Video surveillance is widely applied in modern intelligent systems, such as access control, pedestrian re-identification. However, with the rapid development of urbanization, urban traffic situation becomes more and more complex. It is a big challenge for video surveillance to cope with such massive data. In addition, existing emergency systems are far from modern requirement. So in this paper, we propose an image quality based framework to improve the performance of video surveillance and design a new urban intelligent emergency system. Specifically, we first analyze emergency evacuation of social group security incident for data acquisition including surveillance video and labels. Then, incorporating image quality assessment and convolution neural network, the dataset can be classified into several parts, and each part demonstrates a particular situation. Afterward, we introduce entropy theory to study the application of urban intelligence emergency. The results show that the research method proposed in this paper effectively obtains the evacuation parameters of evacuation personnel in the evacuation of sudden social group events, and improves the timeliness of information transmission in the evacuation process. The results show that the research method of this paper significantly improves the pertinence, effectiveness and perfection of the emergency plan in the application of urban emergency system.