基于DRN和空洞卷积的图像语义分割算法改进

Improvement of Image Semantic Segmentation Algorithm Based on DRN and Dilated ConVolution

为了提高图像语义分割时的识别和分割能力的问题,本文提出了一种基于DeepLabV3+改进的算法。改进的算法以DeepLab模型作为主体,结合了DRN的结构,减少了分割过程中图像出现网格化的情况。同时为了能够检测到更多边缘信息,有效提高检测分割结果,算法中改进了空洞卷积的部分,提高了分割精度,避免遗漏太多图像信息。通过PASCAL VOC 2012数据集开展的语义分割实验显示,改进的算法有效的提高了在分割时的精度和准确率,本文所提出的网络对图像分割有极大的参考价值。

In order to improve the problem of recognition and segmentation ability in image semantic segmentation,this paper proposes an improved algorithm based on DeepLabV3+model.The improved algorithm takes the DeepLab model as the main body and combines the structure of the dilated residual network to reduce the gridding of the image during the segmentation process.At the same time,in order to be able to detect more edge information and effectively improve the detection segmentation results,the dilated convolution part is improved in the algorithm.The segmentation accuracy is improved,and the image avoid missing too much image information.The semantic segmentation experiment carries out through the PASCAL VOC 2012 dataset shows that the improved algorithm effectively improves the accuracy and accuracy of the segmentation.The network proposed in this paper has great reference value for image segmentation.