Abstract | ||
---|---|---|
Analyzing video log taken by drive recorder is so important for finding the potential of accidents during work. On the other hand, pre-accident condition analysis for video log is needed to be automated because manual operation takes an enormous amount of time and effort. We proposed a system that we can find the pre-accident situations from the video log taken by the drive recorder on forklift. We focused on locally of motion on each two frames, and taken histogram of dense optical flow which divided as lattice, then we estimated locally of pixel movement to analyze the difference of motion between safe or danger scene. We used that as a feature and try to recognize pre-accident situation using neural network. Our experimental result shows that we could specify pre accident situation as a certain pre accident situation around 65 percent, but we could recognize pre-accident situation as any pre accident situation around 98 percent. It is useful for actual applications, because we show approximate time that pre accident situation was happening in the video recorded by drive recorder in actual application. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/ICAwST.2018.8517229 | 2018 9th International Conference on Awareness Science and Technology (iCAST) |
Keywords | Field | DocType |
Histogram of Optical Flow,locally of Motion,optical Flow,recognition ofHazardous Events | Histogram,Computer vision,Computer science,Situation analysis,Pixel,Artificial intelligence,Optical flow | Conference |
ISSN | ISBN | Citations |
2325-5986 | 978-1-5386-5827-7 | 0 |
PageRank | References | Authors |
0.34 | 0 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shigeki Kobayashi | 1 | 0 | 0.34 |
Yuichi Yaguchi | 2 | 40 | 16.52 |
Keita Nakamura | 3 | 0 | 0.34 |
Keitaro Naruse | 4 | 47 | 19.98 |
Shuichiro Maekawa | 5 | 0 | 0.34 |