Research on UAV Image Road Recognition Application Based on Improved SRCNN

The application of drones in surveying and monitoring is becoming increasingly important, especially in road information detection. Traditional road detection algorithms have low accuracy and poor anti-interference capabilities, while deep learning methods, although superior in performance, require a large amount of labeled data and have high computational complexity, making them difficult to apply in practice. To address this, this paper proposes an improved drone image road recognition method that combines SRCNN and UNet. The paper improves SRCNN by introducing an attention mechanism to enhance feature extraction capabilities and optimizes the network structure to reduce computational complexity. The improved SRCNN generates high-resolution images as input for UNet, which is further optimized to adapt to drone images. Experimental results show that this method outperforms existing technologies in terms of accuracy, real-time performance, and robustness, meeting practical requirements. In the future, we will explore combining other sensor data and optimizing network structures to enhance road detection performance.