利用可攜式眼鏡型微攝影機輔助視障人士即時識別公車車號

Helping the blind to identify city bus numbers with the mobile eyewear

吳柏翰(Bo-Han Wu)

指導教授 ： 葉榮木 ； 蔡俊明

國立臺灣師範大學/科技學院/機電科技研究所/碩士(2011年)

摘要

視障人士搭乘大眾交通工具(公車)時，面臨許多難題，其中最主要的問題就是無法得知迎面而來的公車車號。目前視障人士能解決的辦法，不外乎是請求旁人協助或手拿自製的公車車號板讓駕駛注意，但上述辦法皆不穩定且變動因素甚大。因此，基於影像處理技術的蓬勃發展，本研究改變以往只採用固定式攝影機處理的方式，利用“可攜式眼鏡型微攝影機”，在皆非固定的情況下(例如：背景、角度、車號等等)，輔助視障人士即時識別公車車號，並以語音輸出告知其資訊。本研究採用主動搜尋與辨識，在不降低準確率的情況下提升系統整體速度。透過事前的分析歸納，直接擷取出輸入影像中感興趣的顏色範圍，並將其轉為二值化影像降低其資料量，再搭配設計的形態學遮罩來確保公車車號的完整性，且透過連通成份分析並挖取出車號區域，送入 MS MODI 做辨識，最後藉由 MS SAPI 在公車停靠前以語音的方式輸出。

關鍵字

移動物偵測 ； 前景擷取 ； 光學字元辨識 ； MS SAPI

並列摘要

Blind people face many problems when they take bus. The major problem is to recognize an approaching city bus. The solution to this torment at present is to ask others’ assistance or to raise a sigh on which the destination is written for appealing bus drivers’ attention. But there exists much unreliability in fore-mentioned solutions. Because image processing is a highly developed research area, in this research, we adopt a mobile eyewear digital video to replace a traditional fixed one to help the blind recognize the bus numbers with vocal message. This research adopts proactive identification for fast response without harm to accuracy. Through beforehand analysis, a block with wanted colors is focused from input video. Then this data proceed with an adaptive binarization method for simplification and Morphology mask for integrity. Bus numbers can be obtained from the analysis with connected component and recognition of MS MODI. A vocal message will be launched with MS SAPI before bus stops in the end.

並列關鍵字

Motion detection ； OCR ； MS SAPI

參考文獻

[2] F. Kamalabadi, “Multidimensional image reconstruction in astronomy,” IEEE Signal Processing Magazine, Vol. 27, Issue: 4, pp. 86-96, 2010.

[3] S.H. Thorne, C.H. Contag, “Using in vivo bioluminescence imaging to shed light on cancer biology,” Proceedings of the IEEE, Vol. 93, Issue: 4, pp. 750-762, 2005.

[4] D. Garcia, J.C. del Álamo, D. Tanné et al., “Two-dimensional intraventricular flow mapping by digital processing conventional color-doppler echocardiography images,” IEEE Transactions on Medical Imaging, Vol. 29, Issue: 10, pp. 1701-1713, 2010.

[5] N.M. Garcia-Aracil, J.M. Azorin Poveda, J.M. Sabater Navarro et al., “Visual Control of robots with changes of visibility in image features,” IEEE Latin America Transactions, Vol. 4, Issue: 1, pp. 27-33, 2006.

[6] H. Zhang, “Image processing for the oil sands mining industry,” IEEE Signal Processing Magazine, Vol. 25, Issue: 6, pp. 198-200, 2008.