Geometric Algebra. Prof. Nigel Boston. Camera Network Research Group Meeting. Nov. 8 & 15, 2007

Similar documents
Density estimation III

EE650R: Reliability Physics of Nanoelectronic Devices Lecture 5: Statistical Projection Date:

Normal Random Variable and its discriminant functions

Gene Expression Data Analysis (II) statistical issues in spotted arrays

Supplemental notes for topic 9: April 4, 6

Module 1: Time Value of Money and Valuation of Cash Flows

EXERCISE - BINOMIAL THEOREM

STATIC GAMES OF INCOMPLETE INFORMATION

Full file at CHAPTER 3

Optimal Reliability Allocation

A Test of Normality. Textbook Reference: Chapter 14.2 (eighth edition, pages 591 3; seventh edition, pages 624 6).

FITTING EXPONENTIAL MODELS TO DATA Supplement to Unit 9C MATH Q(t) = Q 0 (1 + r) t. Q(t) = Q 0 a t,

A Comparison of the Nonparametric Regression Models using Smoothing Spline and Kernel Regression

(SPOT) LIBOR (RATE), denoted L ( t, strictly denote the continuouslycompounding. SHORT (INSTANTANEOUS SPOT) RATE, denoted r (t)

Lecture 9 February 21

1036: Probability & Statistics

Chapter 4. More Interest Formulas

IEOR 130 Methods of Manufacturing Improvement Fall, 2017 Prof. Leachman Solutions to First Homework Assignment

CHAPTER - IV STANDARDIZED CUSUM MEDIAN CONTROL CHART

arxiv: v2 [math.pr] 10 Nov 2015

Task 879.1: Intelligent Demand Aggregation and Forecasting

r i = a i + b i f b i = Cov[r i, f] The only parameters to be estimated for this model are a i 's, b i 's, σe 2 i

MOMENTS EQUALITIES FOR NONNEGATIVE INTEGER-VALUED RANDOM VARIABLES

Deriving & Understanding the Variance Formulas

Probability and Statistical Methods. Chapter 8 Fundamental Sampling Distributions

Probability and Statistical Methods. Chapter 8 Fundamental Sampling Distributions

Financial Econometrics Jeffrey R. Russell Midterm Winter 2011

Economic Growth Continued: From Solow to Ramsey

Sorting. Data Structures LECTURE 4. Comparison-based sorting. Sorting algorithms. Quick-Sort. Example (1) Pivot

Analytical Method for Solution Two-Dimensional Partial Differential Equations with Boundary Integral Conditions

San Francisco State University ECON 560 Summer 2018 Problem set 3 Due Monday, July 23

LECTURE 5: Quadratic classifiers

Thespis No. 10. f mz mz nz n k s k s k s k s k s k s k s k s k s k s k s k s k s k s k s k s k s k s k s. k s o o

Types of Sampling Plans. Types of Sampling Plans. Sampling Procedures. Probability Samples -Simple Random sample -Stratified sample -Cluster sample

Chapter 4. More Interest Formulas

Solution to Tutorial 6

Estimation of the Parameters of the Exact Extreme Value Distribution

GAUTENG DEPARTMENT OF EDUCATION SENIOR SECONDARY INTERVENTION PROGRAMME MATHEMATICS GRADE 12 SESSION 3 (LEARNER NOTES)

Integrating Customer Value Considerations into Predictive Modeling

Performance Analysis of Divide and Conquer Algorithms for the WHT

4452 Mathematical Modeling Lecture 17: Modeling of Data: Linear Regression

Mathematics 1307 Sample Placement Examination

Bond Prices and Interest Rates

Parametric Density Estimation: Maximum Likelihood Estimation

A Note on Missing Data Effects on the Hausman (1978) Simultaneity Test:

Aggregate Demand Aggregate Supply 1 Y. f P

MA Lesson 11 Section 1.3. Solving Applied Problems with Linear Equations of one Variable

4. Methodology. 4.1 Event study methodology

- Inferential: methods using sample results to infer conclusions about a larger pop n.

