Research On Image Segmentation Based On Deep Learning

With the rise of Internet artificial intelligence era,the development of computer vision is also growing.Compared with the traditional image segmentation,the image segmentation based on deep learning has greatly improved the effect and performance,and gradually replaced the traditional image segmentation.The accuracy of segmentation is one of the difficulties in the research of image segmentation based on deep learning,which has a direct impact on the development of automatic driving technology.The main contents of this paper are as follows:(1)In order to make full use of the image context information features,ensure the global consistency of image extraction features,and accurately locate the objects contained in the image,a context based image semantic segmentation network is proposed.The VGG16 network is used as the backbone network,and the context feature extraction module is used to obtain rich image features and take into account the global context information features of the image;The attention module is added to focus on the salient image features in the feature map to further optimize the network precise positioning and improve the final segmentation effect.(2)In order to meet the requirements of automatic driving street scene image segmentation and safe driving of unmanned vehicles,this paper designs an image semantic segmentation network based on multi-scale features and attention mechanism.The network designed the asymmetric convolution module of atrous spatial pyramid pooling,the spatial attention module focusing on the image spatial feature information and the channel attention module focusing on extracting the semantic feature information of the image,which can effectively segment a ll kinds of objects in the street scene image,and solve the problem of single scale and simple extraction of fusion features in the image segmentation model.(3)On the Cityscapes,the paper experiments and simulates the context based semantic segmentatio n network and multi-scale feature and attention mechanism of the semantic segmentation network.The analysis and comparison results show that the proposed method can improve the accuracy and performance of image semantic segmentation.