Utilizing Deep Learning Algorithms for Automatic Recognition of Urban Park Landscape Features

Traditional landscape feature recognition methods often rely on manual feature extraction, facing problems such as insufficient recognition accuracy and difficulty in processing complex scenes. In order to overcome these challenges, this paper proposes an automatic recognition method for urban park landscape features based on deep learning algorithm. This method combines the neural network structure composed of ResNet and ASPP modules, giving full play to the advantages of both in landscape feature extraction. First, ResNet effectively alleviates the common gradient vanishing problem in deep neural networks through its deep residual connection mechanism, and can efficiently extract multi-level features in park landscape images. Secondly, the ASPP module can simultaneously capture different scale information in the image through multi-scale convolution operations, improving the recognition ability of complex landscape elements. This multi-scale information fusion enables the model to effectively improve the recognition accuracy and robustness when processing various detailed features in urban parks. In order to verify the effectiveness of the model, this paper compares the recognition performance of the proposed method with other traditional models through experiments. Experimental results show that the neural network structure based on ResNet and ASPP modules can significantly improve the accuracy in the recognition task of park landscape features, especially in complex images with diverse landscape features, showing strong generalization ability and precision. This method can also effectively analyze the spatial layout and environmental characteristics of park landscapes, providing strong technical support for the intelligent management, environmental protection and sustainable development of urban parks.