Object Detection and Recognition for Visually Impaired People

"Object detection plays a very important role in many applications such as image retrieval, surveillance, robot navigation, wayfinding, etc. In this thesis, we propose different approaches to detect indoor signage, stairs and pedestrians. In the first chapter we introduce some related work in this field. In the second chapter, we introduced a new method to detect the indoor signage to help blind people find their destination in unfamiliar environments. Our method first extracts the attended areas by using a saliency map. Then the signage is detected in the attended areas by using bipartite graph matching. The proposed method can handle multiple signage detection. Experimental results on our collected indoor signage dataset demonstrate the effectiveness and efficiency of our proposed method. Furthermore, saliency maps could eliminate the interference information and improve the accuracy of the detection results. In the third chapter, we present a novel camera-based approach to automatically detect and recognize restroom signage from surrounding environments. Our method first extracts the attended areas which may content signage based on shape detection. Then, Scale-Invariant Feature Transform (SIFT) is applied to extract local features in the detected attended areas. Finally, signage is detected and recognized as the regions with the SIFT matching scores larger than a threshold. The proposed method can handle multiple signage detection. Experimental results on our collected restroom signage dataset demonstrate the effectiveness and efficiency of our proposed method. In the fourth chapter, we develop a new framework to detect and recognize stairs and pedestrian crosswalks using a RGBD camera. Since both stairs and pedestrian crosswalks are featured by a group of parallel lines, we first apply Hough transform to extract the concurrent parallel lines based on the RGB channels. Then, the Depth channel is employed to further recognize pedestrian crosswalks, upstairs, and downstairs using support vector machine (SVM) classifiers. Furthermore, we estimate the distance between the camera and stairs for the blind users. The detection and recognition results on our collected dataset demonstrate that the effectiveness and efficiency of our proposed framework Keywords: Blind people, Navigation and wayfinding, Camera, Signage detection and recognition, Independent travel"

Name thesis:degree_name

Master of Arts (M.A.)

Level thesis:degree_level

Thesis

Year

2012

Author dc:creator

• Wang, Shuihua

• Blind People
• Object detection and recognition
• camera
• Engineering

Repository record dc:identifier

https://academicworks.cuny.edu/cc_etds_theses/96

OAI identifier oai:identifier

oai:academicworks.cuny.edu:cc_etds_theses-1095