基于Ghost-SK-DeblurGAN的钢筋套丝头图像去模糊算法

Image Deblurring Algorithm of Rebar Threading Head Based on Ghost-SK-DeblurGAN

针对视觉测量钢筋套丝头尺寸方法中,由于振动导致采集到的图片存在运动模糊而影响测量结果的问题,提出了一种轻量化的带有注意力机制的基于DeblurGAN-v2的去模糊算法Ghost-SK-DeblurGAN,算法采用GhostNet轻量化模块作为特征提取网络,引入注意力模块SKNet,对生成器损失函数进行修改。采集了不同规格的清晰钢筋套丝头图像,对采集到的清晰图像施加运动模糊处理,得到模糊图像,构建模糊数据集BRT。实验结果表明,与其他基于DeblurGAN-v2的去模糊算法相比,该算法能够兼顾去模糊效果和实时性。

Aiming at the problem of motion blur in the collected images due to vibration in the visual measurement of the size of the rebar threader, which affects the measurement results, a lightweight deblurring algorithm with an attention mechanism based on DeblurGAN-v2, Ghost, is proposed in this paper. The clear images of rebar threading heads of different specifications are collected, and motion blur is applied to the collected clear images to obtain blurred images, and the blur data set BRT is constructed. The experience results show that, compared with other deblurring algorithms based on DeblurGAN-v2, the algorithm can take into account the deblurring effect and real-time performance.