With the acceleration of global urbanization, constructing smart and energy-efficient cities has become a crucial topic for sustainable development. Building energy efficiency, a key component of smart cities, requires accurate identification and classification of building materials. Traditional manual methods are often time-consuming and complex. Consequently, this study primarily uses point cloud data and the PointNet++ neural network to automate the identification and segmentation of building materials. This study involves a data preparation phase, followed by training and evaluating the PointNet++ model. The trained model can categorize building materials into five types: steel, concrete, solar panels, aluminum and brick. In order to demonstrate the practical application of the results of the identification of building materials, some methods have also been introduced in this study, including the estimation of the surface area of different building materials based on the point cloud output from the model as well as calculating the R-values and U-values representing the thermal insulation performance of the building facades based on the material properties and the surface area. This approach can assist in the decision-making process related to thermal insulation performance. Future applications may include real-time building energy efficiency monitoring, aiding in energy management strategy formulation. The techniques could extend to areas such as historical building maintenance, urban planning, and disaster assessment, contributing to the advancement of smarter, more sustainable urban environments.