Research on Semantic Segmentation Algorithm for Remote Sensing Images Based on DeepLabv3+

Currently, for high-resolution remote sensing images with huge data volume, it leads to slow processing speed of the algorithm, which is difficult to meet the real-time application requirements. In addition, small objects occupy fewer pixels in the image, which are easily interfered by background noise, and the model is difficult to learn enough features during training, which leads to poor detection and segmentation of small targets. Thus, we propose a method for semantic segmentation of high-resolution remote sensing imagery, which is based on an enhanced version of the DeepLabv3+ model. By replacing the feature extraction backbone network and adding the attention mechanism module, it can effectively improve the accuracy of semantic cutting of remote sensing images, enhance the detection and segmentation ability of small targets, and improve the processing efficiency and adaptability of the model to multiple targets. The experimental results demonstrate the efficacy of the suggested approach. The proposed network model is validated using the Vaihingen public dataset and compared with several commonly used semantic segmentation network models.