隨著定向定位系統與蒐集空間資訊的遙測感測器結合，逐漸的實現了即時性的移動式測量及空間資料蒐集平台－行動測繪系統，而在定向定位系統與影像資料的相互整合中，直接地理定位法是為關鍵技術。 本研究建置一雙鏡頭、高解析度之車載行動測繪系統，以整合GPS、IMU及兩台數位相機，利用直接地理定位法獲取影像的外方位參數，最後對房屋的牆面點進行定位分析。在此三種感測器的結合中有兩種需要率定的系統轉換關係，分別為GPS與數位相機透視中心間的固定臂位移向量及IMU與數位相機三維坐標軸的軸角旋轉量。 研究中使用的相機焦距為28mm，與牆面距離約10到20公尺，影像的物空間解析度約為0.2到0.4公分，且為提升三維幾何定位精度，兩台相機拍攝之交會角約為60度。實驗中共選擇來自不同相機的十七張影像進行固定臂與軸角的率定，並利用率定結果推估影像外方位參數，最後定位分析的成果可分為兩部分：（一）、將物空間牆面點反投影至像空間，像空間之定位精度可達10個像元以內；（二）、將像空間點進行空間前方交會，檢核點的物空間定位精度可達10公分以內。 With the combination of position and orientation system (POS) and remote sensors, a real-time measurement and spatial data collection platforms¬ which called Mobile Mapping System was realized. The integration of position and orientation system and image data, direct georeferencing is the key technology. This paper proposes to design a terrestrial mobile mapping system of low-cost, multiple cameras, high-resolution. The exterior orientation parameters (EOP) of image directly provides by the principle of direct georeferencing and GPS/IMU data. The final results will show the position analysis of the mobile mapping system. There are two conversion relations between GPS, IMU and digital cameras. First, a position vector between GPS and digital camera called Lever arm Vector. Second, the angle between the rotation axis of IMU and digital camera called Boresight Angle. There will calibrate the conversion relation of the mobile mapping system in this paper. The focal length of the camera which used in this research was 28mm and the subject distance is about 10 ~ 20 meters which result the image spatial resolution of 0.2 ~ 0.4 cm for a pixel. For the purpose of increasing its 3D positioning accuracy, the intersection angle of these two cameras is about 60 degrees. In the experiment, 17 scenes were selected from different camera to calculate Borsight and Lever arm calibration, and used a total of 53 ground control points which on the wall were utilized to evaluate the potential of its 3D positioning accuracy. Positioning analysis of the results can be divided into two parts. First, back-projection the point from the object space to the image space, the positioning accuracy in image space can be less than 10 pixels. Second, the positioning accuracy in object space using space intersection can be less than 10 cm.
