Improving visual inspection using binocular rivalry

To design inspection systems for modern manufacturing conditions, an appropriate mix of human and computer abilities is required. Human inspectors are not very effective at the Search function of inspection, meaning that search tasks are more suited to automated vision systems. We assessed a novel visual technique-binocular rivalry arising from stereo viewing conditions-to enhance human search performance. A sequence of three inspection experiments was used to determine whether binocular rivalry is practical and effective. An initial study using circuit board images showed a small but significant effect of binocular fusion. A second study using simulated material with excellent registration between the left and right images showed binocular rivalry to increase performance considerably. Binocular rivalry also removed any negative effect of inspecting for multiple defects simultaneously. The third study returned to realistic inspected items, but ensured good registration by placing images on a computer screen. This study again showed the superiority of binocular rivalry. There was no unusual visual strain from stereo viewing. Together, these studies allow us to delineate the conditions under which the benefits of binocular rivalry can be exploited in quality control practice.