The synergistic interaction of vision and olfaction is critical for both natural and artificial intelligence systems to recognize and adapt to complex environments. However, current bioinspired systems with visual and olfactory sensations are mostly assembled with separate and heterogeneous sensors, inevitably leading to bulky systems and incompatible datasets. Here, we demonstrate on-chip integration of visual and olfactory sensations through a dual-focus imaging approach. By combining lens-based visual imaging and lensless colorimetric imaging, a target object and its odor fingerprint can be captured with a single complementary metal-oxide-semiconductor imager, and the obtained multimodal images are analyzed with a bionic learning architecture for information fusion and perception. To demonstrate the capability of this system, we adapted it to food detection and achieved 100% accuracy in identifying meat freshness and category with a 10 s sampling time. In addition to the highly integrated sensor design, our approach exhibits superior accuracy and efficiency in object recognition, providing a promising approach for robotic sensing and perception.
Keywords: bionic learning; dual-focus imaging; multimodal sensor; object recognition; vision−olfaction.