Person ReIDentification for Detection of Pedestrians in Blind Spots through V2V Communications
| Autor: | Naoko Enami, Kou Asano, Chikara Ohta, Tomio Kamada |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Sight
Artificial neural network business.industry Computer science Blind spot 0202 electrical engineering electronic engineering information engineering 020206 networking & telecommunications 020201 artificial intelligence & image processing Computer vision 02 engineering and technology Pedestrian Artificial intelligence business |
| Zdroj: | ITSC |
| DOI: | 10.1109/itsc.2019.8917140 |
| Popis: | Traffic accidents with pedestrians jumping out from a blind spot are expected to be prevented as much as possible. For this sake, sharing pedestrian information among nearby vehicles is useful for a vehicle to be aware of pedestrians in blind spots. Any pedestrian perceived by other vehicles but unperceived by a certain vehicle itself can be regarded to be basically out of the vehicle’s view or be behind obstacles even if in its sight. In order to confirm this judgment more certainly, in this paper, we propose a classification method using image-based person Re-Identification (ReID) based on deep neural network (DNN) to classify whether a pedestrian is in a blind spot or not based on its corresponding information gathered from nearby vehicles. Further, since there is no suitable dataset to evaluate our method, we newly constructed it, and then evaluated our classification method by using our dataset and the existing dataset. Our experiments showed a possibility that pedestrian ReID can achieve accurate classification thanks to a newly constructed dataset that reflects the resolution and illumination differences among pedestrians in a real environment. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|