Purpose:
A wearable-head-mounted system that shows information about the proximity of objects by color and thickness of the depicted contour has been developed and tested.
Methods:
In this work, a Trivisio ARvision-3D HMD (Head mounted display) has been used. This HMD has integrated two cameras with 752x480 pixels and projects in two 800x600 LCD displays. A proprietary algorithm has been developed to perform a contour map of the scenario. The algorithm is based in the following sequence: contour detection, depth algorithm computation, and assignation of color or thickness to the plotted contours according to distance from the HMD to the objects detected. Contour detection is performed first because in order to accelerate the depth computation this is only processed in the vicinity of contour regions. Disparity maps are generated with the stereoscopic correlation between both images. A laptop with Intel Core i5 processor, 2GB RAM memory and three USBs and VGA video output, necessary to connect to ARvision-3D HMD, is used to implement the aid.
Results:
Processing times of 11 to 15 frames per second are achieved. A graphic interface is also implemented for the Visual Rehabilitation Professional to control the different parameters to apply in the algorithm, such as color, thickness and their related calibrated distances to objects.
Conclusions:
The system demonstrates that two cameras and a HMD can be used to achieve a wearable vision aid, in such a way that the obstacles could be detected easily on real time. The application also shows information about the proximity of objects by color and thickness of the depicted contour. Future trials include clinical tests.
Keywords: binocular vision/stereopsis • detection • low vision