Building information models (BIMs) have been widely adopted in current construction projects to enhance the efficiency of facility maintenance operations. As-built BIMs can reflect the actual conditions of facilities and thus as-built BIM reconstruction has shown great significance in digital twin generation, building health monitoring, facility management and urban renewal. Laser scanners are capable to capture dense 3D measurements of the environment in a fast and highly accurate way. Therefore, laser scanning data have been widely used for as-built BIM generation. Although research efforts have been made on how to automatically achieve “Scan-to-BIM”, there are still gaps from applying current solutions to real scenarios. One of the challenges is that some irrelevant point clusters may be wrongly recognized as the desired object in the detection stage. This study presents a novel object verification approach based on deep learning point feature comparison to improve the accuracy of automated BIM reconstruction process. Firstly, a KPConv-based deep neural network is developed and trained to perform 3D point feature computation. Then through comparing point features calculated for extracted point clusters and as-designed BIM generated point clouds, point feature distance maps are generated. Afterwards, to automatically analyze the generated feature distance maps, a dataset including simulated positive and negative instances is created based on ModelNet40. And a tiny neural network is established and trained on the prepared dataset to acquire ability of distinguishment. To validate the feasibility of the proposed technique, experiments were conducted on both artificial point clouds and real scan data collected in one MEP room in a water treatment work in Hong Kong. It is demonstrated that the proposed technique can successfully filter out all the false positives in the Scan-to-BIM process, improving reconstruction accuracy significantly.