Zhencheng Hu

U-V-disparity: an efficient algorithm for stereovision based scene analysis

Reliable understanding of the 3D driving environment is vital for obstacle detection and adaptive cruise control (ACC) applications. Laser or millimeter wave radars have shown good performance in measuring relative speed and distance in a highway driving environment. However the accuracy of these systems decreases in an urban traffic environment as more confusion occurs due to factors such as parked vehicles, guardrails, poles and motorcycles. A stereovision based sensing system provides an effective supplement to radar-based road scene analysis with its much wider field of view and more accurate lateral information. This paper presents an efficient solution using a stereovision based road scene analysis algorithm, which employs the U-V-disparity concept. This concept is used to classify a 3D road scene into relative surface planes and characterize the features of road pavement surfaces, roadside structures and obstacles. Real-time implementation of the disparity map calculation and the U-V-disparity classification is also presented.

A complete U-V-disparity study for stereovision based 3D driving environment analysis

Reliable understanding of the 3D driving environment is vital for obstacle detection and adaptive cruise control (ACC) applications. Laser or millimeter wave radars have shown good performance in measuring relative speed and distance in a highway driving environment. However the accuracy of these systems decreases in an urban traffic environment as more confusion occurs due to factors such as parked vehicles, guardrails, poles and motorcycles. A stereovision based sensing system provides an effective supplement to radar-based road scene analysis with its much wider field of view and more accurate lateral information. This paper presents an efficient solution using a stereovision based road scene analysis algorithm which employs the U-V-disparity concept. This concept is used to classify a 3D road scene into relative surface planes and characterize the features of road pavement surfaces, roadside structures and obstacles. Real-time implementation of the disparity map calculation and the U-V-disparity classification is also presented.

Real-time data fusion on tracking camera pose for direct visual guidance

To properly align objects in the real and virtual world in an augmented reality (AR) space, it is essential to keep tracking cameras exact 3D position and orientation, which is well known as the Registration problem. Traditional vision based or inertial sensor based solutions are mostly designed for well-structured environment, which is, however, unavailable for outdoor uncontrolled road navigation applications. This paper proposed a hybrid camera pose tracking system that combines vision, GPS and 3D inertial gyroscope technologies. The fusion approach is based on our PMM (parameterized model matching) algorithm, in which the road shape model is derived from the digital map referring to GPS absolute road position, and matches with road features extracted from the real image. Inertial data estimates the initial possible motion, and also serves as the relative tolerance to stabilize output. The algorithms proposed in this paper are validated with the experimental results of real road tests under different conditions and types of road.

SOLUTION OF CAMERA REGISTRATION PROBLEM VIA 3D-2D PARAMETERIZED MODEL MATCHING FOR ON-ROAD NAVIGATION

This paper presents a dynamical solution of camera registration problem for on-road navigation applications via a 3D-2D parameterized model matching algorithm. The traditional camerafs three dimensional (3D) position and pose estimation algorithms have always employed fixed and known-structure models as well as the depth information to obtain the 3D-2D correlations, which is however unavailable for on-road navigation applications since there are no fixed models in the general road scene. With the constraints of road structure and on-road navigation features, this paper presents a 2D digital road-based road shape modeling algorithm. Dynamically generated multi-lane road shape models are used to match real road scenes to estimate the camera 3D position and pose data. Our algorithms have successfully simplified the 3D-2D correlation problem to the 2D-2D road model matching on the projective image. The algorithms proposed in this paper are validated with the experimental results of real road tests under different conditions and types of road.

Tracking cycle: a new concept for simultaneous tracking of multiple moving objects in a typical traffic scene

Because of the complicated motion pattern of moving objects in a typical on-road traffic scene, detecting the real moving objects and properly tracking each one have become very difficult and always employ optical flows and fixed template matching to detect and track the moving objects. Unfortunately, it has to pay a huge calculation cost and can not deal with some complicated motion pattern like overlapping of tracked objects. A new concept of tracking cycle is proposed in the paper. With the continuous estimation of moving objects vitality and reliability values, our method can detect and track up to more than 30 moving objects in real time. The flexible dynamic template matching method is also presented in the paper, which makes our system faster and more efficient for real-time ITS applications. Experiments on real outdoor road scenes show the accuracy and efficiency of our approach.

