Jungwon Kang

Stereo-vision based free space and obstacle detection with structural and traversability analysis using probabilistic volume polar grid map

The detection of free space and obstacles in a scene is essential for safe driving. Among sensors for environment perception, a stereo-vision is promising as it provides 3D perception information. Moreover current decreasing price of a camera module makes a vision sensor attractive, taking into account the consumer product. In this paper we propose an algorithm for the detection of free space and obstacles in a scene, using a stereo-vision. Contrary to previous generic obstacle detection methods that have a strong assumption of camera placement with respect to the ground or estimate ground parameters for free space/obstacles separation, our method analyzes 3D reconstructued structures of an environment. The environment is represented on the proposed probabilistic volume polar grid map. The probabilistic volume polar grid map is a simplified representation of the environment, using the volumes generated from the reconstructed point hypotheses. The volume polar grid map combines with an image plane, so that the reconstruction uncertainty representation and the free space computation are straightforward. The map is analyzed by the two ways: (1) the analysis of the structural characteristics of volumes and (2) the traversability analysis. The traversability analysis gives the free space and the nearest obstacle in each search direction, while the structural characteristics analysis provides potential obstacles in a more wide range. The results from both analysis modules are combined to provide information of the free space, obstacles and potential obstacles in the given scene, which is useful for safe driving. Our system is expected to be used as a driving assistance system.

The Development of Sensor System and 3D World Modeling for Autonomous Vehicle

This paper describes a novel sensor system for 3D world modeling of an autonomous vehicle in large-scale outdoor environments. When an autonomous vehicle performs path planning and path following, well-constructed 3D world model of target environment is very important for analyze the environment and track the determined path. To generate well-construct 3D world model, we develop a novel sensor system. The proposed novel sensor system consists of two 2D laser scanners, two single cameras, a DGPS (Differential Global Positioning System) and an IMU (Inertial Measurement System). We verify the effectiveness of the proposed sensor system through experiment in large-scale outdoor environment.

Moving Region Segmentation Using Sparse Motion Cue from a Moving Camera

This paper presents a method for pixel-wise segmentation of moving regions using sparse motion cues on an image from a freely moving camera. The main idea is to utilize residual motion, i.e., motion relative to a background, on sparse grid points. Our algorithm consists of three parts: global motion estimation, characterization of points based on sparse motion cue, and pixel-wise labeling of moving regions. Experimental results on real image sequences are presented, showing the effectiveness of the proposed method.

Online motion segmentation using spatially-constrained J-linkage in dynamic scene

We present a method for online motion segmentation in dynamic scenes. Here the dynamic scene is a scene without restrictions on the motion of objects and the motion of a camera. Such a scene causes complex interaction of objects and a camera, leading to the generation of trajectories corrupted by noise and outliers. Moreover, no prior knowledge of the number of objects is given, and the number of objects can vary as the motion of objects. In this paper, we deal with clustering of trajectories in dynamic scenes, while estimating the number of objects at the same time. The basic idea is to find motion models that support the motion of trajectories, and cluster the trajectories according to the support motion models, instead of handling trajectories directly for clustering. To do so, we adopted online J-linkage algorithm proposed in [7], an online multiple model estimation method. Based on the observation that points on one object are located nearby, we applied spatially-constrained agglomerative clustering to the J-linkage algorithm. This spatial constraint can drastically reduce searching time in clustering. The proposed method operates in an online and incremental fashion, so that it is applicable to real-time applications. We tested our method on the Hopkins datasets, demonstrating the effectiveness of the method in dynamic scenes.

Robust clustering of multi-structure data with enhanced sampling

This paper deals with the improvement of clustering results by enhancing performance of sampling. The proposed clustering framework is established on a J-linkage framework [4] with guided sampling technique for multi-structures [7]. We tested the proposed method on publicly available dataset, verifying the validness of the proposed method.

A practical 6D robot pose estimation using GPS and IMU in outdoor

The aim of this study is to estimate 6D robot pose. We estimate 6D robot pose using GPS (Global IMU(Inertial Measurement System) in outdoor. We proposed reliability method for 6D robot pose. We get 3D mapping results and compare the 3D mapping results with and without our method. The 3D mapping results with our method was more accurate.

Building a mobile platform for spatiotemporal integration of 3D outdoor world models

This paper introduces our mobile platform, which equips multiple sensors for 3D outdoor world modeling. The data from GPS and IMU on the platform are fused to provide vehicle pose on the ground. The laterally installed LRFs(Laser Range Finder) on each side of the vehicle give environment perception information. The data from the LRFs are fused with cameras in order to make colored LRF points. With these multiple sensors, we have successfully built spatiotemporal integrated 3D world models of outdoor environments.