基于改进DeepLabV3+的铣床碎屑图像分割方法

针对铣床碎屑形状不规则导致图像分割中碎屑轮廓不清晰、分割精度低的问题，本文提出一种改进的DeepLabV3+铣床碎屑分割算法。首先在DeepLabV3+的Xcepetion模块中嵌入通道与空间注意力机制(convolutional block attention module, CBAM)模型，优化通道的权重和位置信息，加强碎屑图像区域的特征学习；其次将DeepLabV3+的空洞空间卷积池化金字塔(atrous spatial pyramid pooling, ASPP)模块改为密集连接(dense conolutional network, DenseNet)方式，增大碎屑图像特征点的感受野，提升铣床碎屑图像特征的复用效率；最后在解码过程中采用多尺度自适应特征融合方法，聚合多尺度特征作为解码器的输入特征，提高碎屑图像分割的精度与鲁棒性。实验结果表明，本文算法优于其他分割算法，改进后算法相比DeepLabV3+,像素准确率提高0.026,平均交并比(mean intersection over union,MIOU)提高0.020,F₁值提高了0.013。

Segmentation method of milling machine debris image based on improved DeepLabV3+

Aiming at the problem that the irregular shape of the milling machine debris leads to the unclear outline of the debris and the low segmentation accuracy in the image segmentation, this paper proposes an improved DeepLabV3+ milling machine debris segmentation algorithm. First,the CBAM model is embedded in the Xcepetion module of DeepLabV3+ to optimize the weight and position information of the channel,and to strengthen the feature learning of the debris image area.Secondly,the ASPP module of DeepLabV3+ is changed to the dense connection method to increase the receptive field of the feature points of the debris image.The feature reuse efficiency of the milling machine debris image is improved.Finally,the multi-scale adaptive feature fusion method is adopted in the decoding process,and the multi-scale features are aggregated as the input features of the decoder to improve the segmentation accuracy and robustness of the debris image.The experimental results show that the algorithm in this paper is better than other segmentation algorithms.Compared with DeepLabV3+,the improved algorithm improves the pixel accuracy by 0.026,mean intersection over union (MIOU) by 0.020,and the F₁-measure by 0.013.