In this paper, we used a fisheye lens to acquire the wider rear view camera images. Fisheye lens provided about 170° wide rear view images but the images were distorted because of the characteristics of the lens. In order to compensate this, distorted images were transformed to flat images using a distortion model (FOV). Then, noises of compensated images were removed using filters and the images were converted to grayscale. Grayscale images were then applied to HOG (histogram of gradient) to detect cars approaching from back and the positions of the cars were indicated as colored rectangles. As a future study, we will test our system with harsher environments (such as in darkness, with brightness changes, in different climates) and adjust it to have better recognition rates regardless of environments. We also have a plan to improve our system to evaluate the risk of collision by predicting the velocity and direction of approaching cars.
Wide rear vehicle recognition using a fisheye lens camera image
2016-10-01
408351 byte
Conference paper
Electronic Resource
English
Object detection and classification using a rear in-vehicle fisheye camera
Automotive engineering | 2012
|Object Detection and Classification Using a Rear In-Vehicle Fisheye Camera
Springer Verlag | 2012
|Object Detection and Classification Using a Rear In-Vehicle Fisheye Camera
Tema Archive | 2012
|Rear obstacle detection system with fisheye stereo camera using HCT
Tema Archive | 2015
|Object Detection and Classification Using a Rear In-Vehicle Fisheye Camera F2012-F07-009
British Library Conference Proceedings | 2013
|