Market and Information Economics

ON MAXIMAL IDEAL OF SKEW POLYNOMIAL RINGS OVER A DEDEKIND DOMAIN

Fitting Least squares

Exam 1. Econ520. Spring 2017

Performance Analysis of Divide and Conquer Algorithms for the WHT

Interest Rate Derivatives: More Advanced Models. Chapter 24. The Two-Factor Hull-White Model (Equation 24.1, page 571) Analytic Results

Consult the following resources to familiarize yourself with the issues involved in conducting surveys:

Transmuted Inverse Rayleigh Distribution: A Generalization of the Inverse Rayleigh Distribution.

MA Advanced Macro, 2016 (Karl Whelan) 1

Michał Kolupa, Zbigniew Śleszyński SOME REMARKS ON COINCIDENCE OF AN ECONOMETRIC MODEL

8.0% E(R) 6.0% Lend. Borrow 4.0% 2.0% rf rf 0.0% 0.0% 1.0% 2.0% 3.0% 4.0% STD(R) E(R) Long A and Short B. Long A and Long B. Short A and Long B

An alternative estimator for estimating the finite population mean in presence of measurement errors Rajesh Singh, Sachin Malik

Problem Set 6 Finance 1,

Speeding up k-means Clustering by Bootstrap Averaging

Stylized fact: high cyclical correlation of monetary aggregates and output

A New Nonlinear Fuzzy Robust PCA Algorithm and Similarity Classifier in Classification of Medical Data Sets

Lecture 9: The law of large numbers and central limit theorem

Using Data Mining for Learning and Clustering FCM

UCLA Department of Economics Fall PhD. Qualifying Exam in Macroeconomic Theory

Poverty indices. P(k;z; α ) = P(k;z; α ) /(z) α. If you wish to compute the FGT index of poverty, follow these steps:

Econ 546 Lecture 4. The Basic New Keynesian Model Michael Devereux January 2011

Data Mining Anomaly Detection. Lecture Notes for Chapter 10. Introduction to Data Mining

Data Mining Anomaly Detection. Lecture Notes for Chapter 10. Introduction to Data Mining

Estimating the Common Mean of k Normal Populations with Known Variance

5. Best Unbiased Estimators

Macroeconomics. Part 3 Macroeconomics of Financial Markets. Lecture 8 Investment: basic concepts

Sample Survey Design

ANSWER ALL QUESTIONS. CHAPTERS 6-9; (Blanchard)

= 1. UCLA STAT 13 Introduction to Statistical Methods for the Life and Health Sciences. Parameters and Statistics. Measures of Centrality

0.07 (12) i 1 1 (12) 12n. *Note that N is always the number of payments, not necessarily the number of years. Also, for

The Change Effect for Boundary Conditions to Minimize Error Bound in Lacunary Interpolation by Forth Spline Function

Effective March 16, 2015

4.5 Generalized likelihood ratio test

Lecture 5 Point Es/mator and Sampling Distribu/on

FINANCIAL MATHEMATICS GRADE 11

A New Constructive Proof of Graham's Theorem and More New Classes of Functionally Complete Functions

STAT 135 Solutions to Homework 3: 30 points

Data Aggregation with Sequential Detection in Wireless Sensors Network

Affine Term Structure Pricing with Bond Supply As Factors

Forecasting with Judgment

Confidence Intervals. CI for a population mean (σ is known and n > 30 or the variable is normally distributed in the.

7 pages 1. Hull and White Generalized model. Ismail Laachir. March 1, Model Presentation 1

CAUCHY'S FORMULA AND EIGENVAULES (PRINCIPAL STRESSES) IN 3-D

11.7 (TAYLOR SERIES) NAME: SOLUTIONS 31 July 2018

Noise and Expected Return in Chinese A-share Stock Market. By Chong QIAN Chien-Ting LIN

1. (S09T3) John must pay Kristen 10,000 at the end of 1 year. He also must pay Ahmad 30,000 at the end of year 2.

