Upload
elmer-wheeler
View
216
Download
0
Embed Size (px)
Citation preview
A Comparison of Image Aligning and Correcting Software with an Unmanned Aerial System
Kangsan Lee, Jinwoo Park, Jinmu Choi
1
Contents
• Introduction
• Data Acquisition
– Image Gathering
– Correcting Lens Distortion
• Image Alignment
– Image Alignment Software
– Comparing Image Stitching Software
– Results
• Conclusion
Spatial Analysis & GIS Lab 2
Introduction
• Why Unmanned Aerial System: Drone Aircrafts & Satellites: Major role in Remote Sensing Structural problems: Time, Cost and Weather condition UAS, Drone...?
• Purpose of study: Comparing image stitching softwares Variety of softwares in the market Each software was developed ‘On their own needs’
Spatial Analysis & GIS Lab 3
Q. What is the suitable program for small UAS system?
Data Acquisition
• Image Gathering– Hardware: 3DR Robotics IRIS+(Drone) and GoPro(Camera) are used– Area of interest: Baramarae beach - west coast of South Korea– App. 200 images are collected and used (Flight date: Jan, 2015)
Spatial Analysis & GIS Lab 4
Figure 1. IRIS+ and GoPro (Right), Mission Planner for autonomous flight (Left)
Data Acquisition
• Correcting Lens Distortion– GoPro(small digital camera) has a lens distortion (Tauro et al., 2014)
– Adobe Lightroom is selected to remove lens distortion
Spatial Analysis & GIS Lab 5
Figure 2. Lens distortion correction (before – later)
Image Alignment
• Experimental Design
Spatial Analysis & GIS Lab 6
Component Specification
CPU Intel Xeon CPU E3-1231 3.40GHz
RAM 16GB (DDR3, Dual Channel)
GPU GeForce GTX 750Ti
Cuda Core 640
Storage SSD, 256GB
Table 1. Specification of the testing environment
Computing EnvironmentProcessing
TimeTotal Image
Volume
Image Qual-ity
Image Alignment
• Image Alignment Software– Five major software are selected to test about stitching performance– Open-Source software(GPL); Hugin– Commercial software; Photoscan, Photoshop, MS ICE, Pix4D
Spatial Analysis & GIS Lab 7
Name Developer License Notes
Photoscan Agisoft Proprietary Using photometry method
Photoshop Adobe Proprietary User-friendly interface
Image CompositeEditor
Microsoft Proprietary(Free for non-commercial use)
Variety of pre-defined projection
Hugin Pablo d'Angelo et al.
GPL Based on panorama tools
Pix4Dmapper Pix4D SA Proprietary Optimized for UAS
Table 2. Software information
Image Alignment
• Comparing image stitching software– The amount of input data has been changed; range from 50 to 200
Spatial Analysis & GIS Lab 8
Figure 3. Concept of software comparing experiment
Image Alignment
Results
Spatial Analysis & GIS Lab 9
1) Processing Time• Gradually increased• MS ICE → Fastest• Hugin → X working: Software limitation• Photoshop → X working 150↑: Hardware limitation
Image Alignment
2) Image Volume• Image volume gradually increased according to numbers of images• MS ICE exports the lowest volume• Photoscan shows the highest rate of increase
Spatial Analysis & GIS Lab 10
Image Alignment
3) Image Quality
• Photoscan shows the best quality without any problem
• MS ICE and Photoshop have some distorted
• Pix4D cannot calibrate colour and image matching
Spatial Analysis & GIS Lab 11
Conclusion
• Limitation– Researchers can easily find alternative Open-Source software– But the performance of open-source software does not reach to com-
mercial one
• Potential– The potential of open-source software is sufficient– E.g. Visual SFM became the best 3D image reconstruction open-source
software
• Future Market– The sales market of UAS(Drone) have been increased– Needs of image stitching will be increased in the near future
Spatial Analysis & GIS Lab 12
Thank you