利用多隻相機之電腦視覺與自動車航行技術作建築物走廊之安全巡邏

Abstract

本論文中提出了一套基於電腦視覺並且利用自動車航行來作建築物走廊安全巡邏的有效方法，而航行與巡邏的工具則是利用一部架有多隻相機的車子。透過無線網路，車子的位置資訊與環境的狀況可以在監控端與車子之間傳輸。關於航行的方法，本論文亦提出兩種航行模式，包括無線網路遙控與自動導航。為了達到巡邏的需求，車子於航行時能偵測非預期出現的物體，並且利用形狀比對、電腦視覺與連續影像的技術來判斷偵測到的物體為靜態或是動態。對於靜態的物體，我們利用避開障礙物的操作來作無碰撞之航行；對於動態的物體，則利用物體追蹤的方式來跟蹤此移動之物體。此外，對於建築物中的安全而言，門的開啟與存在物體之消失與否也是必須要偵測的事件。一但發現門意外的被開啟或者是原本存在的物體已經消失，系統即將照得之環境影像壓縮後透過無線網路傳送至監控端。最後，良好的實驗結果也證實了本論文所提出方法與系統之可行性。

An effective vision-based approach to security patrolling in building corridors using vehicle navigation is proposed in this study. A commercial vehicle mounted with multi-cameras is designed to navigate and patrol in building corridors. The information of the location of the vehicle and the environment conditions can be transmitted to a security monitor site through the wireless network. Two modes of vehicle guidance, remote control and automatic guidance, are proposed for use in navigation session. During navigation, the vehicle system detects unexpected objects. A distance-weighted correlation measure, computer vision, and image sequence techniques, are used to discriminate between stationary and moving objects. An obstacle avoidance process is performed to achieve collision-free navigation. An object tracking process is performed to follow moving objects. In addition, door opening and disappearance of existing objects can be detected, and the environment images are then compressed and transmitted to the monitor site through the wireless network. Experimental results show the feasibility of the proposed approach and system.