"Is it as I left it?": a computer vision aid for the blind

Blind people navigate safely through a familiar room based on strong expectations about the location of objects. If something has been moved, added or removed, it can present a difficulty and potentially a danger. This paper reports a novel prototype vision-based monitoring application developed as part of research into supportive home environments. It uses ceiling-mounted cameras to monitor rooms in the home for significant changes and automatically generates spoken warnings when appropriate. Significant object-level changes are detected and characterized based on fused region-level change detection modules. These use complementary pixel-level change detectors to obtain a degree of invariance to illumination changes. Results obtained using realistic video data are presented.