Oncoming Vehicle Detection with Variable-Focus Liquid Lens

摘要

Computer vision plays an important role in autonomous vehicle, robotics and manufacturing fields. Depth perception in computer vision requires stereo vision, or fuse together a single camera with other depth sensors such as radar and Lidar. Depth from focus using adjustable lens has not been applied in autonomous vehicle. The goal of this paper is to investigate the application of depth from focus for oncoming vehicle detection. Liquid lens is used to adjust optical power while acquiring images with the camera. The distance of the oncoming vehicle can be estimated by measuring the oncoming vehicle’s sharpness in the images with known lens settings. The results show the system detecting oncoming vehicle at ±2 meter and ±4 meter using depth from focus technique. Estimation of oncoming vehicles above 4 meter can be done by analysing the relative size of the vehicle detected.