This paper proposes a systematic approach to evaluate algorithms for extracting road marking features from images. This specific topic is seldom addressed in the literature while many road marking detection algorithms have been proposed. Most of them can be decomposed into three steps: extracting road marking features, estimating a geometrical marking model, tracking the parameters of the geometrical model along an image sequence. The present work focuses on the first step, i.e. feature extraction. A reference database containing over 100 images of natural road scenes was built with corresponding manually labeled ground truth images (available at http://www.lcpc.fr/en/produits/ride/). This database enables to evaluate and compare extractors in a systematic way. Different road marking feature extraction algorithm representing different classes of techniques are evaluated: thresholding, gradient analysis, and convolution. As a result of this analysis, recommendations are given on which extractor to choose according to a specific application.

/pdf/evaluation-of-road-marking-feature-extraction-1s6rv5oowd.pdf

Evaluation of Road Marking Feature Extraction

We propose a new cooperative fusion approach between stereovision and laser scanner in order to take advantage of the best features and cope with the drawbacks of these two sensors to perform robust, accurate and real time-detection of multi-obstacles in the automotive context The proposed system is able to estimate the position and the height, width and depth of generic obstacles at video frame rate (25 frames per second) The vehicle pitch, estimated by stereovision, is used to filter laser scanner raw data Objects out of the road are removed using road lane information computed by stereovision Various fusion schemes are proposed and one is experimented Results of experiments in real driving situations (multi-pedestrians and multi-vehicles detection) are presented and stress the benefits of our approach

Cooperative Fusion for Multi-Obstacles Detection With Use of Stereovision and Laser Scanner

The development of wireless sensor networks, such as researchers Advanced Driver Assistance Systems (ADAS) requires the ability to analyze the road scene just like a human does. Road scene analysis is an essential, complex, and challenging task and it consists of: road detection (which includes the localization of the road, the determination of the relative position between vehicle and road, and the analysis of the vehicle's heading direction) and obstacle detection (which is mainly based on localizing possible obstacles on the vehicle's path). The detection of the road borders, the estimation of the road geometry, and the localization of the vehicle are essential tasks in this context since they are required for the lateral and longitudinal control of the vehicle. Within this field, on-board vision has been widely used since it has many advantages (higher resolution, low power consumption, low cost, easy aesthetic integration, and nonintrusive nature) over other active sensors such as RADAR or LIDAR. At first glance the problem of detecting the road geometry from visual information seems simple and early works in this field were quickly rewarded with promising results. However, the large variety of scenarios and the high rates of success demanded by the industry have kept the lane detection research work alive. In this article a comprehensive review of vision-based road detection systems vision is presented.

Keeping the vehicle on the road: A survey on on-road lane detection systems

In this study, fuzzy approach with fault-tolerance has proposed to fuse heterogeneous sensed data and overcome the problem of imprecise collision warning due to perturbed input signal when processing the pre-crash warning. Meanwhile, another problem relevant to the danger in drowsy driving, involving fatigue level, carbon monoxide concentration, and breath alcohol concentration, was considered and has approximately reasoned to an extra reaction time to modify NHTSA algorithm. A vision-sensing analysis cooperating with global-positioning system is applied for lane marking detection and collision warning, particularly exchanging the dynamic and static information between neighboring cars via inter-vehicle wireless communications. In addition to pre-crash warning, event data recording very useful for accident reconstruction on scene is also established here. In order to speed up data fusion on both quantum-tuned back-propagation neural network (QT-BPNN) and adaptive network-based fuzzy inference system (ANFIS), a distributed dual-platform DaVinci+XScale_NAV270 has been employed. Several tests on system's reliability and validity have been done successfully, and the comparison of system effectiveness showed that our proposed approach outperforms two current well-known collision-warning systems (AWS-Mobileye and ACWS-Delphi).

Intelligent data fusion system for predicting vehicle collision warning using vision/GPS sensing

Road traffic incidents analysis has shown that a third of them occurs without any conflict which indicates problems with road following. In this paper a driving safety assistance system is introduced, whose aim is to prevent the driver drifting off or running off the road. The road following system is based on a frontal on-board monocular camera. In order to get a high degree of reliability and robustness, an original combination of three different algorithms is performed. Low level results from the first two algorithms are used to compute a reliability indicator and to update a high level model through the third algorithm using Kalman filtering. Searching areas of the road sides for the next image are also updated. Experimental results show the reliability and the robustness of this original association of three different algorithms. Various road situations are addressed, including roads with high curvature. A multi-lanes extension is also presented.

/pdf/a-reliable-and-robust-lane-detection-system-based-on-the-23y5cvdelz.pdf

A Reliable and Robust Lane Detection System Based on the Parallel Use of Three Algorithms for Driving Safety Assistance

In the paper, a road edge identification algorithm is developed. The new idea of this method is to use natural road edge, as well as the white strip for road information acquisition. The natural road edge does not be easily polluted like the white lane maker does, so it indicates better adaptability to the outdoor environment. In the algorithm, we use both the pixel feature and the frame feature to identify the road edge, which is referred to as the whole road model. Because several road constrains is used to ensure the road edge detection, the algorithm is immune to the influence of the image disturbance. The algorithm of the road edge identification includes two stages: initialization detection and tracing detection. The initialization stage detects the road edge from the whole road image. The trace algorithm uses the region of interest (ROI) to limit detecting area, which can save much time. In order to give a measure of the reliability of the road detecting result, this paper presents a road edge identification estimation function, which can estimate the reliability of the road edge.

Libin

Papers

A vision-based road edge detection algorithm