publications

2026

1. M3FNet: Multi-modal multi-temporal multi-scale data fusion network for tree species composition mapping

   Yuwei Cao, Nicholas C. Coops, Brent A. Murray, and 2 more authors

   Jan 2026

   DOI

2025

1. Enhancing tree species composition mapping using Sentinel-2 and multi-seasonal deep learning fusion

   Yuwei Cao, Nicholas C. Coops, Brent A. Murray, and 1 more author

   International Journal of Remote Sensing, Nov 2025

   DOI

2024

1. The Significance of Porches in Urban Applications: A Method for Automated Modeling and Integration

   Yuwei Cao, Daniele Treccani, and Andrea Adami

   The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Jun 2024

   Abs DOI

   Abstract. Porches, as defined by the Art & Architecture Thesaurus, serve as vital transitional spaces linking indoor and outdoor environments. Despite their historical and contemporary significance, porches lack explicit representation in prevalent standards like CityGML and IndoorGML, posing challenges for comprehensive spatial modeling and its application. This paper proposes a method for modeling porches that aligns with the existing OGC standard CityGML 3.0, ensuring accuracy and compatibility. Drawing upon geomatics techniques, the method aims to bridge the gap in representing these spaces, critical for applications such as navigation systems, urban planning, and energy simulations. By integrating geometric, machine learning, and informative modeling approaches, this method seeks to provide a robust foundation for various practical applications. The paper outlines a comprehensive state-of-the-art review, describes the proposed method from digitalization to random forest (RF)-based point cloud classification and vectorization, presents case studies and results, and offers critical discussions and conclusions. Through this endeavor, the paper contributes to enhancing the representation and understanding of porches within the digital spatial landscape.

2023

1. A 3D BUILDING INDOOR-OUTDOOR BENCHMARK FOR SEMANTIC SEGMENTATION

   Y. Cao, and M. Scaioni

   The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Dec 2023

   Abs DOI

   Abstract. Both machine learning (ML) and deep learning (DL) algorithms require high-quality training samples as well as precise and thorough annotations in order to work effectively. The 3D building indoor-outdoor dataset (BIO dataset), which is a highly accurate, high level of detail, and high coverage dataset for 3D building point cloud and mesh semantic segmentation, is established as a canonical benchmark dataset. It contains 100 building models, in which building structural elements are annotated into 11 semantic categories. Each building in this dataset has an average of 75,587 triangular faces, and the total area of the dataset is 481,769 square meters. Furthermore, semantic segmentation of the dataset was carried out using the Random Forest ML algorithm to verify the dataset’s accessibility. A weighted F1 score of 96.64% was obtained with 10% of the segments of each building randomly chosen as training data. For applications involving building geometry data, the BIO dataset can support a broad class of recently developed ML and DL methodologies.

2. A 3D INDOOR-OUTDOOR BENCHMARK DATASET FOR LoD3 BUILDING POINT CLOUD SEMANTIC SEGMENTATION

   Y. Cao, and M. Scaioni

   The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Oct 2023

   Abs DOI

   Abstract. Deep learning (DL) algorithms require high quality training samples as well as accurate and thorough annotations to work effectively. Up until now a limited number of datasets are available to train DL techniques for semantic segmentation of 3D building point clouds, except a few ones focusing on specific categories of constructions (e.g., cultural heritage buildings). This paper presents a new 3D Indoor/Outdoor building dataset (BIO dataset), which is aimed to provide a highly accurate, detailed, and comprehensive dataset to be used for applications related to sematic classification of buildings based on point clouds and meshes. This benchmark dataset contains 100 building models generated from existing polygonal models and belonging to different categories. These include commercial buildings, residential houses, industrial and institutional buildings. Structural elements of buildings are annotated into 11 semantic categories, following standards from IFC and CityGML. To verify the applicability of the BIO dataset for the semantic segmentation task, it has been successfully tested by using one machine learning technique and four different DL algorithms.

2022

1. A Comprehensive Understanding of Machine Learning and Deep Learning Methods for 3D Architectural Cultural Heritage Point Cloud Semantic Segmentation

   Yuwei Cao, Simone Teruggi, Francesco Fassi, and 1 more author

   In Geomatics for Green and Digital Transition, Oct 2022

   Series Title: Communications in Computer and Information Science

   DOI
2. Integration of Point Clouds from 360° Videos and Deep Learning Techniques for Rapid Documentation and Classification in Historical City Centers

