基于机器视觉的焊缝图像识别预处理的研究

Seam image recognition preprocessing based on machine vision

针对传统的图像处理算法在处理焊缝图像时由于外界环境干扰而导致提取出来的中心线图像出现断续或者偏离实际中心线,从而引起机械臂焊接过程中出现偏差的情况,文中提出基于改进的暗通道先验焊缝图像预处理算法。首先,对每次焊接前的焊接环境进行视觉检测以获取像素灰度变化差值,此后,利用改进暗通道先验理论结合像素灰度差值对焊缝图像进行增强,最后,通过实验完成图像预处理的一系列过程,并将融合了改进暗通道先验理论的图像预处理结果与传统图像预处理结果进行比较。结果表明,该算法能够获得清晰、连续、居中的焊缝中心线,具有更好的处理效果。

An improved dark channel prior-based welding seam pre-processing algorithm is proposed to handle the deviation during welding operation by the manipulator due to image discontinuity of or deviation from the center line caused by external interference when traditional image processing algorithms used to process welding seam images. Visual inspection of welding surroundings is carried out before every welding operation in order to work out the pixel gray difference, followed by welding seam image enhancement using the improved dark channel prior theory together with pixel gray difference. Finally, image pre-processing is completed through experiments. Comparison between the result of image pre-processing integrating improved dark channel prior theory and that of traditional image processing suggest that the algorithm developed in this study is able to obtain clear, continuous and centered weld center line with favorable processing effect.