This paper describes a new vision chip architecture for high-speed target tracking. The processing speed and the number of pixels are improved by hardware implementation of a special algorithm which utilizes a property of high-speed vision and introduction of bit-serial and cumulative summation circuits. As a result, 18 objects in a 128 128 image can be tracked in 1 ms. Based on the architecture, a prototype chip has been developed; 64 64 pixels are integrated in 7 mm square chip and the power consumption for obtaining the centroid of an object per every 1 ms is 112 mW. Some experiments are performed on the evaluation board which is developed for evaluation under the condition of actual operation. High-speed target tracking including multitarget tracking with collision and separation has successfully been achieved. Index Terms-Target tracking, vision chip.

No abstract

No abstract

Abstract-This paper describes real-time shape measurement using a newly developed high-speed vision system. Our proposed measurement system can observe a moving/deforming object at high frame rate and can acquire data in real-time. This is realized by using two-dimensional pattern projection and a high-speed vision system with a massively parallel co-processor for numerous-point analysis. We detail our proposed shape measurement system and present some results of evaluation experiments. The experimental results show the advantages of our system compared with conventional approaches. I. INTRODUCTIONThere has been an increasing need for shape measurement techniques in a wide range of application fields. In particularly, approaches using vision are useful because of their contactless nature and flexibility. On the other hand, there are demands for improved performance such as accuracy, measurement range, time, cost and so on, and many kinds of systems have been developed to meet these demands [1].In this paper, we focus on the improvement of the measurement time and realize high-speed time-sequential shape measurement. This type of measurement will have a great impact on a wide variety of applications, such as control tasks for robotics; surgery support, for example, observation of a beating heart; rapid visual inspection of products; vehicle application, for example, checking road surfaces; and humanmachine interaction with high accuracy and flexibility.On the other hand, in such application fields, we need to observe a high-speed moving rigid-body and a deforming or vibrating non-rigid body. Conventional shape measurement, however, is mainly performed in situations where the relationships between the objects being measured and observation devices are completely known by ensuring stable or controllable conditions. Also, the performance of conventional systems in terms of throughput and latency is usually not good enough for real-time control applications; for example, throughput on the order of a thousand frames per second and latency on the order of milliseconds is generally required for robotic control.Based on these backgrounds, the feasible system is considered to have three requirements: (A) the ability to observe a moving object, (B) continuous observation at high frame rate, and (C) real-time acquisition of shape information. To realize these requirements, our proposed system employs an approach in which multiple spots in a single image of a shape are acquired using high-speed vision for numerouspoint analysis. As a result, the constructed system achieved 955-fps measurement of shape information with 4.5 ms

No abstract