基于改进SSD算法的工件检测

Workpiece Detection Based on Improved SSD Algorithm

针对自动化生产线上运动的工件检测准确度要求高,部分小工件难以检测的问题,提出一种基于SSD改进的工件目标检测算法。在SSD算法的基础上,将SSD的骨干网络VGG16替换为ResNet50,并增加输出特征层的预选框数量,其次加入特征金字塔(FPN)的算法思想,将网络中的底层特征与高层特征融合,解决了SSD算法中对底层特征提取能力不足的问题,从而改善了对小目标工件检测效果差的缺陷。研究结果表明,相比于SSD算法改进的SSD工件识别方法降低了对小目标工件的漏检率,平均准确率提高了2.5%,保证了工件检测的准确率。

Aiming at the problem that the detection accuracy of moving workpieces in automated production lines is high,and some small workpieces are difficult to detect,an improved workpiece target detection algorithm based on SSD is proposed.On the basis of the SSD algorithm,replace the SSD backbone network VGG16 with ResNet50,and increase the number of preselected boxes in the output feature layer,and then add the algorithm idea of feature pyramid(FPN)to integrate the low-level features in the network with the high-level features to solve The problem of insufficient ability to extract underlying features in the SSD algorithm is solved,so as to improve the defect of poor detection effect on small target workpieces.The results show that compared with the SSD algorithm,the improved SSD workpiece recognition method reduces the missed detection rate of small target workpieces,and the average accuracy rate is increased by 2.5%,which ensures the accuracy of workpiece detection.