Evaluation of Effectivity of Omnidirectional Image Sensor COPIS in a Real Environment

We designed a new omnidirectional image sensor COPIS (COnic Projection Image Sensor) to guide the navigation of a mobile robot. Under the assumption of the known'motion of the robot (COPIS), an environmental map.of an indoor scene is generated by monitoring azimuth change in the image. We did several experiments in the simple indoor environment to examine the potential of COPIS against effects of observational errors in realtime navigation. The precision of obtained environmental maps was sufficient for robot navigation in such environment. In this paper, we improve the image processing method for extracting and tracking vertical edges, and also evaluate the effectivity of omnidirectional image sensor COPlS in a real indoor and outdoor environment. 1.Introduction While there has been much work on mobile robots with vision systems [I ][2][3], most robots view things only in front of them and avoid static obstacles and as a result, they may collide against objects moving from the side or from behind. To overcome these drawbacks, a sensor is needed to view the environment around the robot so that it may navigate safely. We have proposed an omnidirectional image sensor COPlS (COnic Projection Image Sensor) for guiding navigaiion of a mobile robot [4][5]. The key feature of COPIS is passive sensing of the omnidirectional environment in real-time (at the frame rate of a TV camera), using a conic mirror. Mapping of the scene onto the image by COPlS involves a conic projection and vertical edges in the environment appear as lines radiating from the image center. The main field viewed by COPlS is a side view and the resolution along a radial direction of conic mirror is sufficient to extract vertical edges. We also reported a method of map generation, which is based on azimuth information in the image sequence [6][7]. Under the assumption of known motion of the robot, locations of objects around the robot can be estimated by detecting their azimuth changes in the omnidirectional image. One can recognize the surface of objects and estimate the free space for a mobile robot from the geometrical relation between object points in the image. Using this method, the robot generates an environmental map of an indoor scene while i t is moving in the environment. The method used properties common to many indoor environments. The floor is almost flat and many visible vertical edges are present a b u t the room. heref fore, the method can be used for mobile robots work'ing in most buildings or plants. Using this method, we did several experiments in the simple indoor environment. The obtained precision of environmental maps was sufficient for robot navigation in such environment. However, in this experiment, we simplified the image processing method and the experimental environment to examine the potential of COPIS against effects of observational errors in real-time navigation. In this paper, we improve the image processing method for extracting and tracking vertical edges to discriminate these edges and evaluate the effectivity of omnidirectional image sensor COPlS in a real indoor and outdoor environment. Experiments done in an actual environment support the validity of our method. 2.Conic Projection Image Sensor (COPIS) COPIS mounted on a robot consists of a conic mirror and a TV camera, with its optical axis aligned with the cone's axis, in a cylindrical glass tube. A prototype of the COPlS system is shown in Fig.1. COPlS maps the scene onto the image plane through a conic mirror and a lens. We call this mapping " conic projection ". We describe here how features of the scene appear in the image. Let us use the three dimensional coordinate system 0XYZ, aligned with the image coordinate system o-xy and the Z-axis pointed toward the cone's vertex (see Fig.2). We fix origin 0 at the camera center, thus the image plane is on a level off , where f is the focal length of the camera. A conic mirror yields the image of a point in space on a vertical plane through the point and its axis. Thus, the point P at (X,Y,Z) is projected onto the image point p at (x,y) such that Y Y tan0 = X -x ' (1) If a line is vertical, it appears radially in the image plane, as shown in Fig.2. Thus, COPIS can easily find and track the vertical edges by searching consecutive images for radial edges. For further details, we refer to reader to our preceding report [S]. 3.Assumptions The following properties of the environment and the mobile robot are assumed for image analysis. The floor is almost flat and horizontal while walls and static objects such as desks or shelves have vertical planes. The robot moves in a man-made environment such as a room or a corridor. Its motion parameters are two translational components (U,V) and one rotational 4.Ceneration of an Environmental Map projection. In this projection. the data are segmented into Let us denote the robot motion by (u(t),v(t)) and cp(t). regions of each edge and the azimuth of the edges is cp(t) is the pan angle at time t. Defining the position of P at estimated by calculating the average angle of each region. time t=O by PI(X(O),Y(O),Z(O)), the relative velocity of In the previous method, i t is difficult to distinguish the the point p in the environment at time t is represented by projection data Of all edges in the area where there are (-u(t). -v(t), 0). We get the location of point P at time t as dense vertical edges. Dense areas often appear in the real