论文部分内容阅读
针对汽车喷油器接杆的激光环焊焊缝图像,提出了一种基于边界方向直方图的缺陷检测方法。通过设定门限的方法对待测图像进行分割,再对分割后的图像提取边界点;采用边界方向直方图表示目标的边界方向特征,并采用边界点的不同梯度方向上的灰度平均值表示目标的边界灰度特征,两者结合能有效地描述目标的形状轮廓信息,准确地识别出焊缝图像中的正常收焊点和焊缝缺陷。实验证明:这种基于边界方向直方图的缺陷检测方法具有令人满意的准确性。
Aiming at the welding image of laser torch joint of automobile injector, a defect detection method based on the boundary direction histogram is proposed. The threshold value is used to segment the image under test and the boundary points are extracted from the segmented image. The boundary direction histogram is used to represent the target’s boundary direction features, and the gray values of different gradient directions of the boundary point are used to represent the target The combination of the two can effectively describe the contour information of the target and accurately identify the normal welding point and the weld defect in the weld image. Experiments show that this method based on the boundary-oriented histogram has satisfactory accuracy.