Compact vision sensor for motion control

Abstract

Authors are developing a rotorcraft with autonomous controller for the remote sensing of agricultural fields. Accurate position sensing is essential to the development of the autonomous control system. One major method for the position sensing is to use the image processing technique. Among the image processing techniques labeling processing is most practical and is widely used for detecting the position of target objects. The labeling technique often requires high computing performance. In order to make our remote sensing system available at various practical fields, our labeling technique should be realized in compact system with enough processing speed. In this study, a FPGA based compact vision sensor to achieve fast labeling procedures is proposed. A feature of our vision sensor is that labeling process is simplified and fastened by introducing the horizontal center line information on behalf of contour data of the target objects. Therefore the performance of the processing speed and memory usage are enhanced. Due to our original technique, real-time labeling process could be achieved on the FPGA based compact vision sensor. Through several experiments, the excellent performance of the prototyped vision sensor on target detection was demonstrated.