Object Distance Measurement System Using Monocular Camera on Vehicle
Research on autonomous vehicles in the next generation has a challenge to improve the safety of the vehicle and surrounding objects through an object distance measurement system. As for some of the components that need to be considered in this research are measuring the distance of the object and wa...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/41919 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Research on autonomous vehicles in the next generation has a challenge to improve the safety of the vehicle and surrounding objects through an object distance measurement system. As for some of the components that need to be considered in this research are measuring the distance of the object and warning the autonomous system to take preventative measures in avoiding accidents. Distance measurement system consists of object detection subsystem and object distance calculation subsystem. The object detection subsystem using MobileNets SSD can detect objects captured by monocular cameras. Distance measurement system with a monocular camera uses three approaches to measure, namely distance calculation using ratios, the calculation method by Jamzad, and an approach with linear regression. Tests carried out in static and moving conditions. Static testing is carried out on flat roads, with the vehicle positioned at a distance of between 5 and 20 m from the camera. Moving testing is carried out to get the relative speed of the detected object based on the calculated distance measurement. Speed information relative to the camera is needed to complete the distance information needed for the vehicle's warning system. The results of this study indicate that the performance of MobileNets is very good for this system, this can be seen from the computation time of 0.084 seconds per frame. Both estimation distance measurements provide distance information between the camera and the detected object with an average error ratio calculation of 6% and a standard deviation value of ± 0.35 meter, for measurement by the Jamzad method an average error value of 9% and a standard deviation value of ± 0.43 meter, while distance measurement uses linear regression with an average error of 1% and a standard deviation of ± 1.65 meters. |
|---|