吊车危险区域侵入行人机器视觉定位算法  

Machine Visual Positioning Algorithm for Pedestrians Entering Dangerous Areas of Cranes

作　　者：石怀涛 李刚 范丽婷 刘建昌[2] 冯大阔 SHI Huai-tao;LI Gang;FAN Li-ting;LIU Jian-chang;FENG Da-kuo(School of Mechanical Engineering,Shenyang Jianzhu University,Shenyang 110168,China;College of Information Science and Engineering,Northeastern University,Shenyang 110819,China;Technology Center,China Construction Seventh Engineering Division.Corp.Ltd,Zhengzhou 450004,China)

机构地区：[1]沈阳建筑大学机械工程学院,辽宁沈阳110168 [2]东北大学信息科学与工程学院,辽宁沈阳110819 [3]中国建筑第七工程局有限公司技术中心,河南郑州450004

出　　处：《控制工程》2021年第4期759-765,共7页Control Engineering of China

基　　金：国家重点研发计划资助项目(2017YFC0703903);国家自然科学基金资助项目(51705341,51675353);东北大学轧制与自动化国家重点实验室开放基金资助项目(2018RALKFKT007)。

摘　　要：为了提高吊车施工的安全性能,并针对作业面行人目标像素较小不利于检测定位的问题,提出了一种吊车危险区域侵入行人机器视觉定位算法。首先,改进了用于目标识别的YOLOv3网络,更改了检测输入像素值,增加了小目标检测尺度并简化了不必要的网络结构,获得了更高的检测精度和更快的检测速度;然后,采用相机坐标变换的方法测算行人与危险区域中心点的距离;最后,通过真实样本数据集训练模型网络并测距来验证该算法的性能。实验结果表明,相比于其他算法,该算法在检测精度、检测速度和测距误差上均更优,具有更广的应用前景。In order to improve the safety performance of the crane construction, and to solve the problem of small pedestrian target pixels on the work surface, which is not conducive to detection and positioning, a machine visual positioning algorithm for pedestrians entering dangerous areas of cranes is proposed. Firstly, the YOLOv3 network for target recognition is improved, the detection input pixel value is changed, the detection scale of small targets is increased, the unnecessary network structure is simplified, and higher detection precision and faster detection speed are obtained. Then, the camera coordinate transformation method is used to detect the distance between the pedestrian and the center point of the dangerous area. Finally, the model network is trained on real sample data sets and the distance is measured to verify the performance of the proposed algorithm. The experimental results show that compared with other algorithms, the proposed algorithm is better in detection accuracy, detection speed and distance detection error, and has a wider application prospect.

关 键 词：YOLOv3 行人检测 视觉定位 吊车安全 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]