Smart streetlight based IoT system is the best carrier to accelerate the development of smart cities. Generally, the system architecture takes the ‘cloud’, ‘tube’, and ‘end’ centralized structure, easily causing computing resources centralizing, data transmission delay, and network congestion. In this paper, the decentralized edge structure is applied to replace the traditional centralized structure and hence overcome its deficiencies. Firstly, the embedded hardware platform on streetlight edge-side is virtualized to obtain a unified resource pool and two individual virtual machines. Secondly, the two virtual machines are used to establish an edge intelligence management platform and an edge computing service system, respectively. Finally, brightness and color temperature control algorithms are designed by utilizing fuzzy control technology, and are calculated on the obtained computing service system. Experimental results show that our structure and system can efficiently achieve the edge intelligent computing task of brightness and color temperature.