Abstract | ||
---|---|---|
With the increasing population of drivers and the popularity of intelligent cars, traffic safety is becoming more concerned. Aiming to reduce the probability of road accidents, many works have explored how to provide an early warning when the distances between the current car (or the in-car camera) and vehicles ahead are dangerous. In this paper, benefiting from the robust pixel-level understanding of cutting-edge deep networks, we introduce a vehicle segmentation method based on monocular vision, detecting vehicles in front of the current one and generating a segmentation mask for each vehicle instance detected. Instead of concentrating our efforts on measuring distance precisely, we categorize distances into three classes (“near”, “medium”, and “far”), which are pretty enough for an early collision warning. Our model is trained and evaluated on our hand-made dataset, after which we transfer and deploy it onto an Android TV virtual machine to test its performance. All the results we got show that our method does a good job in vehicle segmentation with coarse distance estimation, being applicable to aided driving. |
Year | DOI | Venue |
---|---|---|
2021 | 10.1109/ICEBE52470.2021.00019 | 2021 IEEE International Conference on e-Business Engineering (ICEBE) |
Keywords | DocType | ISBN |
vehicle segmentation,distance estimation,monocular vision | Conference | 978-1-6654-4419-4 |
Citations | PageRank | References |
0 | 0.34 | 0 |
Authors | ||
7 |
Name | Order | Citations | PageRank |
---|---|---|---|
Fangyuan Xie | 1 | 0 | 0.34 |
Pengshuai Yin | 2 | 0 | 0.34 |
Zehui Ke | 3 | 0 | 0.34 |
Ruizhou Sun | 4 | 0 | 0.34 |
Ke Ding | 5 | 0 | 0.34 |
Jian Chen | 6 | 42 | 8.66 |
Wu Qingyao | 7 | 259 | 33.46 |