Dynamic recognition of objects labeled by visible light IDs is numerically investigated for obtaining a high optical signal-to-noise ratio in received optical signals when a relative motion between the objects and observer considerably alters. An averaging effect of pixel values in an image sensor shows a trade-off between reductions in accumulated image sensor noises and spatially varying noises originating from the relative motion. Our results suggest that a real-time optimization of the averaging process is suitable for an efficient recognition of visible light IDs toward applications for the disaster risk reduction.