To move efficiently in an unknown or uncertain environment, a mobile robot must use observations taken by various sensors to construct information for path planning and execution. A reasonably accurate representation of the external world would also be very useful for robot self-localization. One of the merits of applying multiple sensors to a mobile robot is the enhancement of environment recognition. In this paper, we propose to use sensory information combined from double ultrasonic sensors and a CO) camera. We developed an algorithm based on a dual-transducer design to eliminate errors resulting from the beam opening angle of ultrasonic sensors. An extended discrete Kaiman filter (EDKF) was designed to fuse raw sensory data and to reduce the influence of specular reflection of ultrasonic type transducers, thereby providing a more reliable representation for environment perception. Computer simulation, as well as practical experimental results demonstrate that this sensory system can provide useful and comprehensive environment perception for intelligent robotics.