Multispectral imaging technology has made significant progress in photodetection due to its high-dimensional information. This progress is driven by two factors: innovations in multispectral technology combined with the development of interdisciplinary fields such as materials science, nanotechnology, and artificial intelligence, which have transformed multispectral technology from traditional to novel detection techniques. The spectral splitting methods of novel detectors have transitioned from macro-scale to micro-scale, leading to enhanced performance that surpasses traditional systems. The demand for miniaturized, lightweight detectors has further accelerated the development of on-chip multispectral photodetectors. Recent innovations in high-integration and high-sensitivity methods and materials have laid a solid foundation for future on-chip multispectral technologies. It is clear that multispectral photodetectors will evolve from large-scale systems to on-chip designs, highlighting the significance of this survey on multispectral photodetectors. First, the spectral splitting methods of multispectral photodetectors are introduced, which can be categorized into three types: dispersion-based, filter-based, and interference-based methods. These methods typically rely on large discrete optical components, such as gratings, triangular prisms, large-aperture lenses, and multiple sets of filters. However, these components often suffer from poor stability and are difficult to integrate, which increases the overall volume and complexity of the system. As a result, such designs deviate from the current trend toward miniaturization in multispectral photodetection. In contrast, the new generation of on-chip multispectral detectors offers greater flexibility in light splitting. These detectors either utilize ultra-thin or pixel-based splitting surface arrays that can be directly integrated with the detector or employ innovative optoelectronic materials that simultaneously function as both photosensitive and spectroscopic components. This approach enables the integrated design and manufacturing of multispectral detectors, significantly advancing the field.After introducing the principles of multispectral detection, the development progress of novel on-chip multispectral detectors is detailed. Semiconductor nanowires-based and active nanoantenna-based multispectral detectors can achieve wavelength-selective and efficient detection from the visible light to the infrared region by simply adjusting the diameter of the nanowires/nanopillars, making them ideal candidates for photodetection across a wide spectral range from ultraviolet to infrared. Quantum dot-based multispectral detectors effectively compensate for the deficiencies in the infrared wavelength range of multispectral detection. By reasonably designing the photoresponse range, they can almost cover the entire infrared range from near infrared to long-wave infrared. However, the uniformity and stability of quantum dots are challenges that need to be urgently overcome. Two-dimensional material-based multispectral detectors take advantage of the atomic thin-layer characteristics of two-dimensional materials, which can achieve a high detectivity, but the impact of filter usage on system complexity cannot be completely eliminated. Perovskite-based multispectral detectors benefit from the flexibility and designability brought by the octahedral symmetry structure of perovskites. Their photoresponse range can be flexibly tuned from the ultraviolet to the near-infrared range. However, their lack of stability and the presence of toxic lead currently hinder the commercialization and large-scale application of perovskite materials.The primary application fields for multispectral detectors include remote sensing satellites, drones, and airborne platforms, and the application scenarios are highly diverse. They could be utilized not only in natural environments, such as estimating cloud top height, detecting thick cloud positions, inverting aerosol optical thickness, identifying land features, measuring rice nitrogen content, detecting floods, inverting ocean depth, and identifying marine plastic waste, but also in human-related fields. These include assisting in skin cancer screening and treatment, monitoring diabetic retinopathy, assessing human stress levels, aiding in the diagnosis of psychological conditions such as depression and anxiety, detecting fire points, inspecting food quality, protecting natural heritage, and supporting archaeological research.At the conclusion of this paper, the main challenges multispectral photodetectors are facing are summarized, and several recommendations are proposed. These insights aim to provide valuable guidelines for the development and research of multispectral photodetectors in China. The field of multispectral imaging photodetectors is transitioning from traditional to emerging technologies. Traditional detectors provide excellent stability and reliability, establishing a strong foundation for commercialization. However, their complexity and sensitivity limitations constrain performance enhancement and application expansion. In contrast, new detectors, which incorporate innovative methods and materials, are driving revolutionary advancements. Currently, both material selection and device fabrication are in the early stages of exploration and optimization, requiring more efforts to accelerate the development of these technologies. Despite ongoing technological and cost challenges, multispectral detectors show significant potential. As global demand for multispectral imaging continues to rise, the development of high-precision, compact, fast-response, and high-resolution detectors has become a key area of research.