Environmental noise affects people’s lives and poses challenges for urban sound classification. Traditional algorithms such as Mel frequency cepstral coefficients (MFCCs) struggle due to audio signal complexity. This study applied an attention mechanism to a deep residual network (ResNet) deep learning network to overcome the structural impact of urban noise on audio signals and improve classification accuracy. We propose a three-feature fusion ResNet + attention method (Net50_SE) to maximize information representation in environmental sound signals. This method uses residual structured convolutional neural networks (CNNs) for feature extraction in sound classification tasks. Additionally, an attention module is added to suppress environmental noise impact and focus on different feature map channels. The experimental results demonstrate the effectiveness of our method, achieving 93.2% accuracy compared with 82.87% with CNN and 84.77% with long short-term memory (LSTM). Our model provides higher accuracy and confidence in urban sound classification.

Estimating forest carbon content typically requires the precise measurement of the trees’ diameter at breast height (DBH), which is crucial for maintaining the health and sustainability of natural forests. Currently, Terrestrial Laser Scanning (TLS) systems are commonly used to acquire forest point cloud data for DBH estimation. However, traditional circular fitting methods face challenges such as a reliance on forest elevation normalization and underfitting of large trees. This study explores a novel approach, the Shape Diameter Function (SDF) algorithm model, leveraging the advantages of three-dimensional point cloud information to replace traditional circular fitting methods. This study employed a parallel approach, combining the Digital Elevation Model (DEM) with Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to segment tree point clouds at breast height. Additionally, a point cloud SDF algorithm based on an octree structure was proposed to accurately estimate individual tree DBH. The research data were obtained from tropical secondary forests located in Cameroon, Peru, Indonesia, and Guyana, with forest ground point cloud data acquired via TLS. The experimental results demonstrated the superior performance of the SDF algorithm in estimating DBH. Compared with the Random Sample Consensus (RANSAC) and Hough transform methods, the Root Mean Square Error (RMSE) decreased by 28.1% and 47.8%, respectively. Particularly in estimating DBH for large trees, the SDF algorithm exhibited smaller errors, indicating a closer alignment between the estimated individual tree DBH values and those obtained from manual measurements. This study presented a more accurate DBH estimation algorithm, contributing to the exploration of improved forest carbon content estimation methods.