本論文使用智慧型手機的相機鏡頭所拍到的畫面進行影子偵測和邊線偵測,用於行人行走中使用手機時,偵測前方的路是否有障礙物,並以警告通知。此方法特色在於只需利用單個鏡頭來判斷障礙物,達到在走路也能安全的使用手機。 本研究分為四個部分,首先是背景偵測,我們先將擷取影像的上方和下方比較,判斷是否有明顯變化,若有變化再往下進行影子偵測的障礙物判斷。第二部份是影子偵測,我們使用物體的影子來判斷,如果物體周圍有影子表示物體是有高度且危險的,但是考慮到並不是每一個情況下物體都會有影子,所以再使用邊線偵測來輔助。第三部份是邊線偵測,我們利用平面和立體物體邊線上的差異來判斷,以提升障礙物判斷的準確率。最後為實驗結果,我們將結果分為兩種,一種為只用影子偵測的結果,另一種為影子偵測加上邊線偵測輔助的結果。 本論文貢獻: 1.安全性:讓一邊行走一邊使用手機時不會因為撞到前方的障礙物而受 傷。 2.避障即時性:只利用單個鏡頭的圖片進行影像處理達到避障的效果, 這樣能提升辨識速度。 3.方法的創新:提出用影子和邊線來判斷障礙物,讓避障方法更多元。
In this thesis, we use smartphone camera lens capturing screen to do shadow detection and sideline detection, when using a smartphone for pedestrians, it can detect whether there is an obstacle in front of the road, and give a warning notice. Feature of this method is that we only use a single camera lens to determine the obstacles. It can achieve the goal of safety with using smartphone while walking. There are four parts in our thesis. The first part is background analysis. We compare photos of above and below, to determine whether there is significant change. If it changes, we use shadow detection to do obstacle determination. Second is Shadow Detection. We use the shadow of the object to determine if there is a shadow around the object to represent an object is high and dangerous. Considering the situation of the obstacles does not have shadows, we use sideline detection to assist the obstacles identification. Third is sideline detection. It is used to compare plane object sidelines and three-dimensional object sidelines. Finally part is simulation results. The results are divided into two types. The first one is only use shadow detection. Another one is shadow detection combining with sideline detection. In this thesis, we have some contributions as follows: 1.Safety: Let people use mobile phones while walking will not be injured by front obstacles hitting. 2.Immediate avoidance: Using only a single lens image to do image processing to improve recognition speed. 3.Innovative approaches: Propose the algorithms combining shadow and sideline detection to determine obstacles.