基於深度與彩色資訊之互動式數位積木遊戲

本論文針對學齡前的孩童設計出一套結合實體物體與虛擬情境的親子互動裝置，讓孩童可以方便地透過日常生活中唾手可得的物品，堆疊組合成各式複合形狀。本論文使用體感器取得深度影像與彩色影像，先透過深度影像進行空間中物體的偵測與切割(Object Detection and Segmentation)，並取得物體的外型，在形狀的描述以邊界點作為特徵，再以形狀上下文(Shape Context)演算法進行辨識。另外，再擷取物體的彩色資訊之尺度及旋轉不變SURF(Speeded-Up Robust Features)特徵點，進行商標圖案與文字的辨識。最後將辨識結果從關聯表找到事先對應的關聯圖，將對應的關聯圖依照實體物體的長寬大小與在空間中的位置合成於圖畫上。測試結果可知具有實用性。

In this paper, we implement a parent-child game for a whole family by using a sensor has both the depth and the RGB camera. This interactive game will use the objects the parents can get everywhere in the family as the materials. This game is designed by extracting the shapes and colors of the material objects using several image processing segmentation methods, and matching with the templates in the predefined database. Once the shape of object has been classified, a corresponding associated image will be selected and composited with background and other image layers, according to the current playing scenario. This composite image can provide a lovely virtual scenario for the parents and children to let them tell any stories as they want, and also let them immerse into the joy of this game scenario.