Design of a low power multiple object tracking vision sensor for use in robot localization and surveillance tasks

Abstract

This paper presents a smart CMOS image sensor with capability of detecting motion and positioning of moving objects such as robots on the chip. The proposed method is based on in-pixel frame difference extraction and event generation upon threshold level detection. Due to the localization unit of the designed sensor, it is able to detect multiple moving objects, obtain each object's size and position, and ultimately localize them. By simulation of the sensors processing algorithms and behavior in MATLAB software, activity rate of the sensors data lines has been studied. According to the simulations and comparisons with conventional image sensors, reduction of sensors activity rate and consequently power consumption has been observed in the proposed method.