【摘要】 针对运行中的风力机叶片，提出一种基于机器视觉特征分类的故障诊断方法。通过对叶片叶尖进行圆形标记，利用工业相机周期性获取叶片尖端的图像，并在Halcon软件上对图像进行预处理，对大雾天气下采集的图像利用暗通道除雾算法进行清晰化处理。利用叶尖标记检测算法提取标记、计算区域圆度和区域中心等区域特征。对相邻叶片上标记计算位移差，并与系统预警阈值比较，判断叶片在扭转或偏摆方向的变形程度和故障趋势，从而实现风力机叶片变工况运行状态在线检测和自适应预警。更多还原

【Abstract】 A fault diagnosis method based on machine vision feature classification is proposed for wind turbine blades in operation. By circularly marking the blade tip,images of the blade tip are acquired periodically using an industrial camera and pre-processed on Halcon software,and images acquired in foggy weather are clarified using a dark channel defogging algorithm. A leaf tip marker detection algorithm is used to extract markers,calculation area features such as area roundness and area centre. The markers on the adjacent blades are then compared with the system’s warning threshold to determine the degree of blade deformation and fault trend in the direction of torsion or deflection,thus enabling online detection and adaptive warning of the variable operating conditions of wind turbine blades.更多还原