Pedestrian Traffic Light Recognition for the Visually Impaired

In this research, we employed the technology of computer vision to recognize traffic light. We manipulated the color information of traffic light under HSI color space. And then, adding motor tracking technology, we can trap the traffic light which is subjecting to region of interesting (ROI) to the center area of monitor. Meanwhile, we assign different frequency (peach) according to red light or green light to inform the people what color the light is.
We developed the algorithm at PC via MATLAB with powerful image process toolbox. Then, we ported the whole system to DSP platform made by Texas Instrument, TMS320 DM642 EVM. The result is that in the condition of non-complex environment, the system can distinguish the red light from green light, and also can output audio signal by means of speaker or headphone.