Advances in devices and on-device/edge computing algorithm for infrared machine vision and infrared thermography: a review

Infrared machine vision (IRMV) and infrared thermography (IRT) enable non-contact imaging, temperature-field visualization, and intelligent target analysis, and have been widely applied in critical scenarios such as industrial inspection and aerospace engineering. With recent advances in infrared detectors, intelligent infrared cameras, and edge artificial intelligence, IRMV and IRT are evolving from conventional “imaging–offline processing” paradigms toward integrated “sensing–computing–decision” systems. However, existing reviews have mainly focused on individual devices, specific algorithms, or isolated applications, while a systematic synthesis of the co-evolution of infrared hardware, on-chip/in-sensor computing, and edge computing algorithms remains insufficient. To address this gap, this paper reviews recent progress in infrared optical lenses, uncooled and cooled infrared detectors, multifunctional infrared cameras, and infrared on-chip/in-sensor computing devices from a hardware–algorithm co-design perspective. Furthermore, on-device and edge-computing algorithms for real-time IRMV and IRT tasks are summarized, including infrared image denoising, enhancement, non-uniformity correction, super-resolution, and segmentation. Compared with existing infrared reviews, the distinctive contribution of this work lies in establishing a full-chain analytical framework covering “infrared devices–imaging and sensing systems–on-device and edge computing algorithms–representative applications”. This framework highlights the applicable scenarios, performance boundaries, and engineering bottlenecks of different technical routes. Finally, the deployment patterns and development trends of IRMV and IRT in typical application domains are summarized, and future research directions are discussed in terms of low-power array integration, hardware-friendly algorithms, standardized datasets, and reliable engineering validation. This review is expected to provide a systematic reference for the development of next-generation intelligent infrared sensing systems.