基于LabVIEW和Mask R-CNN的柱塞式制动主缸内槽表面缺陷检测

Surface defect detection of inner groove in plunger brake master cylinder based on LabVIEW and Mask R-CNN

为了解决传统图像处理方法对于铸铝材料表面缺陷检测通用性不高、准确度低等问题,研究了一种基于Mask R-CNN神经网络的缺陷检测系统。首先,采用自主研发的缺陷检测装置采集柱塞式制动主缸内槽表面图像,对其进行预处理,制作成Microsoft COCO格式数据集;其次,搭建适用于该数据集的Mask R-CNN神经网络结构,并绘制训练过程损失函数与平均精度均值曲线;最后,将检测结果与基于SVM和Faster R-CNN模型的检测结果进行比较,统计了3种神经网络模型的单图检测平均时间和识别率。试验结果表明,在相同样本条件下,该方法的识别率比另外2种方法高,达到了93.6%,能够更精确地检测柱塞式制动主缸内槽的表面缺陷。

In order to solve the problems that the traditional image processing method has low versatility and low accuracy for surface defect detection of cast aluminum materials,a defect detection system based on Mask R-CNN neural network was studied.Firstly,the images of the inner groove in plunger brake master cylinder were collected by the self-developed defect detection device and pre-processed to form a Microsoft COCO format data set.Secondly,the Mask R-CNN neural network structure suitable for the data set was built and the loss function and mean average precision curve of the training process were drawn.Finally,the test results were compared with the test results based on SVM and Faster R-CNN model,the average time of single picture detection and the recognition rate of three neural network models were counted.The experimental results show that under the same sample conditions,the recognition rate of this method is higher than that of the other two methods,reaching 93.6%,which can more accurately detect the groove surface defects in plunger brake master cylinder.