mode.)
|
||||||
|
|
||||||
|
**Report Requirements:**
|
||||||
|
|
||||||
|
1. Briefly discuss the basics of geometric camera modeling.
|
||||||
|
2. Briefly discuss the linear approach to camera calibration.
|
||||||
|
3. Show the simulation results by figures (Part 3).
|
||||||
|
4. Include some sample frames of the video (Part 4) in the report.
|
||||||
|
5. The Python source code should be included as the appendix of the
|
||||||
|
report with detailed comments. A separate Python file should also be
|
||||||
|
provided for testing.
|
||||||
|
6. Zip all files (DOC, AVI, M-file) into one package and upload it to
|
||||||
|
Blackboard by the due date.
|
||||||
|
|
||||||
|
**Useful Python functions:**
|
||||||
|
|
||||||
|
You need to import the required libraries:
|
||||||
|
|
||||||
|
> import torch\
|
||||||
|
> import numpy as np\
|
||||||
|
> import matplotlib.pyplot as plt\
|
||||||
|
> from PIL import Image\
|
||||||
|
> import time
|
||||||
|
|
||||||
|
> import cv2\
|
||||||
|
> import os
|
||||||
|
|
||||||
|
- Inner product: torch.dot(t1,t2), Cross product : torch.cross(t1,t2).
|
||||||
|
|
||||||
|
> v0 = (s \*\* 2) \* torch.dot(a2, a3)\
|
||||||
|
> \
|
||||||
|
> cross_a1_a3 = torch.cross(a1, a3)
|
||||||
|
|
||||||
|
>
|
||||||
|
|
||||||
|
- Singular value decomposition (SVD) :
|
||||||
|
|
||||||
|
> \# \-\-- Solve for the projection matrix M \-\--\
|
||||||
|
> \# The solution to Qm=0 is the last column of V from the SVD of Q\
|
||||||
|
> \# In PyTorch, torch.linalg.svd returns U, S, Vh (V transpose)\
|
||||||
|
> U, S, Vh = torch.linalg.svd(Q)\
|
||||||
|
> \
|
||||||
|
> \# The solution is the last row of Vh, which corresponds to the
|
||||||
|
> smallest singular value\
|
||||||
|
> m = Vh\[-1, :\]\
|
||||||
|
> \
|
||||||
|
> \# Reshape the 12x1 vector m into the 3x4 projection matrix M\
|
||||||
|
> M = m.reshape(3, 4)
|
||||||
|
|
||||||
|
- Norm of a vector: torch.linalg.norm(vector).
|
||||||
|
|
||||||
|
> torch.linalg.norm(cross_a1_a3)
|
||||||
|
|
||||||
|
>
|
||||||
|
|
||||||
|
- Inverse of a matrix: torch.linalg.inv(M)
|
||||||
|
|
||||||
|
> torch.linalg.inv(M)
|
||||||
|
|
||||||
|
>
|
||||||
|
|
||||||
|
- Reshape of a vector to a matrix:
|
||||||
|
|
||||||
|
> \# Reshape the 12x1 vector m into the 3x4 projection matrix M\
|
||||||
|
> M = m.reshape(3, 4)
|
||||||
LFS
BIN
Binary file not shown.
Reference in New Issue
Block a user