中文摘要 目標偵測和追蹤對於機器視覺系統而言是相當重要且基本的技術之一,然而目前常見的應用多為單一目標偵測和追蹤,因此發展可以同時偵測和追蹤多個目標的視覺系統是一個重要的研究方向。本文的研究目的即在設計一套根基於立體視覺的多目標追蹤系統,此系統由二組影像追蹤平台構成,分別具有二部攝影機和二個旋轉自由度,可以在動態影像中追蹤各自的移動目標。首先藉由顏色辨識和影像處理方法自連續影像中偵測分別具有不同顏色的移動目標,再經影像標籤化與面積過濾後即可求出各個目標在影像中的質心位置,最後運用立體視覺演算法估算個別目標與攝影機之間的相對位置,將此位置轉換成馬達控制訊號,影像追蹤平台即可追蹤移動目標物。 關鍵字:機器視覺、立體視覺、多目標追蹤、影像追蹤
Abstract Target detection and tracking is one of the most important and fundamental technologies for a machine vision system, however, current common application are mostly one target detection and tracking. Hence, developing a vision system which can detect and track multiple targets simultaneously is an important research topic. The purpose of this paper is to design a stereo vision-based multi-target tracking system. The system consists of two image tracking platforms, where one platform includes two cameras and two rotational degrees of freedom, can track their moving target in live video streaming. To begin with, we use color recognition and image processing methods to detect moving targets with different colors in continuous image streaming. Next, after the process of image labeling and filtering, the centroid of each target in image sequences will be determined. Finally, we can estimate the relative position between targets and cameras by stereo vision algorithm. With converting the relative position into motor control signals, the two image tracking platforms will be able to track their moving targets. Keywords: machine vision; stereo vision; multi-target tracking; image tracking