集装箱正面吊人机安全防撞行人智能检测算法研究

Intelligent Pedestrian Detection Algorithm for Human-Machine Safety and Collision Avoidance of Container Reach Stacker

针对集装箱装卸作业环境行人小目标检测精度较低以及行人检测易受背景干扰问题,提出了一种融合行人掩模的多尺度人机安全防撞行人智能检测模型。该模型以VGG-16作为主干网络,通过构建行人注意力模块和特征融合模块,将像素级行人注意力掩模与主干网络特征进行特征融合,增强小尺度行人的特征识别,显著减少了背景干扰,并将行人智能检测信息作为参数,提出基于标定法的人机距离测量方法,设计集装箱正面吊自动刹车执行机构,研发集装箱正面吊人机安全防撞行人智能预警系统,实际应用验证了集装箱正面吊人机安全防撞行人智能检测算法的有效性和可靠性,满足铁路货场行人小目标实时准确检测要求,能有效防止集装箱正面吊与行人相撞,避免人身伤亡事故发生。

In view of the low accuracy of small pedestrian target detection in container loading and unloading operation environments and the susceptibility of pedestrian detection to background interference,an intelligent pedestrian detection model for multi-scale human-machine safety and collision avoidance integrating pedestrian masks was proposed.This model used VGG-16 as the backbone network and constructed pedestrian attention modules and feature fusion modules to fuse pixel-level pedestrian attention masks with the backbone network features.As a result,it enhanced the feature recognition of small-scale pedestrians,significantly reducing background interference.Intelligent pedestrian detection information was used as a parameter,and a human-machine distance measurement method based on the calibration method was proposed.The automatic brake actuator of the container reach stacker was designed,and an intelligent pedestrian warning system for human-machine safety and collision avoidance of the container reach stacker was developed.The practical application verified the effectiveness and reliability of the intelligent pedestrian detection model for human-machine safety and collision avoidance of container reach stackers,which met the real-time and accurate detection requirements of small pedestrian targets in railway freight yards,effectively prevented collisions between container reach stackers and pedestrians,and avoided personal injury accidents.