Ultrafast diagnosis technologies are primarily employed to acquire images and information from ultrafast phenomena in the fields of physics, optics, and materials at nanosecond even femtosecond scale. Compressed Ultrafast Photography (CUP) is a major breakthrough in the field of ultrafast diagnosis, which is one of the cutting-edge technologies to acquire information efficiently in ultrafast dynamic events.According to Nyquist Theorem, the sampling frequency must be at least twice the frequency of the original signal to ensure that the sampled signal can accurately reconstruct the original signal. But in the condition where the Compressed Sensing (CS) theory is applicable, when the signal satisfies sparsity, the original signal can be reconstructed with high quality in conjunction with a priori knowledge, even when the sampling rate is significantly lower than that required by Nyquist Theorem. The streak camera is one of the major devices for ultrafast diagnosis with ultra-high temporal resolution which is at picosecond scale. The streak camera converts the optical signal into a photoelectron beam through the photocathode. Whereafter, the streak camera deflects the photoelectron beam with a time-varying deflecting electric field, and the imaging position of photoelectron beam changes over time. Therefore, the streak camera is able to realize time-resolved images of the photoelectron beam with high temporal resolution. In order to avoid overlapping of images of different moments, a slit is added to limit the spatial range of the optical signal, thus the streak camera is only able to realize one-dimensional imaging in general. However, since CUP combining CS theory with streak camera, it has both high temporal resolution and two-dimensional spatial-resolved capability. Consequently, CUP is able to realize time-resolved two-dimensional spatial diagnosis of ultrafast events with one single shot.In this paper, a CUP system is designed and presented. Two-dimensional ultrafast diagnosis experiment is simulated, and the experimental platform of CUP system is built to diagnose the intensity evolution process of the picosecond laser pulse in both spatial and time dimensions. The CUP system is mainly composed of streak camera, DMD, optical path system and synchronous modulation module. The system first uses DMD to encode the original ultrafast image I(x, y, t). Then the encoded image transmits to the streak camera through fully opened slit. After being deflected by electric field, images of different moments are overlapped, and finally the overlapped integrated image is sampled by the CCD. In order to reconstruct the image sampled by the CCD, it is necessary to reverse solve I (x, y, t). However, this is an underdetermined problem, which can not be solved by direct operation. According to CS theory, this underdetermined problem can be transformed into an unconstrained optimization problem by regularization method based on the principle of Total Variation (TV) and the prior knowledge of the DMD coding matrix. Once CS theory was applied, each frame of the original images of different moments can be reconstructed. According to the working principle and mathematical model of the CUP system, the sampling and reconstruction processes are simulated. The peak signal-to-noise ratio of the reconstructed images is better than 45 dB, and the structural similarity is better than 99%. The intensity evolution process of the picosecond laser pulse is diagnosed by the experimental platform of CUP system. To measure the accuracy rate of the image reconstruction of the CUP system, the intensity evolution process of the picosecond laser spot is measured by the streak camera in one-dimensional spatial imaging. The accuracy rate of the reconstructed signal images of the system is 96.06%, demonstrating its ability to accurately diagnose ultrafast dynamic events.