CMOS image sensor for spatiotemporal image acquisition

Abstract

Abstract. We present a 64-ϫ 64-pixel CMOS image sensor chip that can acquire a 16-gray-level image containing both spatial and temporal information of the moving luminous IntroductionThe determination of the speed of a moving object requires the knowledge of the trajectory and the time taken by the object to cover a certain distance along the trajectory. To achieve this by machine vision using classical cameras, one needs to take several images of the object at different time instances so as to obtain the trajectory, displacement, and time information. Even for relatively slow moving objects this requires that the images be taken and transferred to the computer at a relatively high rate. Let us consider, for instance, an object moving horizontally at 2 m / s and suppose that the width of the view field of the vision system is 10 cm. The object would take 0.05 s to cover the 10-cm distance. Since at least two images of the object are necessary for computing the average speed, this means that the images must be acquired at a minimum rate of one image every 0.05 s ͑note that the video rate, 25 images/ s, is one image every 0.04 s͒. For faster objects or if more than two images are required then the image rate would increase.It is well known that in classical systems the image transfer rate is a bottleneck and one solution is to store the images on the chip at a very high rate and later transfer them to the computer at a much lower rate. The CMOS