Night Patrol Alarm Cart Based on PID Algorithm and YOLOv5 Algorithm

To address the issues of low target detection accuracy and poor path tracking stability in low-light environments for traditional electric patrol vehicles, this paper proposes a nighttime automatic patrol system that integrates YOLOv5 target detection with PID motion control. The system is based on the MaixCAM edge computing platform and uses the SE5 series AI acceleration chips to achieve hardware-accelerated inference of the YOLOv5s model (detection frame rate <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\geq 30$</tex> FPS, power consumption 3.8 W)<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3]</sup>. It addresses low-light imaging issues at 0.01 Lux illumination through an intelligent IRCUT dual-filter module and an 850 nm infrared illuminator. It introduces Kalman filter optimisation for target tracking to improve the accuracy of human image positioning in dynamic environments. It also uses PID algorithms to construct a path tracking closed-loop control system to compensate for the trajectory deviation of the three-wheeled chassis. Experiments show that the system has a target detection accuracy of over 95% at night. In sunny weather, it maintains 90 % accuracy in adverse weather conditions, has a path tracking error of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\pm 5 ~\text{cm}$</tex>, an alarm response time of less than 3 seconds, and a battery life of over 8 hours. This solution provides a low-power, high-robustness solution for nighttime fully autonomous patrols, driving the application of intelligent security technology in complex environments.