Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames
Artificial intelligence plays a significant role in traffic-accident detection. Traffic accidents involve a cascade of inadvertent events, making traditional detection approaches challenging. For instance, Convolutional Neural Network (CNN)-based approaches cannot analyze temporal relationships amon...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-06-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/11/13/2884 |
| _version_ | 1797591294859542528 |
|---|---|
| author | Yihang Zhang Yunsick Sung |
| author_facet | Yihang Zhang Yunsick Sung |
| author_sort | Yihang Zhang |
| collection | DOAJ |
| description | Artificial intelligence plays a significant role in traffic-accident detection. Traffic accidents involve a cascade of inadvertent events, making traditional detection approaches challenging. For instance, Convolutional Neural Network (CNN)-based approaches cannot analyze temporal relationships among objects, and Recurrent Neural Network (RNN)-based approaches suffer from low processing speeds and cannot detect traffic accidents simultaneously across multiple frames. Furthermore, these networks dismiss background interference in input video frames. This paper proposes a framework that begins by subtracting the background based on You Only Look Once (YOLOv5), which adaptively reduces background interference when detecting objects. Subsequently, the CNN encoder and Transformer decoder are combined into an end-to-end model to extract the spatial and temporal features between different time points, allowing for a parallel analysis between input video frames. The proposed framework was evaluated on the Car Crash Dataset through a series of comparison and ablation experiments. Our framework was benchmarked against three accident-detection models to evaluate its effectiveness, and the proposed framework demonstrated a superior accuracy of approximately 96%. The results of the ablation experiments indicate that when background subtraction was not incorporated into the proposed framework, the values of all evaluation indicators decreased by approximately 3%. |
| first_indexed | 2024-03-11T01:35:22Z |
| format | Article |
| id | doaj.art-95db5638d20145da90673ae681e268b6 |
| institution | Directory Open Access Journal |
| issn | 2227-7390 |
| language | English |
| last_indexed | 2024-03-11T01:35:22Z |
| publishDate | 2023-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj.art-95db5638d20145da90673ae681e268b62023-11-18T17:02:43ZengMDPI AGMathematics2227-73902023-06-011113288410.3390/math11132884Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video FramesYihang Zhang0Yunsick Sung1Department of Autonomous Things Intelligence, Dongguk University-Seoul, Seoul 04620, Republic of KoreaDivision of AI Software Convergence, Dongguk University-Seoul, Seoul 04620, Republic of KoreaArtificial intelligence plays a significant role in traffic-accident detection. Traffic accidents involve a cascade of inadvertent events, making traditional detection approaches challenging. For instance, Convolutional Neural Network (CNN)-based approaches cannot analyze temporal relationships among objects, and Recurrent Neural Network (RNN)-based approaches suffer from low processing speeds and cannot detect traffic accidents simultaneously across multiple frames. Furthermore, these networks dismiss background interference in input video frames. This paper proposes a framework that begins by subtracting the background based on You Only Look Once (YOLOv5), which adaptively reduces background interference when detecting objects. Subsequently, the CNN encoder and Transformer decoder are combined into an end-to-end model to extract the spatial and temporal features between different time points, allowing for a parallel analysis between input video frames. The proposed framework was evaluated on the Car Crash Dataset through a series of comparison and ablation experiments. Our framework was benchmarked against three accident-detection models to evaluate its effectiveness, and the proposed framework demonstrated a superior accuracy of approximately 96%. The results of the ablation experiments indicate that when background subtraction was not incorporated into the proposed framework, the values of all evaluation indicators decreased by approximately 3%.https://www.mdpi.com/2227-7390/11/13/2884artificial intelligencedeep learningtraffic-accident detectionbackground subtractionCNN encoderTransformer decoder |
| spellingShingle | Yihang Zhang Yunsick Sung Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames Mathematics artificial intelligence deep learning traffic-accident detection background subtraction CNN encoder Transformer decoder |
| title | Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames |
| title_full | Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames |
| title_fullStr | Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames |
| title_full_unstemmed | Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames |
| title_short | Traffic Accident Detection Using Background Subtraction and CNN Encoder–Transformer Decoder in Video Frames |
| title_sort | traffic accident detection using background subtraction and cnn encoder transformer decoder in video frames |
| topic | artificial intelligence deep learning traffic-accident detection background subtraction CNN encoder Transformer decoder |
| url | https://www.mdpi.com/2227-7390/11/13/2884 |
| work_keys_str_mv | AT yihangzhang trafficaccidentdetectionusingbackgroundsubtractionandcnnencodertransformerdecoderinvideoframes AT yunsicksung trafficaccidentdetectionusingbackgroundsubtractionandcnnencodertransformerdecoderinvideoframes |