Front Car Lane-Change Detection for Advanced Driver Assistance Systems
Front Car Lane-Change Detection for Advanced Driver Assistance Systems
Front Car Lane-Change Detection for Advanced Driver Assistance Systems
Front Car Lane-Change Detection for Advanced Driver Assistance Systems
Organization:
Southeast University
Organization:
Southeast University
Organization:
Southeast University
Duration:
Aug 2020 - Nov 2020
Duration:
Aug 2020 - Nov 2020
Duration:
Aug 2020 - Nov 2020
Overview
Designed a model for detecting the line-change behavior of a preceding vehicle based on in-vehicle video recordings, attaining ~77% accuracy under severe weather conditions
Collected, cleaned, and labeled in-vehicle video recording data from multiple sources to obtain a comprehensive dataset, covering various camera angels, and weather and illumination conditions
Built a module integrating YOLOv5 for preceding vehicle detection and Deep SORT for vehicle dynamic tracking
Built a lane detection module based on ERFNet and H-Net for balancing accuracy and computational efficiency
Proposed a novel approach for accuracy assessment, which extended the traditional 2-D Intersection-over-Union (IoU) metric to 3-D, by integrating lane crossing time
Overview
Designed a model for detecting the line-change behavior of a preceding vehicle based on in-vehicle video recordings, attaining ~77% accuracy under severe weather conditions
Collected, cleaned, and labeled in-vehicle video recording data from multiple sources to obtain a comprehensive dataset, covering various camera angels, and weather and illumination conditions
Built a module integrating YOLOv5 for preceding vehicle detection and Deep SORT for vehicle dynamic tracking
Built a lane detection module based on ERFNet and H-Net for balancing accuracy and computational efficiency
Proposed a novel approach for accuracy assessment, which extended the traditional 2-D Intersection-over-Union (IoU) metric to 3-D, by integrating lane crossing time
Overview
Designed a model for detecting the line-change behavior of a preceding vehicle based on in-vehicle video recordings, attaining ~77% accuracy under severe weather conditions
Collected, cleaned, and labeled in-vehicle video recording data from multiple sources to obtain a comprehensive dataset, covering various camera angels, and weather and illumination conditions
Built a module integrating YOLOv5 for preceding vehicle detection and Deep SORT for vehicle dynamic tracking
Built a lane detection module based on ERFNet and H-Net for balancing accuracy and computational efficiency
Proposed a novel approach for accuracy assessment, which extended the traditional 2-D Intersection-over-Union (IoU) metric to 3-D, by integrating lane crossing time
Overview
Designed a model for detecting the line-change behavior of a preceding vehicle based on in-vehicle video recordings, attaining ~77% accuracy under severe weather conditions
Collected, cleaned, and labeled in-vehicle video recording data from multiple sources to obtain a comprehensive dataset, covering various camera angels, and weather and illumination conditions
Built a module integrating YOLOv5 for preceding vehicle detection and Deep SORT for vehicle dynamic tracking
Built a lane detection module based on ERFNet and H-Net for balancing accuracy and computational efficiency
Proposed a novel approach for accuracy assessment, which extended the traditional 2-D Intersection-over-Union (IoU) metric to 3-D, by integrating lane crossing time
