Name
Title | ||
|---|---|---|
Evaluating Optical Flow Vectors Under Varying Computer-Generated Snow Intensities and Pixel Density for Autonomous Vehicles |
Abstract | ||
|---|---|---|
The success of Advanced Driver Assistance Systems (ADAS) in self-driving vehicles depends on the accuracy of underlying algorithms used for vision and range-based sensors. In this paper, we present an evaluation model, that measures the performance of optical flow algorithms in noisy conditions and different data processing methods used to determine a mean flow vector of objects. To validate the evaluation model, we run the dense polynomial expansion Farneback [1] algorithm and then perform the evaluation (i) under light, mild and heavy snow intensities, (ii) pixel densities of 0.5, 0.16 and 0.1 pixels/pixel and (iii) data processing methods: Moving Average, Voted Mean and Weighted Mean on Edges. We provide experimental evidence about the quality of robustness of algorithms through Jaccard Index and deviation of displacement vectors from their expected value under different practical criteria that are relevant for the automotive domain. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1109/DSN-W.2018.00071 | 2018 48th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W) |
Keywords | Field | DocType |
opticalflow,ADAS,virtualdataset | Data processing,Pixel density,Computer science,Advanced driver assistance systems,Algorithm,Robustness (computer science),Real-time computing,Jaccard index,Pixel,Optical flow,Moving average | Conference |
ISSN | ISBN | Citations |
2325-6648 | 978-1-5386-6708-8 | 0 |
PageRank | References | Authors |
0.34 | 0 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Vikas Agrawal | 1 | 0 | 0.34 |
| Marcel Frueh | 2 | 0 | 0.34 |
| Oliver Bringmann | 3 | 586 | 71.36 |
| Wolfgang Rosenstiel | 4 | 1462 | 212.32 |