隨著無線網路與行動裝置普及,基於WiFi訊號的室內定位系統是近年來熱門的研究議題之一。在過去的研究方法中主要分為:鄰近式定位法、三角定位法與環境分析定位法。然而,鄰近式定位法的缺點為精準度低,三角定位法因室內環境多路徑與障礙物的影響,使得定位精準度大幅下降,且此種方式須知道每一個AP位置。環境分析式雖擁有較高的精準度,但在系統離線狀態時,需要耗費大量的系統架設時間收集訊號的特徵資料。本論文針對上述的問題,提出基於WiFi訊號強度的快速混合式室內定位系統,將三角定位法環境分析式定位法的特徵結合,能以較少的取樣點密度、不需建立AP位置的快速系統建置特性,應用在多層樓空間的室內定位系統。系統主要分為離線與線上狀態:離線狀態分為資料取樣模式與系統校正模式;線上狀態分為樓層判斷與平面定位兩個定位流程。並以研究中的雙矩陣法、權重篩選法、矩陣權重篩選法與權重篩選矩陣法做比較,實驗結果顯示系統的權重篩選法能達到理想的定位精準度。
In recent years, the research of WiFi-based indoor positioning system is prone to attracted with the widely deployed of WLAN and mobile devices. In the past, the positioning method has been divided into three methods mainly: proximity, trilateration and scene analysis method. However, the shortcoming of proximity method is low accuracy. The trilateration method becomes worse accuracy in complex environment cause by multipath and obstacles. And it must be known the location of APs (access points). The scene analysis method has higher accuracy. But it needs to spend a lot of time for samples fingerprint of signal in system offline. In view of this, this thesis proposes a WiFi-based indoor positioning system that combines both of the characteristic of trilateration and scene analysis methods. It has fast system installation feature by using less sample points and don’t need to know each locations of AP. The proposed method determines which floor the user maybe on first and then determines the position where the user maybe at. The implementation of the proposed system has divided into two steps: offline and online. The offline mode includes data collection and system calibration. The online mode includes floor estimation and plane positioning processes. In this thesis, there are four methods in plane positioning: double matrix, weighted screening (WS), matrix WS and WS matrix. The study shows the WS is workable with acceptable accuracy.