Abstract
People get information of the external world by senses. The information from the sense of sight that is especially important as we known. To make a machine more intelligent, many researches have been reported. For an intelligent machine the acquisition of the circumference situation is most important. We considered a vision sensing system which imitate man's visual function and understand external world information, and use it as a driver assistant device in automotives.
One of the visual functions of man is that the wide circumference information can be obtained quickly by the eyeball motility. And the eyeball motility is divided into a high-speed element (saccardic movement) and of a low-speed element (pursuit movement). A lot of visions sensing systems are used, which are fixed permanently or installed in a mover which can rotate in a certain range. Therefore, the wide circumference information can not be obtained quickly by these systems. We have developed a new vision sensing system. This system includes a new vision sensor that can imitate intermittence movement of an eye and acquire the information on the large range quickly, dynamic image composition and obstacle detection algorisms.
Recently, there are many reports about compositing panorama image. Panoramic image composition is useful technique for many applications such as virtual reality, security surveillance and so on. However, most of the conventional techniques are focused on creating a static panoramic image from a set of input image sequences from static cameras. We propose a new compositing method which can dynamically composite the input images to panoramic images while the intermittency movement vision sensor (IMVS) is mounted in a vehicle. And an obstacle detection algorism is used to detect and recognize the appearance of the obstacle object and the action pattern of the movement in the panoramic images. Some experiments had been made in a vehicle provide with the vision sensing system in the front of it as a driver Assistant system. Wide surrounding information can be obtained, and be useful for the road safety to decrease the traffic accident.