Identification and prediction of a moving object using real-time global vision sensing

Abstract

This paper deals with a global vision based optical sensing system employed for identifying and intercepting moving objects. Using a color thresh-holding identification algorithm, the system can detect the instantaneous position and the direction angle of moving objects. The vision processing is done in real-time, effectively within 16.67 ms sample time of an interlaced NTSC video image. Incremental tracking is employed to save on the vision processing time. Since odd and even fields are processed separately, there is inherent quantization noise in the system, which can be smoothed by using Kalman filtering. A case study of a robot goalkeeper behavior, including interception and clearance of ball, has been presented in detail. Based on the vision sensor data, a prediction technique is used to intercept the ball traveling towards the goal. A state transition based algorithm for goalkeeper behavior is also introduced.