We present a robust and real time approach to lane marker detection in urban streets. It is based on generating a top view of the road, filtering using selective oriented Gaussian filters, using RANSAC line fitting to give initial guesses to a new and fast RANSAC algorithm for fitting Bezier Splines, which is then followed by a post-processing step. Our algorithm can detect all lanes in still images of the street in various conditions, while operating at a rate of 50 Hz and achieving comparable results to previous techniques.

Real time detection of lane markers in urban streets

Real time Detection of Lane Markers in Urban Streets

Given a single image of an arbitrary road, that may not be well-paved, or have clearly delineated edges, or some a priori known color or texture distribution, is it possible for a computer to find this road? This paper addresses this question by decomposing the road detection process into two steps: the estimation of the vanishing point associated with the main (straight) part of the road, followed by the segmentation of the corresponding road area based upon the detected vanishing point. The main technical contributions of the proposed approach are a novel adaptive soft voting scheme based upon a local voting region using high-confidence voters, whose texture orientations are computed using Gabor filters, and a new vanishing-point-constrained edge detection technique for detecting road boundaries. The proposed method has been implemented, and experiments with 1003 general road images demonstrate that it is effective at detecting road regions in challenging conditions.

/pdf/general-road-detection-from-a-single-image-1vxtfzpegg.pdf

General Road Detection From a Single Image

By using an onboard camera, it is possible to detect the free road surface ahead of the ego-vehicle. Road detection is of high relevance for autonomous driving, road departure warning, and supporting driver-assistance systems such as vehicle and pedestrian detection. The key for vision-based road detection is the ability to classify image pixels as belonging or not to the road surface. Identifying road pixels is a major challenge due to the intraclass variability caused by lighting conditions. A particularly difficult scenario appears when the road surface has both shadowed and nonshadowed areas. Accordingly, we propose a novel approach to vision-based road detection that is robust to shadows. The novelty of our approach relies on using a shadow-invariant feature space combined with a model-based classifier. The model is built online to improve the adaptability of the algorithm to the current lighting and the presence of other vehicles in the scene. The proposed algorithm works in still images and does not depend on either road shape or temporal restrictions. Quantitative and qualitative experiments on real-world road sequences with heavy traffic and shadows show that the method is robust to shadows and lighting variations. Moreover, the proposed method provides the highest performance when compared with hue-saturation-intensity (HSI)-based algorithms.

Road Detection Based on Illuminant Invariance

Autonomous vehicle navigation is challenging since various types of road scenarios in real urban environments have to be considered, particularly when only perception sensors are used, without position information. This paper presents a novel real-time optimal-drivable-region and lane detection system for autonomous driving based on the fusion of light detection and ranging (LIDAR) and vision data. Our system uses a multisensory scheme to cover the most drivable areas in front of a vehicle. We propose a feature-level fusion method for the LIDAR and vision data and an optimal selection strategy for detecting the best drivable region. Then, a conditional lane detection algorithm is selectively executed depending on the automatic classification of the optimal drivable region. Our system successfully handles both structured and unstructured roads. The results of several experiments are provided to demonstrate the reliability, effectiveness, and robustness of the system.

A Sensor-Fusion Drivable-Region and Lane-Detection System for Autonomous Vehicle Navigation in Challenging Road Scenarios

An eye contour extraction method which combines a simplied version of Active Shape Model(ASM) with a gradient method is proposed. Considering the large amount of calculations required by ASM, it is only used to extract eyelids. As iris is considered to have some more regular shape, the detection of iris is done by the simple but fast gradient method, which is improved by introducing gradient value to the weight matrix. Our detection method has been implemented in the C programming language and experimental results shows good accuracy and efficiency.

A composite method to extract eye contour

The ability of obstacles detection and tracking is essential for safe visual guidance of autonomous vehicles, especially in urban environments. In this paper, we first overview different plane projective transformation (PPT) based obstacle detection approaches under the planar ground assumption. Then, we give a simple proof of this approach with relative affine, a unified framework that includes the Euclidean, projective and affine frameworks by generalization and specialization. Next, we present a real-time hybrid obstacle detection method, which combined the PPT based method with the region segmentation based method to provide more accurate locations of obstacles. At last, with the vehicle's position information, a Kalman Filter is applied to track obstacles from frame to frame. This method has been tested on THMR-V (Tsinghua Mobile Robot V). Through various experiments we successfully demonstrate its real-time performance, high accuracy, and high robustness.

Vision-based real-time obstacle detection and tracking for autonomous vehicle guidance

Pedestrian detection, as one of the most important modules of intelligent vehicles, is a challenging topic for researchers. In this paper, we propose a stereo-vision-based and shape-based approach for pedestrian and bicyclist detection. An efficient stereo system and an obstacle detection algorithm based on v-disparity map help us locate potential regions. Using the shape of the rigid part (upper body) of pedestrians and bicyclists, the matching criterion of partial Hausdorff distance, can efficiently detect them from front or back views. Our algorithm is tested off-line on a large mount of data, and the experiments show its realtime and reliable performance.

Shape-based Pedestrian/Bicyclist Detection via Onboard Stereo Vision

Pose estimation is one of the key issues in the research of intelligent vehicles. In this paper, a real-time pose estimation algorithm based on vision is proposed and implemented. The ground plane assumption is used to simplify the interframe motion model to a 2D plane motion model, which reduces the computation and avoids the difficulty in feature point selection in outdoor environments. This algorithm is composed of two parts: the Gradient Angle Histogram algorithm and the Iterative Gradient Closest Point algorithm. The fusion of these two algorithms successfully addresses the local minimum problem and the high computation problem with the ICP algorithm. Experimental results with both synthetic data and real data prove the high accuracy, low computation, and high robustness to outliers in this algorithm.

http://ieeexplore.ieee.org/iel5/9278/29469/01336392.pdf

Vision based real-time pose estimation for intelligent vehicles

Ego-motion estimation is a key issue of outdoor mobile robot navigation, especially in the demand of moving object tracking. In this paper, we proposed correspondence based method, which apply iterative closest point (ICP) algorithm to match feature points on ground plane. However the outliers in the scene contribute false measurement for estimation, thus we introduce a stereovision-based method to detect free-space on the road plane. Then we extract edge points in the free-space as primitives, which avoid the limit of the rigid scene hypothesis. This method has been tested on THMR-V (Tsinghua mobile robot V), which is an outdoor mobile robot developed by Tsinghua University. Through various experiments we successfully demonstrate its real-time performance and high robustness.

https://ieeexplore.ieee.org/iel5/9294/29577/01343605.pdf

Hong Wang

Papers

A composite method to extract eye contour

Vision-based real-time obstacle detection and tracking for autonomous vehicle guidance

Shape-based Pedestrian/Bicyclist Detection via Onboard Stereo Vision

Vision based real-time pose estimation for intelligent vehicles

Real-time vision based ego-motion estimation for outdoor mobile robot