| 摘要: | 本實驗設計的空間編碼結構光，藉由偽隨機序列、絕對編碼方式，達到87的編碼數量，在1.6m橫向可解析最小距離為6mm，到3m橫向可解析最小距離約為1cm左右；在1.6m深度誤差為2mm，到3m深度誤差約為6mm左右。並將本實際的編碼圖樣模擬成繞射光學元件並投影在3m，針對投影光進行SNR、Contrast、MBR的可行性分析，在繞射光學元件尺寸為0.90x0.70mm，相位區分為四階，波長λ=840nm，投影距離d=3m，線寬L_width=225nm的情況下，SNR約為37，Contrast約為5，MBR約為27且該DOE遞迴傅立葉模擬達到理論極限MBR之93.48%。最後還與市面上Kinect v1作比較分析，從系統上來看，在3m下v1掃描之平面深度誤差厚度為25cm，本實驗架構為1.5cm。另外，發現v1資訊點密度為在2.75m時，共約有430個點資訊/m2，而本實驗的架構在2.75m時，約有3215個點資訊/m2；從編碼圖樣設計來說，在相同解析度和DOE尺寸的條件下，Kinect v1有較好的表現在DOE製成模擬，而透過低通模糊成像結果，觀察得到本實驗的編碼對於影像模糊有更好的容忍度。;In this study, we design a coded structured light pattern whose coding number is 87 by spatial neighborhood, absolute coding strategy and pseudorandom sequences.
First, lateral resolution of our system at 1.6m and 3m is 6mm and 1cm respectively, and deep Resolution at 1.6m and 3m is 2mm and 6mm respectively.
Then, we use this coding pattern performing diffractive optical element simulation projecting at 3m, and do the feasibility analysis of SNR, contrast and MBR. The SNR is 37, contrast is 5, and MBR is 27 which is 93.48% of MBR limitation under conditions that diffractive optical element size is 0.90x0.70mm, lambda is 840nm, projecting distance is 3m, line width is 225nm and divide phase into four orders.
Also, we compare our system with the system of Kinect v1, and have a result that the error at 3m of v1 is 25cm, while only 1.5cm in our system. By the way, the density of signal points of v1 and our system at 2.75m is 430 and 3215 points/m2 respectively. Last but not least, we compare our patterns with the patterns of Kinect v1, the patterns of Kinect v1 have higher contrast and MBR in diffractive optical element simulation, but our patterns have better tolerance for blurred image under the same signal points numbers and camera resolution. |
