技術簡介
使用YoloV4深度網路架構,透過AMR上搭載之攝影機,偵測路徑上門扉開閉狀態
Abstract
Door open / close status decision module include two phases. In the first phase, part of door is detected by YoloV4. In the second phase, the door open / close status is concluded by the features extracted from the part of door detected in the first phase.
技術規格
每次偵測花費0.11秒,正確率97.14%
Technical Specification
The proposed module detection the door open / close status in 0.11sec per image, and achieve accuracy 97.14%.
技術特色
針對部份門扉之特性,以光達難以辨別門扉之開閉,導至AMR移動可能碰撞。透過YoloV4以偵側方式偵測門扉特定位置,搭配規則式判斷,以達到有效門扉狀態偵測
應用範圍
搭載RGB攝影機之移動載具
接受技術者具備基礎建議(設備)
Windows 10 x64 PC
接受技術者具備基礎建議(專業)
資工相關背景
聯絡資訊
聯絡人:周茂良 數位轉型技術組
電話:+886-6-3847128 或 Email:norman@itri.org.tw
客服專線:+886-800-45-8899
傳真:-