基于改进Yolo v3的电连接器缺陷检测

Defect Detection of Electrical Connector Based on Improved Yolo v3

针对电连接器缺陷检测目前存在自动化程度低、检测精度低、检测速度慢以及鲁棒性较差等问题,提出了一种改进的Yolo v3算法来检测电连接器的缺陷。首先采用K-means聚类算法对本文数据集进行聚类分析得到3种宽高比的候选框,针对本文缺陷对象提高检测精度;对主干网络第三个残差块输出的8倍降采样特征图进行4倍上采样,将得到的特征图与第二个残差块输出的2倍降采样特征图进行拼接得到融合特征检测层;将目标检测层之前经过的6个DBL单元改为2个DBL单元加上2个残差单元,以提高特征的复用与获取;另外本文选择单尺度特征图进行目标检测,而不是原网络的多尺度预测,既节省了计算量,又一定程度上避免误检。通过定性与定量的实验结果表明,本文改进后的Yolo v3算法对电连接器缺陷检测有着更好的性能以及速度,准确率为93.5%,相较于Faster R-CNN更加准确,原Yolo v3更加快速,基本上满足了工业现场对电连接器检测的要求。

Aiming at the problems of electrical connector defect detection,such as low automation,low detection accuracy,slow detection speed,and poor robustness,this paper proposes an improved Yolo v3 algorithm to detect electrical connector defects. First,the K-means clustering algorithm is used to perform cluster analysis on the data set of this paper to obtain three kinds of candidate frames with aspect ratios to improve the detection accuracy for the defective objects in this paper;the 8-times downsampling feature of the third residual block output of the backbone network The image is up-sampled 4 times,and the obtained feature map is merged with the down-sampled feature map of the second residual block to obtain the fusion feature detection layer;the 6 DBL units passed by the target detection layer are changed to 2 DBL unit plus 2 residual units to improve feature reuse and acquisition;In addition,this paper chooses single-scale feature maps for target detection instead of multi-scale prediction of the original network,which not only saves the calculation amount,but also avoids it to a certain extent False detection. The qualitative and quantitative experimental results show that the improved Yolo v3 algorithm has better performance and speed for electrical connector defect detection,with an accuracy rate of 93.5%,which is more accurate than Faster R-CNN,and the original Yolo v3 is more It is fast and basically meets the requirements of industrial field for electrical connector testing.