This paper considers a method for detection of road surface markings using a camera mounted on top of a vehicle.The detection is done with an orientation-aware detector based on a convolutional neural network. To successfully detectthe orientation and position of road surface markings, the input frontal image is converted to a bird’s-eye view imageusing inverse perspective matching. Synthetic image dataset is constructed with aid of MSER (maximally stable extremalregions) algorithm to solve data imbalance problem. The detector is trained to estimate orientations of the detectedobjects in addition to the class labels and positions. Pretrained DenseNet based YOLOv2 model is modified to detectrotated rectangles with an additional cost function and new efficient IOU (intersection of union) measure. Instead ofdirectly estimating the orientation angle of the road surface markings, probabilistic estimation is done with quantizedangular bins. Benchmark dataset is formulated for evaluation and the experimental results showed that the consideredalgorithm provides promising result while running in a real-time.
展开▼