基于改进YOLOv4的工业棒料识别算法

针对工业棒料存在遮挡干扰时难以快速有效识别的问题,提出了一种基于改进YOLOv4的棒料识别算法。首先对YOLOv4进行轻量化改进,将改进的Mobilenetv3作为YOLOv4的主干网络,以减少模型参数量,提高算法的检测速度。然后提出在YOLOv4原损失函数基础上串联Repulsion损失函数,此新增损失函数包含2部分:RepGT损失和RepBox损失,RepGT损失函数计算目标预测框与相邻真实框所产生的损失值,用来减少棒料误检;RepBox损失函数计算目标预测框与相邻的其他目标预测框所产生的损失值,用来减少棒料漏检。实验结果表明,改进算法的检测速度为63帧/s,比原YOLOv4算法提升了20帧/s;识别准确率达到97.85%,比原YOLOv4算法提升了1.62%。

Industrial Bar Recognition Algorithm Based on Improved YOLOv4

Aiming at the problem that it is difficult to recognize the industrial bar quickly and effectively when there is partial occlusion interference,a bar recognition algorithm based on improved YOLOv4 is proposed.Firstly,the lightweight improvement of YOLOv4 is carried out,and the improved Mobilenetv3 is used as the backbone network of YOLOv4,so as to reduce the amount of model parameters and improve the detection speed of the algorithm.Then the Repulsion loss function is cascaded on the basis of YOLOv4s original loss function.The improved loss function contains two parts:RepGT loss and RepBox loss.The RepGT loss function calculates the loss value generated by the target prediction box and the adjacent real boxes,which is used to reduce the bar fault detection;The RepBox loss function calculates the loss value generated by the target prediction box and the adjacent other target prediction boxes,which is used to reduce the missing detection of bars.The experimental results show that the detection speed of the improved algorithm is 63 FPS,which is 20 FPS higher than that of the original YOLOv4 algorithm.The recognition accuracy is 97.85%,which is 1.62%higher than the original YOLOv4 algorithm.