Deep Learning Algorithm Based Remote Sensing Image Classification Research

In the field of remote sensing image recognition, compared with deep learning related image recognition algorithms, traditional image recognition algorithms have limitations in recognizing different types of land, so a large amount of deep learning is used for remote sensing image recognition. However, different models have different effects on different tasks, and selecting an excellent deep learning model is a matter of concern. This paper takes WHDLD dataset as the data source, DeepLabV3Plus, HRNet-OCR, ACGAN and other deep learning models as the experimental objects to study the performance ability of different deep learning models on WHDLD dataset. The experimental results show that ACGAN is the best among the experimental models. accuracy, precision, recall, F1-score are 0.83, 0.95, 0.92 and 0.90 respectively. This study can provide some reference values for remote sensing image land classification.