Due to the advancement of software and hardware technologies such as geographic information systems, for example, optical fiber storage, mobile network, digital camera, and high-definition network, it has promoted the establishment of a data digital management system. Network sharing, to a certain extent, makes it the smart city video sensor of today's era. It is a key component of the city's smart management system. In the context of the rapid development of the information age, I explore the role of video sensors in the visual design of smart cities. I propose three methods to analyze the theory of big data video sensors: theoretical analysis of big data video sensors, information visualization, and visual signal output, and then model the video signal output to visualize the complex multilayer network decision. Starting from the RBF neural network, I simulate a video perception prediction model system, import the data into Matlab for experimental analysis and simulation, and compare RBF and BP are based on both data predictions. RBF obviously has stronger predictive video perception. Experimental data show that the error between video perception and actual data is within 3%. It can be seen that video sensors play a key role in the construction of smart cities.