Precise curvature estimation by cooperating with digital road map

This paper proposed a novel algorithm for road geometry estimation ahead of vehicle depending on present vehicle localization. In order to provide precise and robust on-coming road geometry estimation under different environmental conditions, we employ a new hybrid vehicle localization approach combining the commercial GPS and inertial sensors. In this work, road geometry (planar curvature) is initially estimated from the prevalent digital road map following the rules of road construction design and basic geometrical elements (straight line, clothoid curve and circle curve). Therefore, road trajectory between vehicle and road ahead could be calculated by flexible road model, relying on precise curvature which has been designed beforehand. In simulation test, road geometrical curve within camerapsilas view range is built by variance curvature in meters. Meanwhile, road tests show similarity between simulation results and real road scenes. This approach is also proved effectiveness and robustness under different kinds of environment conditions.

Vehicle Localization with Global Probability Density Function for Road Navigation

This paper presents a novel approach of real-time vehicles localization (position and orientation) estimation. Fusion of GPS, gyroscope, speedometer and visual data is employed here to provide real time and accurate localization information. Global probability density function(PDF) is adopted to be the blending factor instead of general Kalman gain, which allows our approach to be robust and accurate for most of practical systematic problems, since the basic measurements from GPS may cause data drift or large infrequent data jumps during the fusion processing. Combining with visual data for lane shape recognition and tracking, our approach can provide as accurate as 3 to 5 meters RMS location accuracy at about 30 Hz, with less then 35 ms delay. This approach has been adapted to the direct visual navigation system in VICNAS.

3D Vehicle Extraction and Tracking from Multiple Viewpoints for Traffic Monitoring by using Probability Fusion Map

This paper presents a novel solution of vehicle occlusion and 3D measurement for traffic monitoring by data fusion from multiple stationary cameras. Comparing with single camera based conventional methods in traffic monitoring, our approach fuses video data from different viewpoints into a common probability fusion map (PFM) and extracts targets. The proposed PFM concept is efficient to handle and fuse data in order to estimate the probability of vehicle appearance, which is verified to be more reliable than single camera solution by real outdoor experiments. An AMF based shadowing modeling algorithm is also proposed in this paper in order to remove shadows on the road area and extract the proper vehicle regions.

High-speed lane detection for road geometry estimation and vehicle localization

This paper presents a fast and robust lane detection method designed to track 3D road shape and estimate accurate vehicle location in real-time. This method is based upon road curvature calculated from digital road network. This curvature gives cues for detection in some difficult situation like road lane markers are occupied by other vehicles, dirt and shadow. Moreover, to stabilize estimated road parameters, we applied AMF&PDF to handle the variation of road parameters in every frame and provide continuous vehicle location data. Experiment results show the effectiveness and robustness of our approach under different kinds of environment conditions.

Dynamical road modeling and matching for direct visual navigation

This paper proposes a new concept of direct visual navigation, (DVN), which superimposes virtual direction indicators and traffic information into the real road scene to give drivers efficient and direct visual navigation guidance. To align the virtual objects properly with respect to the real world, we need to solve the so-called Registration Problem in Augmented Reality (AR) context. Traditional solutions always employ a fixed and known-structure model as well as the object depth information to obtain the 3D-2D correlations, which is not possible in the case of on-road driving navigation. With the constraints of road structure and on-road vehicle motion features, this paper presents a dynamical multi-lane road shape modeling method as well as a road model matching method to simplify the 3D-2D correlation problem to the 2D-2D road model matching on projective image. Additional road shape lookup table (RSL) concept is also presented in this paper to calculate the road model matching score. The algorithms proposed in this paper are validated with the experimental results from real road test under different conditions and types of road.