Real-time Small-size Pixel Target Perception Algorithm Based on Embedded System for Smart City

With the continuous development of technology in the artificial intelligence era, smart city applications based on high-performance servers in large data centers have penetrated all walks of life. However, the current mainstream smart city application model is only data collection on the device side, and then calculations and inferences in the data center. Data transmission is difficult to achieve the real-time performance of the system, resulting in poor effects in many smart city applications. The intelligent perception for smart city is required to perceive the whole urban environment comprehensively. Among them, small-size pixel target detection and recognition is particularly critical. To this end, a real-time small-size pixel target perception algorithm based on embedded system for smart city is proposed, which uses lightweight neural networks and model pruning optimization to realize terminal intelligence for smart city applications, and integrates traditional machine learning filtering algorithms for improving the detection speed and the accuracy. The experimental results of the method show that the real-time performance and the accuracy of detection are greatly improved for different sizes of small-size pixel targets.