Robust detection of shady and highlighted roads for monocular camera based navigation of UGV

Abstract

Abstract-This paper addresses the problem of UGV navigation in various environments and lightning conditions. Previous approaches use a combination of different sensors, or work well, only in scenarios with noticeable road marking or borders. Our robot is used for chemical, nuclear and biological contamination measurement. Thus, to avoid complications with decontamination, only a monocular camera serves as a sensor since it is already equipped. In this paper, we propose a novel approach -a fusion of frequency based vanishing point estimation and probabilistically based color segmentation. Detection of a vanishing point, is based on the estimation of a texture flow, produced by a bank of Gabor wavelets and a voting function. Next, the vanishing point defines the training area, which is used for self-supervised learning of color models. Finally, road patches are selected by measuring of the roadness score. A few rules deal with dark cast shadows, overexposed highlights and adaptivity speed. In addition to the robustness of our system, it is easy-to-use since no calibration is needed.

Rapid vanishing point estimation for general road detection

“…The data set consists of 1003 images, among them, about 430 images are from photographs taken on a scouting trip along a possible Grand Challenge route in the Southern California desert and the other part is downloaded from internet by Google Image. All images are normalized to the same height and width of 128, as is suggested in [14]. The ground truth data were obtained by manual labeling: 5 persons marked the vanishing point location, a median of these results is used as the initial ground-truth position.…”

“…More recently, Miksik et al [14] proposed a fusion of the frequency based estimation of a vanishing point and probabilistically based texture segmentation. The texture segmentation utilizes the continuously re-learned Gaussian Mixture Model, which is used to determine the free road ahead of the robot.…”

Visual road following using intrinsic images

“…Their vanishing point detection closely follows the method presented in [12], which fuses the responses of Gabor filters applied on the image by means of particle filtering, and the 'roadness' of the image pixels is classified by means of Gaussian Mixtures. The authors of [11] tested their algorithm only on datasets due to its computational complexity and report that its real-time implementation requires specialized DSP-or FPGA-based hardware solutions. Most of the aforementioned approaches either assume that the color model of the shadowed and illuminated path is known in advance or use additional, expensive sensors that provide their autonomous navigation systems with range measurements.…”

“…The paper [11], which is based on monocular sensing as well, proposes to achieve path following by combination of vanishing point detection and color-based probabilistic segmentation. Their vanishing point detection closely follows the method presented in [12], which fuses the responses of Gabor filters applied on the image by means of particle filtering, and the 'roadness' of the image pixels is classified by means of Gaussian Mixtures.…”

Fast Vanishing Point Estimation Based on Particle Swarm Optimization

“…Moghadam et al used skyline to reduce the candidates of vanishing point [10]. Furthermore, Haar functions, integral image [9] and orthogonal Gabor [4] filters are also used to improve the efficiency of the vanishing point estimation [9]. This paper also focuses on improving the efficiency of the conventional vanishing point estimation based on C. Russmusen and Kong's method.…”