基於Kinect的盲人環境及障礙物偵測系統

視障人士因為無法利用眼睛去觀察周遭的環境，容易因為環境的改變或是物體的移動而造成碰撞產生危險，因此本論文將要探討基於Kinect的盲人環境及障礙物偵測系統，希望能夠透過此研究提供視障者在探索環境時，可以避開障礙物的盲人輔助系統。我們採用Kinect做為此研究的感應器，藉由其接收到的深度影像資訊和彩色影像資訊進行影像處理。本研究分成三個部分，第一部分是利用深度影像去做室內的樓梯偵測以及地面坑洞偵測。透過形態學的前處理去除掉深度影像的雜訊，之後使用Canny邊緣偵測和霍夫轉換取得直線的特徵，最後再設定樓梯和地面坑洞的條件，判斷出畫面中是否有這兩種場景。第二部分是藉由深度影像去做室內的障礙物偵測。同樣以形態學做為前處理，透過區域成長進行物件偵測以得到候選障礙物，再來基於四個原則選出符合條件的障礙物和其資訊。第三部分是彩色影像去做室外的障礙物偵測。先將擷取到的彩色影像做Peak-and-Valley濾波器的前處理，再透過Canny邊緣偵測取得垂直畫面的線，做為區域成長的種子點擺設。將偵測出來的物件經由條件的設定來過濾候選的障礙物，最後得到障礙物的資訊。透過實地拍攝多個不同環境與障礙物的場景進行測試，我們證實了此系統能夠有效的偵測到室內和室外的障礙物以及分辨出室內樓梯和室內地面坑洞的場景，再由震動器和語音來提示使用者障礙物、樓梯和地面坑洞的方向與距離。此系統一張影像所處理的時間約為0.4秒，可以達到即時視障輔具的需求。

關鍵字

障礙物偵測； Kinect Sensor ；盲人輔助系統；樓梯偵測

並列摘要

The visually impaired people lose their vision and are apt to be in danger when situated in unfamiliar environments or confronted with moving object. This study proposes an environment and obstacle detection system for the blind based on the Kinect sensors. This system aims to assist the visually impaired avoiding when they explore the environments. We used Kinect sensors to obtain the environment information. Digital image processing techniques were applied to process the color and depth images generated by the Kinect sensors. This research can be separated into three parts. In the first part, we processed the depth image to detect stair and concave ground in the indoor environment. Morphology preprocessors were used to eliminate noises in the depth images. Then we use Canny edge detection and Hough transform were employed to search for line patterns. Finally constraints were set to determine the appearance of these two scenes. The second part was obstacle detection in the indoor environment based on depth images. We used morphology as preprocessors to eliminate noise. Then we obtained obstacle candidates by using region growing. Finally four rules were used to determine if the candidates were real obstacles. The third part was obstacle detection in the outdoor environment based on color images. Peak-and-Valley filter was used as the preprocessor. Then Canny edge detection and Hough transform were used to detect the vertical lines for possible position of the seeds for the following region growing process. We obtained the obstacle candidate by using region growing. Finally principles were set to determine if the candidates were real obstacles or not. After tested in the scenes of different environments and obstacles, we demonstrated the capability of this system in detecting obstacles in the indoor and outdoor environments. It can also recognize the stair and concave ground scenes. The visually impaired is able to know the position and distance of the obstacles, stairs, and concave ground through the vibration module and sound alert. The processing time of this system is 0.4 sec per frame, a speed which is fast enough for real time assistant system for the blind.

並列關鍵字

Obstacle Detection ； Kinect Sensor ； Blind Aid System ； Stair Detection

參考文獻

[8] 蘇仕瑋, “以Kinect配合智慧型手機實做盲人輔具系統”, 國立成功大學工科科學學系碩士論文, 2013

[4] S. Gupta, P. Arbelaez, J. Malik, , “Perceptual Organization and Recognition of Indoor Scenes from RGB-D Images”, Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on, Page(s): 564 - 571

[6] W. Mark and D. Gavrila, “Real-Time Dense Stereo for Intelligent Vehicles”, IEEE Transactions on Intelligent Transportation Systems, VOL. 7, NO. 1, MARCH 2006

[7] J. Choi, D. Kim, H. Yoo, and K. Sohn, “Rear Obstacle Detection System Based on Depth From Kinect”, 2012 15th International IEEE Conference on Intelligent Transportation Systems, pp. 98-101, September, 2012

[9] S. Wang, Y. Tian, “Detecting stairs and pedestrian crosswalks for the blind by RGBD camera”, City Coll. of New York, 2012.

國際替代計量

基於Kinect的盲人環境及障礙物偵測系統

未授權

主題瀏覽