| Paper Abstract and Keywords |
| Presentation |
2022-05-20 10:20
Improvement of Traveling Distance Estimation of a Vehicle from Images Captured by In-Vehicle or CCTV Camera Based on Perspective Projection Model
-- Using Multiple Frames Based on Geometric Constraint --
Daisuke Imoto, Masakatsu Honma, Norimitsu Akiba, Manato Hirabayashi (NRIPS), Shintaro Onozuka (Niigata Prefect. Police Hqrs.), Kimiya Akita (Hokkaido Prefect. Police Hqrs.), Kenji Kurosawa (NRIPS) |
| Abstract |
(in Japanese) |
(See Japanese page) |
| (in English) |
This study focuses on the problem of traveling distance estimation of a straight-ahead vehicle from multiple images captured by a single in-vehicle or security camera at night for forensic science (scientific investigation) applications. Depending on both the image quality and viewing direction, there are problems that the existing method cannot be applied or the analysis accuracy is reduced. In this study, we propose a new method for estimating the vehicle traveling distance by utilizing multiple frames based on a perspective projection model due to geometric constraints. Accuracy evaluation elucidated that when wheel feature points cannot be extracted due to the image quality and viewing direction, the proposed method can be applied in all cases whereas an existing method cannot be applied. Including the cases where feature points can be extracted, the proposed and existing methods are comparable, suggesting the usefulness of the proposed method. |
| Keyword |
(in Japanese) |
(See Japanese page) |
| (in English) |
forensic science / moving distance estimation / perspective projection / DLT / geometric constraint / multiple frames / / |
| Reference Info. |
ITE Tech. Rep. |
| Paper # |
|
| Date of Issue |
|
| ISSN |
|
| Download PDF |
|