Optimization of Rural Landscape Classification Model Based on Convolutional Neural Network

Rural landscape classification is of great significance to land planning, agricultural management and cultural protection. Traditional classification methods are difficult to adapt to the complexity, diversity and seasonal changes of rural landscape. Aiming at the problem of rural landscape classification, this paper proposes a convolutional neural network model, constructs a large-scale data set containing 10 typical rural landscapes, designs the data strategy of rural landscape feature enhancement and spatial attention self-calibration, introduces the spatial attention module, solves the landscape boundary ambiguity and scale change problems, and improves the learning rate adjustment strategy. Experimental data in this study show that the classification accuracy of this model reaches 93.27 %, which is 7.83% higher than that of the basic ResNet-50, and has achieved significant advantages.