| 摘要: | 手術導航系統現今已廣泛的使用在臨床上，可協助醫師更安全、更精確地完成手術，受到許多醫師及患者的青睞。由於手術導航系統需要精度較高的器材以及許多相關技術的整合，以保持其準確性與穩定性，因此通常整套系統要價不菲。若能將其中的硬體設備以其他較低廉的方案取代，便能大幅地降低其成本。
深度相機相較於傳統相機，除了可得知環境的影像之外，還可以直接獲得視野內物體的深度，如今已廣泛的應用在機器視覺領域，可搭載在自走車、機器手臂上等，許多廠商也推出消費級的深度相機，使得深度相機的使用門檻與成本降低了不少。而深度相機通常用於感測周遭事物的距離，因此若能使用深度相機取代導航系統中造價昂貴的感測器，用來得知目標物的位置，便能使成本大幅度的降低。
本研究之目的是以消費級的深度相機為基礎搭建一套器械追蹤系統。本研究結合彩色影像、紅外線影像、深度資訊等，透過影像處理與演算法來找出目標物的位置。本研究將深度相機與目標物固定在滑軌上，使滑軌移動特定距離，測量本研究之誤差。實驗包含測量單一球體球心的誤差、動態參考框架距離誤差和角度誤差，而各個項目又包含在不同距離所測得的誤差。擬和球心的誤差可在3mm以內，而動態參考框架的幾何相似度可達98%以上，角度誤差平均為2.74度。
;Surgical navigation system is now widely used in clinics. They can assist doctors in completing operations more safely and accurately. It is widely accepted by physicians and patients nowadays. Surgical navigation system requires high-precision equipment and the integration of many related technologies to maintain its accuracy and stability, so the entire system is usually expensive. Compared with traditional camera, depth camera can not only acquire the color image of the environment, but also obtain the depth of objects in the field of view directly. Nowadays, depth camera has been widely used in the field of machine vision. It can be mounted on automatic vehicles, robotic arms, etc. Many manufacturers have also introduced consumer-grade depth cameras, which lower the threshold and cost of depth cameras. Depth cameras are usually used to sense the distance of things around. Therefore, if the depth camera can be used to replace the expensive sensors in the navigation system, the cost of the system can be greatly reduced.
The purpose of this research is to build a spatial positioning system based on consumer-grade depth cameras. This research combines color images, infrared images, depth information, etc. to find the location of the target through image processing and algorithms. In this research, the depth camera and the target are fixed on the slide rail, so we can make the target move a specific distance and measure the error of the experiment. The experiment includes measuring the error of the sphere center, the distance error of the dynamic reference frame, and the angle error of the dynamic reference frame, and each item includes the error measured at different distances. The minimum error of the fitting sphere can be within 3 mm, and the geometric similarity of the dynamic reference frame can reach more than 98%, the angular error is 2.74 degree on average. |
