A novel disparity transformation algorithm for road segmentation

Abstract

The disparity information provided by stereo cameras has enabled advanced driver assistance systems to estimate road area more accurately and effectively.In this paper, a novel disparity transformation algorithm is proposed to extract road areas from dense disparity maps by making the disparity value of the road pixels become similar. The transformation is achieved using two parameters: roll angle γ and fitted disparity value d with respect to each row. To achieve a better processing efficiency, golden section search and dynamic programming are utilised to estimate γ and d, respectively. By performing a rotation around γ, the disparity distribution of each row becomes very compact. This further improves the accuracy of the road model estimation, as demonstrated by the various experimental results in this paper. Finally, the Otsu's thresholding method is applied to the transformed disparity map and the roads can be accurately segmented at pixel level.

Road Damage Detection Based on Unsupervised Disparity Map Segmentation

“…This fact leads to gradual disparity change in the horizontal direction (see Fig. 2(a)), making the way of representing road disparity projections using (1) somewhat problematic [13]. Furthermore, compared to the case that the roll angle is zero, the disparity distribution of each row becomes less compact and E min becomes much higher.…”

“…2 http://apolloscape.auto/stereo.html whered = [D(p 1 ),D(p 2 ), · · · ,D(p m )] stores the transformed disparity values. We compare our proposed method with GSS-DP [13] and GD [15]. The comparisons of σ and runtime are illustrated in Table I.…”

“…On the other hand, 2D image analysis-based road damage detection methods can be grouped into two categories: computer vision-based [6], [7], [12]- [15] and machine learningbased [16]- [19]. The former typically pre-processes a 2D image, i.e., an RGB/gray-scale image or a depth/disparity map, using some image processing techniques, e.g., various image filters, to reduce image noise and enhance road damage outline [12], [13]. The pre-processed image is then segmented using some thresholding methods, such as Otsu [14], triangle [6] or watershed [7], to extract damaged road areas.…”

“…The pre-processed image is then segmented using some thresholding methods, such as Otsu [14], triangle [6] or watershed [7], to extract damaged road areas. In [13], we proposed a disparity transformation algorithm which can better distinguish between damaged and undamaged road areas. The transformation parameters were estimated by minimizing an energy function using golden section search (GSS) and dynamic programming (DP).…”

A Robust Pavement Mapping System Based on Normal-Constrained Stereo Visual Odometry

“…cracks and potholes, from RGB images. Furthermore, Fan et al [6] proposed an efficient binocular system that is capable of effectively distinguishing road damage from a transformed disparity map [10]. Current visual odometry and mapping frameworks have demonstrated their accuracy and robustness on various opensource datasets [8], [11], [12].…”