基于边缘计算的机巡图像缺陷识别算法研究

Research on image defect recognition algorithm based on edge calculation

机巡图像识别算法(Faster R-CNN算法)存在识别精度低的不足,为此提出基于边缘计算的机巡图像缺陷识别算法。通过对机巡图像缺陷特征提取,再利用RPN网络获取目标候选区域,利用边缘计算优化Faster R-CNN算法,在此基础上,实行Faster R-CNN训练,再通过正负样本和损失函数对目标区域实施精确分类,从而识别出机巡图像缺陷,至此完成基于边缘计算的机巡图像缺陷识别算法研究。通过与传统算法、未经优化的Faster R-CNN算法作对比实验,实验结果表明,提出的基于边缘计算的机巡图像缺陷识别算法能更有效地识别机巡图像缺陷。

The aircraft patrol image recognition algorithm(Faster R-CNN algorithm)has the disadvantage of low recognition accuracy.For this reason,a machine-based image defect recognition algorithm based on edge calculation is proposed.By extracting the defect features of the camera image,and then using the RPN network to obtain the target candidate region,the edge calculation is used to optimize the Faster R-CNN algorithm.On this basis,the Faster R-CNN training is implemented,and then the positive and negative samples and the loss function are used to target the target.The area is accurately classified to identify the defect of the camera image,and the algorithm for defect recognition of the camera image based on edge calculation is completed.Compared with the traditional algorithm and the unoptimized Faster R-CNN algorithm,the experimental results show that the proposed edge recognition based image recognition algorithm can more effectively identify the defect of the camera image.