視訊資料中目標物追蹤及定位之研究

Object Tracking and Positioning on the Video Images

10.6574/JPRS.2006.11(2).7

陳敏新(Min-Hsin Chen)；陳繼藩(Chi-Farn Chen)

物件追蹤 ； 物件定位 ； 視訊影像 ； 形狀為主的物件匹配 ； Object tracking ； Positioning ； Video Image ； Color-based Segmentation ； Shape-based Matching

航測及遙測學刊

11卷2期（2006 / 06 / 01）

201 - 211

繁體中文

數位攝影機由於它的普及性以及相對於一般監視器的高畫質，使得數位攝影機逐漸變成具應用性質的監測工具。本研究提出了一個利用數位攝影機進行目標物追蹤及定位的應用。研究主要的目的是在攝影機裝置於一個移動載台的情況下，能夠自動的追蹤與定位一個靜止的物體。由於攝影機的拍攝的角度與位置會隨著時間改變，因此目標物在視訊影像上的形狀與大小也會隨著改變。本研究提出了一個以形狀為基礎的追蹤方法，透過Shape Matrix Algorithm (SMA)在位移，旋轉與尺度不變的特性，來找出畫面間物件的相似程度。當每個畫面的目標物皆被追蹤之後，就可以利用共線關係與最小二乘法進行目標物地面座標的求定。

This study presents an application of digital video camera for object tracking and positioning. The main purpose of the study is to automatically trace and position a motionless object on the video images that are recorded on a moving vehicle. Because of the change of the camera's focal length and the movement of the vehicle, by observation, the appearance of a motionless object on the video images firstly will gradually change its shape and completely lose its shape after a period of time. The variation of the shape apparently will bring about the complexity for developing the automated algorithm to trace and position the object on the video image. In this study, we develop an object-based matching technique to implement the tracking task. The technique uses shape matrix algorithm (SMA) that has scale and rotation invariant characteristics to calculate the similarity of variant shapes between adjacent images. The technique consists of three steps: (1) the segmentation step that is used to separate object and background on the video image; (2) the matching step that uses SMA to match the variant shapes between adjacent images and obtain the relative orientation parameters of the shapes; and (3) the tracking step that employs both Target Define Ground Operator (TDGO) and relative orientation parameters to trace the feature point of the object. After the object is traced in every video frame, the photogrammetric collinearity condition is used to transform the object from the image coordinates to the ground coordinates. An experiment is performed to trace a motionless ship in the open sea. The result shows that the proposed method can successfu1ly trace and position the ship even the ship had become entirely out of shape on the video images.