Download presentation
Presentation is loading. Please wait.
Published byJessie Ramsey Modified over 8 years ago
1
Panorama Tools H2T2 Group Capstone Project
2
H2T2 Group TungNS00457 - Project Manager HoaHM00556 - DesignerHuongP00282 - DeveloperThoND00288 - Tester
3
Content I.Overview II.PMS Project 1.Requirements 2.Software Process Model 3.Architecture Design 4.Algorithm 5.Test III.Demo IV.Q&A
4
Overview What is Panorama?
5
Overview How to make a Panorama?
6
Some type of panoramic images Sphere Cube Planar of Flat Cylinder
7
Some existing methods and solutions Kolor Autopano
8
Some existing methods and solutions Microsoft’s Image Composite Editor
9
Our idea Free and open source software. High quality. Powerful.
10
PMS Project
11
Requirement User Requirement Specification Functional RequirementNon-functional Requirement Create panorama automatically Create PMS file project Display image Edit image Friendly interface Easy to use
12
Requirement System Requirement Specification Hardware InterfacesSoftware Interfaces 2 GHz processor 32-bit (x86) 2 GB RAM or higher. 128 MB Graphic Card or higher. Microsoft Visual Studio 2010 OpenCV 2.2 Library.NET Framework 4.0
13
Requirement o System features Use case 1
14
Requirement o System features Use case 2
15
Requirement o System features Use case 3
16
Requirement o System features Use case 4
17
Requirement o System features Use case 5
18
Requirement o System features Use case 6
19
Software Process Model WHY CHOOSE? OUR CHOSE PMS team members experience PMS project characteristic Iterative and incremental development
20
Architecture Design Choice of System Architecture The basic of system architecture to build the application “Panorama Tool” Application WPF NET Framework 4.0 OpenCV2.2 GUI Core-PMS.dll
21
Architecture Design Component Diagram
22
Architecture Design Core Package GUI Package
23
Architecture Design Sequence Diagram
24
Architecture Design User Interface Design
25
Architecture Design Data Structure: *.PMS file
26
Algorithm Image Stitching algorithm flow: Reference: [1] Jubiao Li and Junping Du Study on Panoramic Image Stitching Algorithm, 2010 PACCS
27
Feature Extraction: Harris Corner Detection, SIFT, SUFT, etc Feature Matching: Neighbor Matching, SIFT descriptors, SUFT descriptor, etc Mismatch Removal & Image Registration: RANSAC Image Fusion: Using result of Image Registration to stitch images Algorithm
28
Algorithm Feature Extraction: Harris Corner Detection Simple example with function E() = Sum(all pixel in small window) Window around flat: E do not change Window around edge: E change in some directions, do not change along edge Window around corner: E change in all directions
29
Algorithm Feature Extraction: Harris Corner Detection
30
Algorithm window size = 3, threshold = 0.1 Reference: [2] C. Harris and M.J. Stephens. A combined corner and edge detector. In Alvey Vision Conference, pages 147–152, 1988.
31
Algorithm Feature Matching: Neighbor Matching Area to compare two features from two images Distance(X,Y) = SUM (Xi * Yi) / SQRT (Yi * Yi) window size = 51 pixel; adaptive threshold
32
Algorithm Mismatch Removal & Image Registration: RANSAC RANSAC is an abbreviation for "RANdom SAmple Consensus" Reference: [3] Ondrej Chum (2005) - "Two-View Geometry Estimation by Random Sample and Consensus"
33
Algorithm Mismatch Removal & Image Registration: RANSAC We consider a couple matching key features from two image is one point in previous sample of RANSAC, we have to find the model to fit the maximum number of coupe matching key features. Model to use RANSAC: Reference: [1] Jubiao Li and Junping Du Study on Panoramic Image Stitching Algorithm, 2010 PACCS
34
Algorithm Mismatch Removal & Image Registration: RANSAC Sample result of using RANSAC:
35
Algorithm Image Fusion: Using result of Image Registration (matrix M) to stitch images The first imageThe second image
36
Algorithm Image Fusion: Using result of Image Registration (matrix M) to stitch images Result with no blending Result with blending
37
Test The V-Model
38
Test Test Approach Unit testing Integration testing System testing Acceptance testing
39
Test Test cases PCL (Program Check List) test cases Why PCL? Ensure quality of application. Easy to detect defects and issues. Reduce effort.
40
Test Defect logs tracking system
41
Result CasesTime 02 images (1280x800)~ 1.5s 03 images (1280x800)~ 3.5s 06 images (1280x800)~ 10s 08 images (1280x800)~ 15s
43
Compare with Autopano Giga 2.5
44
Assessment
45
Q&A
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.