Huijing Zhao
Center for spatial information science
University of Tokyo
Email: chou@paddy.iis.u-tokyo.ac.jp

A Future Image

Till now, our research work focus on the first three topics. Let me introduce two systems that are developed in our research.

A vehicle-borne 3D mapping system called VLMS (Vehicle-borne Laser Measurement System) has been developed in a joint research project between Asia Air Survey Co. Ltd and the Univ. of Tokyo, where three single-row laser range scanners ("laser scanner") and six line CCD cameras ("line camera") are mounted on a van to map object's geometry and texture along the streets, a GPS (Global Positioning System)/INS (Inertial Navigation System) based navigation unit is also available. The data output of VLMS are used to reconstruct a detailed spatial database of central urban area, where an automated approach was developed to reconstruct textured 3D models of buildings, roads and trees automatically using vehicle-borne laser range and line images [1], a semi-automated interface was designed for extracting a broad range of urban objects, such as commercial signboards, road boundaries, traffic signs/signals, ph poles/cables etc. [2]. On the other hand, in order to use the VLMS output to update an existing geographic database, an algorithm for efficiently correcting the position and orientation solutions of the GPS/INS based navigation unit, which is rather erroneous in center town, is developed [3]. The efficiency of the system for generating a database of urban details was demonstrated by many experiments in central Tokyo. In addition, a framework of using horizontal laser scanning for navigation, and using vertical laser scanning for 3D mapping is also proposed [4].

A novel tracking system has been developed [7], aimed at providing real-time monitoring of pedestrian behaviors in a crowded environment, such as a railway station, shopping mall, or exhibition hall. A network of single-row laser range scanners is exploited, which are set on the floor, horizontally scanning at ground level to monitor the pedestrian feet. A pedestrian's walking model on the horizontal cross section of ground level is defined and a tracking algorithm is developed to detect and trace the rhythmic swing feet. The reason for us to target at pedestrian feet is that (1) the occlusion at ground level is much lower than at waist height; (2) the reflections occur from swing arms, hand bags and coats are difficult to model to obtain an accurate tracking; and (3) the rhythmic swing feet is the common pattern for a normal pedestrian, which can be measured at the same horizontal plane.

Comparing with normal video camera, laser scanner has the advantages in tracking people in large area. First, it is a form of direct measurement. The extraction of moving objects in a real-world coordinate system is not as time-consuming a task as using normal video camera. Second, as the range measurement can be converted into a rectangular coordinate system with a real dimension on a horizontal plane, it is comparatively easy to calibrate multiple laser scanners and integrate the distributed data to cover a relatively large area. Third, the tracking of a large crowd will be achieved in real time in the near future due to the low computation cost. Finally, although range measurements have poor interpretability compared with video images, to some extent this avoids privacy problem, which is a sensitive topic in public places, such as supermarkets and exhibition halls.

Whereas, a laser scanner based tracking system has also limitations. When tracking a large and dense crowd using laser scanners only, it is difficult to tell exactly which feet belongs to the same pedestrian. If a trajectory is broken due to occlusion, it is difficult to link the fragments together. Also, it is impossible to attach other status, such as sex, height, age, face, cloth, etc. to such a trajectory. A hybrid tracking system using a network of video cameras and single-row laser scanners is developed [8], where walking trajectories are mainly tracked from laser points, video images are used to associate contents, e.g. face, height, contour, color, and so on, to each trajectory.

A number of experiments have been conducted at railway stations, exhibition halls, intersections and streets. Moving trajectories are collected for days in either real-time or off-line mode. They are examined to find the distribution of density, speed, direction, collision and the change with time [9], the data of which are quite needed in the studies of architectural design and behavior study.

Reference