In this report we present the methodology we are using in the AMI group to
compare optic flow estimation methods in the context of the FLUID Specific
Targeted Research Project - Contract No 513633 founded by the CEE. The main
goal of this report is to try to unify comparison criteria between the different
parnerts in order to share information and conclusions. We think that it is one
of the topics we have to address in Las Palmas Meeting December next. In
order to describe the methodology we use, we divide it in five steps:
1. We choose a number of optic flow estimation methods
2. We choose an image sequence dataset to apply the methods.
3. We apply the methods and we store the results in a ASCII file using an
standard format
4. We define a number of statistics to compare the methods (we use the
ASCII files as basic information to compare).
5. We compute the statistics for each pair of methods and we generate tables
with the information. To help result interpretation we use arrow images
and in the case we know the optic flow groundtruth we generate image to
provide the information about what method perform better in the different
areas of the image.
To illustrate this approach and to fit ideas we present the results of this
approach in a particular image sequence provided by LaVision in sections from
1.1 to 1.4. Then, in section 2 we remark some of the difficulties we have found
in our work, which 首ld be discussed in order to be able to compare results
provided by all the partners of the FLUID project. 自適應粒子成像測速場儀（PIV） 德國LaVision 表面PIV 德國LaVision PIV/PLIF粒子成像測速場儀 汽車發動機多參量測試系統 顯微粒子成像測速系統(Micro PIV) 雙平面立體PIV 時間分辨粒子成像測速系統(TR-PIV) 內窺式粒子成像測速系統(Endo-PIV) 水下粒子成像測速系統(Under Water PIV) 電動層析顯微粒子成像測速PIV-Mitas 三維立體PIV 體視層析粒子成像測速系統(Tomo-PIV)
在線留言
滬公網安備 31011502008050號
參與評論
登錄后參與評論