Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning
The accurate detection and classification of road markings is required for autonomous vehicles. There have been several previous studies on the detection of road lane markings, but the detection and classification of arrows and bike markings has not received much attention. There exists previous res...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2019-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/8789416/ |
| _version_ | 1818948445829857280 |
|---|---|
| author | Toan Minh Hoang Se Hyun Nam Kang Ryoung Park |
| author_facet | Toan Minh Hoang Se Hyun Nam Kang Ryoung Park |
| author_sort | Toan Minh Hoang |
| collection | DOAJ |
| description | The accurate detection and classification of road markings is required for autonomous vehicles. There have been several previous studies on the detection of road lane markings, but the detection and classification of arrows and bike markings has not received much attention. There exists previous research on the detection and classification of arrows and bike markings, but they comprise a performance limitation owing to the use of the entire input image. Therefore, our approach is focused on enhancing the performance of the detection and classification of arrows and bike markings based on the adaptive region of interest (ROI) and deep convolutional neural network (CNN). In the first stage, a vanishing point is detected in order to create the ROI image. The ROI image that covers the majority of the road region is then used as the input to train the CNN-based detector and classifier in the second stage. The proposed approach is evaluated using three open datasets, namely, the Cambridge dataset, Daimler dataset, and Malaga urban dataset on a desktop computer and NVIDIA Jetson TX2 embedded system. The experimental results show that the proposed method outperforms the state-of-the-art methods in recognition performance even with small road markings at a large distance. |
| first_indexed | 2024-12-20T08:46:55Z |
| format | Article |
| id | doaj.art-959efd9fb4a448b28863db5b7039122c |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-12-20T08:46:55Z |
| publishDate | 2019-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-959efd9fb4a448b28863db5b7039122c2022-12-21T19:46:13ZengIEEEIEEE Access2169-35362019-01-01710981710983210.1109/ACCESS.2019.29335988789416Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep LearningToan Minh Hoang0Se Hyun Nam1Kang Ryoung Park2https://orcid.org/0000-0002-1214-9510Division of Electronics and Electrical Engineering, Dongguk University, Seoul, South KoreaDivision of Electronics and Electrical Engineering, Dongguk University, Seoul, South KoreaDivision of Electronics and Electrical Engineering, Dongguk University, Seoul, South KoreaThe accurate detection and classification of road markings is required for autonomous vehicles. There have been several previous studies on the detection of road lane markings, but the detection and classification of arrows and bike markings has not received much attention. There exists previous research on the detection and classification of arrows and bike markings, but they comprise a performance limitation owing to the use of the entire input image. Therefore, our approach is focused on enhancing the performance of the detection and classification of arrows and bike markings based on the adaptive region of interest (ROI) and deep convolutional neural network (CNN). In the first stage, a vanishing point is detected in order to create the ROI image. The ROI image that covers the majority of the road region is then used as the input to train the CNN-based detector and classifier in the second stage. The proposed approach is evaluated using three open datasets, namely, the Cambridge dataset, Daimler dataset, and Malaga urban dataset on a desktop computer and NVIDIA Jetson TX2 embedded system. The experimental results show that the proposed method outperforms the state-of-the-art methods in recognition performance even with small road markings at a large distance.https://ieeexplore.ieee.org/document/8789416/Detection and recognition of road markingsarrows and bike markingsvanishing pointdeep CNNNVIDIA Jetson TX2 embedded system |
| spellingShingle | Toan Minh Hoang Se Hyun Nam Kang Ryoung Park Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning IEEE Access Detection and recognition of road markings arrows and bike markings vanishing point deep CNN NVIDIA Jetson TX2 embedded system |
| title | Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning |
| title_full | Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning |
| title_fullStr | Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning |
| title_full_unstemmed | Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning |
| title_short | Enhanced Detection and Recognition of Road Markings Based on Adaptive Region of Interest and Deep Learning |
| title_sort | enhanced detection and recognition of road markings based on adaptive region of interest and deep learning |
| topic | Detection and recognition of road markings arrows and bike markings vanishing point deep CNN NVIDIA Jetson TX2 embedded system |
| url | https://ieeexplore.ieee.org/document/8789416/ |
| work_keys_str_mv | AT toanminhhoang enhanceddetectionandrecognitionofroadmarkingsbasedonadaptiveregionofinterestanddeeplearning AT sehyunnam enhanceddetectionandrecognitionofroadmarkingsbasedonadaptiveregionofinterestanddeeplearning AT kangryoungpark enhanceddetectionandrecognitionofroadmarkingsbasedonadaptiveregionofinterestanddeeplearning |