Empirical analysis on China money multiplier

EVA NOPAT Capital charges ( = WACC * Invested Capital) = EVA [1 P] each

Problem Set 1 Answers. a. The computer is a final good produced and sold in Hence, 2006 GDP increases by $2,000.

Correlation of default

Mortality in the Nordic countries Sweden. IAA Life 2007, Stockholm Ellinor Forslund

Transcription:

Geomerc Algebra Ngel Boso /5/07 Geomerc Algebra Prof. Ngel Boso Camera Nework esearch Group Meeg Nov. 8 & 5, 007 Sraegy: Use Clfford algebra o develop varas for projecve rasformaos. (eferece: J. Laseby, e al., New Geomerc Mehods for Compuer Vso: a applcao o srucure ad moo esmao, 996) Idea: Bass-free geomery (sce s bass-free, varas wll aurally arse, herefore should be useful for our purposes.) Gve: vecors, a, b a b (grade 0) a x b (grade ) (oly vald for 3D) More geerally, roduce a /\ b (a wedge b) s a dreced area sweepg from a o b.(see Fg. ) (a parallelogram) (hs s a bvecor) Ad, b /\ a - a /\ b () I geeral, we have a /\ a /\ /\ a m- /\ a m (mul-vecor) (a parallelopped) (grade m) Defe: ab a b + a /\ b () follows ha: ab a /\ b f a, b are orhogoal, ad (3) so he, ab a b f a, b are parallel (4) ba b a + b /\ a a b - a /\ b (5) of 7

Geomerc Algebra Ngel Boso /5/07 a b ½ (ab + ba) (assocaed er produc) (6) a /\ b ½ (ab - ba) (assocaed ouer produc) (7) Defo: A geomerc algebra has a produc sasfyg: (ab)c a(bc) aλ λa, for scalar λ a(b + c) ab + ac a a (b + c)a ba + ca ad he assocaed er ad ouer produc are defed by (6) ad (7). If {σ,, σ } s ay orhoormal bass for, we ge a bass for he algebra: {, {σ }, {σ /\ σ j j}, {σ /\ σ j /\ σ k #{, j, k} 3}, ec.} whch wll have elemes. The hghes grade eleme s σ /\ /\ σ. Example: 3 The bass wll have 3 8 elemes: {, σ, σ, σ 3, σ σ, σ σ 3, σ σ 3, σ σ σ 3 } (Noe: σ σ 0 sce he bass s orhoormal, hece σ σ σ /\ σ.) (Also, sce he bass s orhoormal, σ σ.) 8 (Ths s wh a srage mulplcao.) I follows ha: (σ σ σ 3 ) σ σ σ 3 σ σ σ 3 -σ σ σ σ 3 σ σ 3 -σ σ σ σ 3 -σ σ σ 3 - Also, oe ca verfy ha σ σ σ 3 commues wh every eleme of he algebra (wll be rue for odd). Call σ σ σ 3 (.e., ) I follows ha: σ 3 σ σ, σ σ σ 3, σ σ 3 σ (8) These bvecors roae vecors her ow plae by 90. Noe: se I σ, J - σ, K σ 3, he we ge: I J K IJK - (Hamlo s quaeros). The algebra wh bass, I, J, K s of our geomerc algebra. 4 wh a srage mulplcao,.e., a sub-algebra of 7

