04 / Computer Vision / Robotics
Autonomous Vehicle / Object Detection
An autonomous-vehicle project combining visual object detection with stereo-camera-based distance estimation.
Overview
A computer-vision and robotics project that brings together visual object detection and stereo-camera-based distance estimation for an autonomous vehicle.
Problem
A vehicle perception system needs both information about visible objects and an estimate of how far away they are.
Architecture
Camera observations support two related perception tasks: object detection and stereo-based distance estimation.
- CamerasStereo observations
- PerceptionDetection + stereo vision
- Object / rangePerception information
Implementation
- Use YOLOv5 and OpenCV for visual object-detection work.
- Combine detection with stereo-camera-based distance estimation in a Python / ROS project stack.
Challenges
Camera calibration, scene conditions and consistency between detections and distance estimates are important concerns in this kind of perception system.
Technology
- YOLOv5
- OpenCV
- Python
- Stereo Vision
- ROS
Outcome
A project combining visual detection and distance estimation. Quantitative detection and distance-error measurements are not supplied.
What I learned
A perception result needs more than an object label: range information and the conditions under which it was obtained matter to the wider robotics system.