3-Dimensional Recognition Using a Moving Camera

点击翻译按钮获取中文摘要

Speeds of on-board computers are critical in the problems of autonomous guidance by visual image processing. In this research data-folw processors are applied to this problem.A test vehicle which is equipped with 8 data-flow-pipeline processors are developed for this purpose. The vehicle has a length of 43 cm and a width of 20 cm and its maximum speed is 40 cm/sec.Each of the processor has the computation capacity of 5 MIPS (Million Instructions Per Second), and totally 40 MIPS which is 100 times as high as speed of usual microprocessors is supplied by 8 processors and the controllers for the processors.In the first phase experiments of course tracking are done. The course is designated by white tape and includes diverging points. Image flames acquired through an ITV camera are treated by the data-flow processors. Treating interval for one flame is 6.3 msec, which is satisfactory since it is far less than 33 msec: a flame cycle of video-composite signal.In the second phase obstacle avoidance is accomplished by the system. There are two kinds of obstacles on the course: white blocks which are required to be avoided and white sheets which are to be run through. Signals from two ITV cameras are used in order to detect the difference of these obstacles. The vehicle can run along the course at the maximum speed 40 cm/sec under the environment. Treating interval for one flame is 230 msec.In this study it is confirmed that the computation capacity of the processors is important toward the visual image guidance in the restricted environment like factories and highways and that new LSIs have adequate capacity for this purpose.