Geomerc Algebra Ngel Boso /5/07 Geeral oaos: Frs, sudy reflecos. (easer) (Ay roao s he produc of wo reflecos.) Cosder a vecor, a refleced he plae orhogoal o he u vecor, o ge a (see Fg. ). The, we ca breakup a o compoes, a a +, (9) a he compoes perpedcular ad parallel o, ad he, a a. (0) a Sce s a u vecor, he, ad a a ( a + /\ a) ( a) + ( /\ a), whch mples ha a a ( /\ a) - ( a) a -( a) - ( /\ a) -( a + /\ a) -a. () Thus, reflecg he plae perpedcular o s he map (Noe: hs works ay dmeso) a a a Suppose you reflec he plae perpedcular o, ad he he plae perpedcular o m. (he produc of wo reflecos) Ths s a roao. The, a maps o -m(-a)m (m)a(m) a, where m, m. () s called a roor ( s also a mulvecor) ( ecapsulaes he formao abou he roao.) Noes: ) oly has eve grade elemes (scalar, bvecor, ec.) ad mm m m mm, ad ( s he mulplcave verse of ) ) aa a hadles roaos ay dmeso. 3) Ca roae elemes of ay grade, o jus vecors very geeral! 3 of 7

Geomerc Algebra Ngel Boso /5/07 Cosder he problem of roag o, say by agle θ (see Fg. 3). Wha s? Noe ha oe soluo s +, bu we also eed, so ry ( + ), he ( + )( + ) ( + + + ) ( + ) ( + ) So, + θ exp, (3) ( + ) where s orhogoal o he plae cu ou by ad (he axs of roao). /5/07 Example: Camera moo from wo scee projecos wh rage daa kow. (3D-o-3D correspodeces) Assume cameras wh opcal ceers a O ad O, wh respecve Axes {σ, σ, σ 3 } ad { σ, σ, σ 3 }, where σ 3 s orhogoal o, he mage plae for camera oe, ad smlarly for camera wo. (See Fg. 4.) Le O P, x O M, O O (4) The frame {σ, σ, σ 3 } s roaed o a frame { σ, σ σ } correspodg roor, we have a O, where for beg he, 3 σ σ σ σ (5) Le ( P O The, ( (6) 4 of 7

Geomerc Algebra Ngel Boso /5/07 5 of 7 The observables (measuremes) he {σ } ad { } σ frames are: σ ad σ ( Defe he vecor he {σ }frame as: ( ) ( ) ( ) ( σ σ σ (7) earragg: + (8) For smplcy, assume. Suppose we have po correspodeces he wo vews, ad he coordaes he vews, { } ad { }, ( ), are kow. We wa o recover he camera moo,.e., fd ad ha mmze he sum: ( ) [ ] S (9) To fd hs mmzao, we dffereae wr ad, he se o zero. Frs, wr : ( ) [ ] ( ) S (0) whch wll be zero whe, ( ) [ ] 0 Solve for : [ ] 0 + [ ] [ ] ()

Geomerc Algebra Ngel Boso /5/07 where ad () So, he opmal s he cerods of he pos. (Noe: hs s a kow resul ha we have recovered here.) (Noe: here wll be a ssue wh robusess sce oe ouler ca adversely affec he resuls a so-called black swa ) The, dffereag wr, ca be show ha we ge: [ ( ) ] 0 (3) Plug- he opmal ad we ge: [ u ] v 0 (4) where u ad v We ca fd he roor,, usg he Sgular Value Decomposo (SVD) o he marx F, defed erms of he u s ad v s as: F β σ f ( σ ) ( σ u )( σ v ) β β (5) he he SVD gves F USV T, ad he VU T. Paper for ex me (Ma): A geomerc approach for he heory ad applcaos of 3D projecve varas Bayro-Corrochao, Eduardo; Baarer, Vladmr Joural of Mahemacal Imagg ad Vso, v. 6,., March 00, pp. 3-54 6 of 7

Geomerc Algebra Ngel Boso /5/07 Appedx: Fg. - a /\ b Fg. Vecor a refleced he plae orhogoal o. Fg. 3 oao of o. Fg. 4 Objec po vewed from wo camera posos. 7 of 7