Vehicle Distance Detection and Estimation With Recalibration of a Camera and Lidar System
碩士 === 國立臺北科技大學 === 製造科技研究所 === 106 === Car distance detection is important, no matter in auto-driving or driver assist systems. If we can detect distance between other cars and keep safety distance, we can avoid car accident. Otherwise, when driving it can’t avoid car shaking. Sensors moving make d...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2018
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/v5w769 |
| Summary: | 碩士 === 國立臺北科技大學 === 製造科技研究所 === 106 === Car distance detection is important, no matter in auto-driving or driver assist systems. If we can detect distance between other cars and keep safety distance, we can avoid car accident. Otherwise, when driving it can’t avoid car shaking. Sensors moving make different from initial pose so sensors can’t detect car distance precisely. That is why online calibration is in need. Sensors should be calibrated online after offline calibration to make sure sensors pose anytime. This paper purpose a car distance detection combined with Lidar and camera. In online calibration system, we capture edge feature from Lidar point cloud and camera image. Then align both edge feature to get best calibration result. In car distance detection system, detecting car and motorcycle through using Faster-RCNN object detection model. It can detect point cloud in object detection bounding box so we can get distance for each car and motorcycle. In the end, We collect cars and motorcycles distance data in Taiwan to build a camera-based distance estimation model.
|
|---|