Aerial Video Based Traffic Flow Computing
碩士 === 國立交通大學 === 資訊科學與工程研究所 === 107 === Vehicle trajectory at road intersection can provide rich information, including traffic flow, vehicle speed, driving behavior, even the design of road intersection. With these information, road planners can develop more efficient methods to maintain the road,...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2019
|
Online Access: | http://ndltd.ncl.edu.tw/handle/g59y3f |
id |
ndltd-TW-107NCTU5394067 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-107NCTU53940672019-06-27T05:42:50Z http://ndltd.ncl.edu.tw/handle/g59y3f Aerial Video Based Traffic Flow Computing 空拍影像之車流辨識與計算 Wang, Ruyu 王如玉 碩士 國立交通大學 資訊科學與工程研究所 107 Vehicle trajectory at road intersection can provide rich information, including traffic flow, vehicle speed, driving behavior, even the design of road intersection. With these information, road planners can develop more efficient methods to maintain the road, adjust traffic relief measures and even improve the design of road intersection. In this work, we proposed a system based on video from unmanned aerial vehicles (UAVs) and can process image to locate and recognize vehicles such that the trajectories of vehicles can be discovered and traffic information can be calculated. The clear view of road intersection in UAV videos could help road planners to gain more comprehensive traffic information than from traditional surveillance cameras. Then a deep learning framework is applied to recognized vehicles from each frame of the video. The detection model can achieve 93% accuracy. According to the result of the detection, a tracking algorithm which is based on Kalman filter is adopted to reconstruct trajectories of vehicles in each frame. To present the result, a prototype system is implemented. Users can upload videos to our website and virtual gates can be arbitrarily added on the preview of the video. After processing, the video with calculated traffic information such as vehicle trajectory, estimate speed of each vehicle and traffic flow in each direction is allowed to download. Yi, Chih-Wei 易志偉 2019 學位論文 ; thesis 54 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立交通大學 === 資訊科學與工程研究所 === 107 === Vehicle trajectory at road intersection can provide rich information, including traffic flow, vehicle speed, driving behavior, even the design of road intersection. With these information, road planners can develop more efficient methods to maintain the road, adjust traffic relief measures and even improve the design of road intersection.
In this work, we proposed a system based on video from unmanned aerial vehicles (UAVs) and can process image to locate and recognize vehicles such that the trajectories of vehicles can be discovered and traffic information can be calculated. The clear view of road intersection in UAV videos could help road planners to gain more comprehensive traffic information than from traditional surveillance cameras. Then a deep learning framework is applied to recognized vehicles from each frame of the video. The detection model can achieve 93% accuracy. According to the result of the detection, a tracking algorithm which is based on Kalman filter is adopted to reconstruct trajectories of vehicles in each frame. To present the result, a prototype system is implemented. Users can upload videos to our website and virtual gates can be arbitrarily added on the preview of the video. After processing, the video with calculated traffic information such as vehicle trajectory, estimate speed of each vehicle and traffic flow in each direction is allowed to download.
|
author2 |
Yi, Chih-Wei |
author_facet |
Yi, Chih-Wei Wang, Ruyu 王如玉 |
author |
Wang, Ruyu 王如玉 |
spellingShingle |
Wang, Ruyu 王如玉 Aerial Video Based Traffic Flow Computing |
author_sort |
Wang, Ruyu |
title |
Aerial Video Based Traffic Flow Computing |
title_short |
Aerial Video Based Traffic Flow Computing |
title_full |
Aerial Video Based Traffic Flow Computing |
title_fullStr |
Aerial Video Based Traffic Flow Computing |
title_full_unstemmed |
Aerial Video Based Traffic Flow Computing |
title_sort |
aerial video based traffic flow computing |
publishDate |
2019 |
url |
http://ndltd.ncl.edu.tw/handle/g59y3f |
work_keys_str_mv |
AT wangruyu aerialvideobasedtrafficflowcomputing AT wángrúyù aerialvideobasedtrafficflowcomputing AT wangruyu kōngpāiyǐngxiàngzhīchēliúbiànshíyǔjìsuàn AT wángrúyù kōngpāiyǐngxiàngzhīchēliúbiànshíyǔjìsuàn |
_version_ |
1719213402923991040 |