Research on Segmentation Method of Civil Construction Image based on Unet Algorithm Model

In order to accurately segment buildings and features in high-resolution remote sensing images. First, the boundary distance map is generated according to the ground truth map of the remote sensing image, and the remote sensing image and the ground truth map are used as the input of the Unet network; then, the building feature prediction layer is added to the end of the Unet network constructed based on the ResNet network. With the boundary distance prediction layer, a multi-task network is built; finally, the loss function of the multi-task network is defined, and the Adam optimization algorithm is used to train the network. Experiments are carried out on the Inria aerial remote sensing image building and landmark data set. The results show that compared with the fully convolutional network combined with the multi-layer perceptron method, the intersection ratio of VGG16 network, VGG16+ boundary prediction, ResNet50 and this method are improved by 5.15, 6.94, 6.41 and 7.86 percentage points, the accuracy is increased to 94.71%, 95.39%, 95.30% and 96.10% respectively, which can achieve high-precision extraction of architectural features.