Single-photon avalanche diodes are widely used in various fields because of their single-photon sensitivity and excellent time-resolved performance. With the development of semiconductor technology, single-photon avalanche diode arrays integrating multiple pixels and time measurement circuits are gradually popularized, with the ability to collect photon information in parallel. Imaging is a means of obtaining information of the target object through photons, and imaging systems based on single-photon avalanche diodes can use richer photon counts and temporal information to detect the target in extreme environments. Single-photon avalanche diode arrays have higher detection efficiency, which can replace the detection system of single-pixel detectors and scanning structures, promoting the advancement of biological microscopy, scattering imaging and lidar technologies. This manuscript concludes the development of single-photon avalanche diode arrays and introduces some typical applications of single-photon avalanche diode arrays in imaging. The development of SPAD is similar to other photodetectors, which have gone through the process from single-point detectors to multi-pixel arrays. Because of the application of CMOS technology, SPAD arrays have developed rapidly in terms of pixel scale and circuit integration. Megapixel SPAD arrays with time measurement capabilities are available nowadays. While the pixel scale is gradually increasing, important parameters such as photon detection efficiency, dark count rate, spectral response range, and temporal resolution are also continuously optimized with the development of related technologies. Optical imaging has a long history. With the development of science and technology, people's research interests have gradually expanded from traditional imaging to imaging under extreme conditions, such as super-resolution imaging, extremely low-light imaging, and over-the-horizon imaging. SPAD has single-photon sensitivity and ps-level temporal resolution, which enables obtaining photon information under extreme conditions. As the performance of early SPAD arrays were not perfect, single-pixel SPAD detectors are often combined with a scanning device to obtain two-dimensional images. With the development of SPAD arrays, in applications that require real-time imaging, such as vehicle-based lidar, SPAD arrays have gradually replaced the scanning imaging systems because of their efficient parallel single-photon detection capabilities, providing higher imaging speed. Besides, many biophotonics applications have been explored with SPAD arrays, such as SMLM and FLIM. The use of SPAD arrays in these applications enables higher SNR, higher imaging speed, providing powerful method to investigate the structural details and molecular dynamics of cells. In addition, the high dynamic range images are accessible through a SPAD array, which has the potential to be applied in autonomous driving and object recognition. In scattering imaging and non-line-of-sight imaging, the emergence of SPAD arrays enables the complex photon propagation process caused by multipath, scattering or other factors to be distinguished in time domain and space domain, and the photon information can be combined with physical model or neural network to detect the target object which is outside the field of view or behind the scattering medium. Using the high temporal resolution and parallel acquisition capability of SPAD arrays, one can also track the high-speed laser pulses, providing more details of ultrafast optical phenomena. The low price and high integration of SPAD arrays are unmatched by other devices. In the future, if the cost of megapixel SPAD arrays can be reduced to a reasonable range, they will be widely used in scientific research, industry and military fields. However, the reported megapixel SPAD array is still in the laboratory verification stage, and there is still a long distance from the industrialization and commercialization. In addition, with the continuous increase of the pixel scale of SPAD arrays, the storage, processing and transmission of photonic data will be a difficult problem. Using high-performance FPGA to locally preprocess the photonic data can effectively reduce the demand for data storage, reducing the data transfer bandwidth between SPAD arrays and computers. In terms of spectral response, the spectral response peaks of silicon-based SPAD arrays are mainly within the visible light band. The photon detection probability of silicon-based SPAD arrays in the near-infrared band can be improved through structure and process optimization, enabling utilizing the strong penetration of near-infrared light to improve the detection range of lidar and scattering imaging. SPAD arrays based on InGaAs or InP can respond to short-wave infrared light above 1450 nm, so such SPAD arrays have great application potential in imaging with optical fibers and quantum optics. SPAD arrays can efficiently acquire spatiotemporal information of photons, and how to make full use of these photonic data is also a problem that needs to be solved in the application of SPAD arrays. In addition to establishing the physical model of photon propagation, data-driven algorithms such as deep learning can also be used to establish a correspondence between the spatiotemporal distribution of photon and target features from a higher dimension, enabling efficiently reveal the hidden information in the photonic data. In the future, SPAD arrays will play a more important role in optical imaging, providing more powerful tools to perceive and understand the world.