清洁车智能监测与控制系统研究
2022,30(5):109-114
摘要：针对目前市面上的清洁车容易因为驾驶员的疏忽造成清扫盘与路沿的碰撞，加速清扫盘的损坏，洒水作业容易波及路人，贴地工作的吸盘容易与减速带碰撞造成损失，不能自动根据垃圾量调节清扫功率等问题，提出了一种清洁车智能监测和控制系统。该系统采用超声测距模块避免碰撞，运用YOLOv4算法检测行人和减速带，使用VGG16网络进行垃圾量化进而调节清扫功率，通过CAN通信模块实现对清洁车设备的实时控制。该系统采用多线程实现多任务算法之间的协调与通信。行车实验表明，该系统的整体运行速度为每秒6-7帧，满足实时性。路沿距离检测及避障算法准确率为96%，垃圾量化算法准确率为90%，行人和减速带算法准确率为96.25%，准确率较高，能够有效提高清扫效率，降低设备损耗，提高行车安全。
关键词：清洁车;智能监测 ;超声传感器;YOLOv4;VGG16;多线程;CAN通信
Research on intelligent monitoring and control system of cleaning vehicle
Abstract：Aiming at the problems that the cleaning vehicle on the market is easy to collide with the roadside due to the negligence of the driver, accelerate the damage of the cleaning plate, the watering operation is easy to affect passers-by, the suction cup working close to the ground is easy to collide with the deceleration belt, resulting in losses, and the cleaning power can not be automatically adjusted according to the amount of garbage, an intelligent monitoring and control system for the cleaning vehicle is proposed. The system uses ultrasonic ranging module to avoid collision, uses YOLOv4 algorithm to detect pedestrians and speed bumps, uses VGG16 network to quantify garbage and adjust cleaning power, and realizes real-time control of cleaning vehicle equipment through Controller Area Network (CAN) communication module. The driving experiment shows that the overall running speed of the system is 6-7 frames per second, which meets the real-time performance. The accuracy of roadside distance detection and obstacle avoidance algorithm is 96%, the accuracy of garbage quantification algorithm is 90%, and the accuracy of pedestrian and deceleration belt algorithm is 96.25%. The accuracy is high, which can effectively improve the cleaning efficiency, reduce equipment loss and improve driving safety.
Key words：cleaning vehicle; intelligent monitoring; ultrasonic sensor; YOLOv4 ; VGG16 ;multithreading; CAN communication
收稿日期：2021-11-09
基金项目：
