攝影機自我定位技術與估測物體空間座標方法之研究

Abstract

在這篇論文中，針對移動的目標物，我們設計並完成一組立體影像系統。為了達成這個目的，本系統使用兩部單眼取像設備(CCD)，先以次圖素演算法及影像處理技術完成取像設備自我定位的工作。在往後目標物的座標估測研究上，取像設備都一直保持著定位完成時的姿態。接下來，我們利用同時取得的左、右兩張影像，先偵測出目標物並決定其在影像平面中的座標位置。再者，運用三角定位法求得目標物於三度空間中的實際座標。因此，只要目標物在左、右兩張影像中的相關座標位置，我們可以藉由一組數學關係式，求得該目標物在三度空間中的座標。最後，我們分別在離目標物2公尺、2.5公尺、3公尺、3.5公尺及4公尺等不同的距離上，設計攝影機自我定位與估測物體空間座標的實驗，以證實本論文所推導數學關係式之正確性。

Abstract—A stereo vision system for a moving target is presented in this thesis. In order to achieve the goal, a pair of CCD cameras will be used to capture the images. Then, the system is calibrated firstly by subpixel algorithm and image processing technique. The attitude of the CCD cameras will be fixed when the system is used to estimate 3D position of the target. Next, using the images captured by the left and right CCD cameras, the position of target templet in the image plane will be detected and estimated. Furthermore, 3D coordinate of the moving target will be found by triangulation principle. Therefore, if 2D coordinates of the target templet in two image planes can be obtained, 3D coordinate of the target will be estimated by the algorithm described in the thesis. Finally, the experiments of camera self-calibration and estimating target position will be designed and implemented in the different distances (2m, 2.5m, 3m, 3.5m, and 4m) between the target and the system to prove the algorithm introduced in this thesis. 1.1 Survey............................................1-2 1.2 Motivation........................................1-5 1.3 Organization of the Thesis........................1-7 2. Image Process Algorithm and Stereo Imaging Technique..2-1 2.1 Image Process Algorithm...........................2-2 2.1.1 Image Thresholding..........................2-2 2.1.2 Image Calibration...........................2-4 2.1.3 The Perspective Projection Matrix...........2-5 2.1.1 Change of Coordinate Systems................2-7 2.2 Nonlinear Least-Squares Method....................2-12 2.3 Stereo Imaging....................................2-16 3. Implementation of the Stereo Vision System............3-1 3.1 System Overview...................................3-2 3.2 Cameras Self-Calibration..........................3-4 3.2.1 Cameras Model...............................3-5 3.2.2 Calibrating Cameras.........................3-6 3.2.3 Linear Methods for Estimating P.............3-7 3.2.4 Eigen Analysis..............................3-9 3.2.5 Nonlinear Optimizztion of P.................3-10 3.2.6 The Intrinsic and Extrinsic Parameters......3-10 3.3 Corner Detection..................................3-12 3.3.1 Edge Equations..............................3-13 3.3.2 Corner Idenfication.........................3-15 3.4 Locating 3D Coordinate of Moving Target...........3-18 3.4.1 Detecting Target Templet....................3-18 3.4.2 Calculating 3D Coordinates of Target........3-20 4. Experiments and Results...............................4-1 4.1 Experiments of Camera Self-Calibration............4-2 4.1.1 Results and Analyses of Camera Self-Calibration .............................4-3 4.2 Experiments of Estimating Target Position.........4-10 4.2.1 Communicating with PC.......................4-12 4.2.2 Control Circuit.............................4-13 5. Conclusion............................................5-1