Higher Accuracy and Lower Computational Perception Environment Based Upon a Real-time Dynamic Region of Interest
- Delivery
- Provide download link
- Format
- Price
- Non-members (tax incl.)：￥6,600 Members (tax incl.)：￥5,280
- Paper/Info type
- SAE Paper
No.2022-01-0078
- Pages
- 1-9(Total 9 p)
- Date of publication
- Mar 2022
- Publisher
- SAE International
- Language
- English
- Event
- WCX SAE World Congress Experience 2022
Detailed Information
| Author(E) | 1) Nicolas Eric Brown, 2) Johan Fanas Rojas, 3) Hamzeh Alzu'bi, 4) Qusay Alrousan, 5) Rick Meyer, 6) Zachary Asher |
|---|---|
| Affiliation(E) | 1) Western Michigan University, 2) Western Michigan University, 3) FEV North America, Inc., 4) FEV North America, Inc., 5) Western Michigan University, 6) Western Michigan University |
| Abstract(E) | Robust sensor fusion is a key technology for enabling the safe operation of automated vehicles. Sensor fusion typically utilizes inputs of cameras, radars, lidar, inertial measurement unit, and global navigation satellite systems, process them, and then output object detection or positioning data. This paper will focus on sensor fusion between the camera, radar, and vehicle wheel speed sensors which is a critical need for near-term realization of sensor fusion benefits. The camera is an off-the-shelf computer vision product from MobilEye and the radar is a Delphi/Aptive electronically scanning radar (ESR) both of which are connected to a drive-by-wire capable vehicle platform. We utilize the MobilEye and wheel speed sensors to create a dynamic region of interest (DROI) of the drivable region that changes as the vehicle moves through the environment. The use of the DROI can reduce the need for up to approximately 100% of the detections from radar, for processing of the driveable region. This provides not only accurate and robust detections but also has benefits in lowering computational power and time complexity. We then continue to reduce the number of detections in the driveable region using machine learning techniques such as density-based spatial clustering and applications with noise (DBSCAN). This is followed by KMeans clustering to further reduce detections and lastly fused with the extended Kalman filter. Our experimental results obtained using an instrumented vehicle show a large reduction in the need for radar detections processing after both, fusion with our DROI and further clustering using machine learning techniques for the driveable region. Our proposed complete technique decreases the amount of fused misdetection, decreases computational power, and increases the reliability of the fused perception model which can greatly benefit current Advanced Driver Assistance System products available on the market today. |