   Yuwei Cao, Mattia Previtali, Luigi Barazzetti, and 1 more author

   In Image Analysis and Processing. ICIAP 2022 Workshops, Oct 2022

   Series Title: Lecture Notes in Computer Science

   DOI
3. A PRE-TRAINING METHOD FOR 3D BUILDING POINT CLOUD SEMANTIC SEGMENTATION

   Y. Cao, and M. Scaioni

   ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, May 2022

   Abs DOI

   Abstract. As a result of the success of Deep Learning (DL) techniques, DL-based approaches for extracting information from 3D building point clouds have evolved in recent years. Despite noteworthy progress in existing methods for interpreting point clouds, the excessive cost of annotating 3D data has resulted in DL-based 3D point cloud understanding tasks still lagging those for 2D images. The notion that pre-training a network on a large source dataset may help enhance performance after it is fine-tuned on the target task and dataset has proved vital in numerous tasks in the Natural Language Processing (NLP) domain. This paper proposes a straightforward but effective pre-training method for 3D building point clouds that learns from a large source dataset. Specifically, it first learns the ability of semantic segmentation by pre-training on a cross-domain source Stanford 3D Indoor Scene Dataset. It then initialises the downstream networks with the pre-trained weights. Finally, the models are fine-tuned with the target building scenes obtained from the ArCH benchmarking dataset. Our paper evaluates the proposed method by employing four fully supervised networks as backbones. The results of two pipelines are compared between training from scratch and pre-training. The results illustrate that pre-training on the source dataset can consistently improve the performance of the target dataset with an average gain of 3.9%.

2021

1. LABEL-EFFICIENT DEEP LEARNING-BASED SEMANTIC SEGMENTATION OF BUILDING POINT CLOUDS AT LOD3 LEVEL

   Y. Cao, and M. Scaioni

   The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Jun 2021

   Abs DOI

   Abstract. In recent research, fully supervised Deep Learning (DL) techniques and large amounts of pointwise labels are employed to train a segmentation network to be applied to buildings’ point clouds. However, fine-labelled buildings’ point clouds are hard to find and manually annotating pointwise labels is time-consuming and expensive. Consequently, the application of fully supervised DL for semantic segmentation of buildings’ point clouds at LoD3 level is severely limited. To address this issue, we propose a novel label-efficient DL network that obtains per-point semantic labels of LoD3 buildings’ point clouds with limited supervision. In general, it consists of two steps. The first step (Autoencoder – AE) is composed of a Dynamic Graph Convolutional Neural Network-based encoder and a folding-based decoder, designed to extract discriminative global and local features from input point clouds by reconstructing them without any label. The second step is semantic segmentation. By supplying a small amount of task-specific supervision, a segmentation network is proposed for semantically segmenting the encoded features acquired from the pre-trained AE. Experimentally, we evaluate our approach based on the ArCH dataset. Compared to the fully supervised DL methods, we find that our model achieved state-of-the-art results on the unseen scenes, with only 10% of labelled training data from fully supervised methods as input.

2020

1. UNDERSTANDING 3D POINT CLOUD DEEP NEURAL NETWORKS BY VISUALIZATION TECHNIQUES

   Y. Cao, M. Previtali, and M. Scaioni

   The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Aug 2020

   Abs DOI

   Abstract. In the wake of the success of Deep Learning Networks (DLN) for image recognition, object detection, shape classification and semantic segmentation, this approach has proven to be both a major breakthrough and an excellent tool in point cloud classification. However, understanding how different types of DLN achieve still lacks. In several studies the output of segmentation/classification process is compared against benchmarks, but the network is treated as a “black-box” and intermediate steps are not deeply analysed. Specifically, here the following questions are discussed: (1) what exactly did DLN learn from a point cloud? (2) On the basis of what information do DLN make decisions? To conduct such a quantitative investigation of these DLN applied to point clouds, this paper investigates the visual interpretability for the decision-making process. Firstly, we introduce a reconstruction network able to reconstruct and visualise the learned features, in order to face with question (1). Then, we propose 3DCAM to indicate the discriminative point cloud regions used by these networks to identify that category, thus dealing with question (2). Through answering the above two questions, the paper would like to offer some initial solutions to better understand the application of DLN to point clouds.

2019

1. A Quantitative Analysis of Creating Genius Loci in a Museum Based on Spatial Visibility Analysis

   Jing Chen, Yuwei Cao, and JingJing Chen

   MATEC Web of Conferences, Aug 2019

   Abs DOI

   Analysing building styles and guiding architectural design with architectural phenomenology is essential for spatial design. Architectural phenomenology and especially shaping the genius loci facilitates emotional expression in design. This creates a building-human connection and a meaningful spatial social culture. It also provides a sense of direction and identity. However, architectural phenomenology is abstract and perceptual, helping designers create architecture through qualitative analysis. This complicates practical guidance during their creative process: it lacks a quantitative explanation of its practical design applications. This research focuses on the genius loci (a sense of direction and identity) in architectural phenomenology. It also explores how quantitative analysis creates a museum’s genius loci. Associating architectural legibility and narrative with the genius loci to enhance spatial readability and ensure richer description is the best way to accomplish this goal. A space syntax-based visibility graph analysis then determines how to create a legible space with a rich narrative for museum visitors. This is a quantitative method creating a museum’s genius loci and provides architects with a standardized design method for creating a building’s social culture. Finally, the Memorial Hall of the Victims in Nanjing Massacre by Japanese Invaders explains how spatial visibility creates and affects the genius loci.

2. Using Geographic Ontologies and Geo-Characterization to Represent Geographic Scenarios

   Yi Huang, May Yuan, Yehua Sheng, and 2 more authors

   ISPRS International Journal of Geo-Information, Dec 2019

   Abs DOI

   Traditional Geographic Information Systems (GIS) represent the environment under reductionist thinking, which disaggregates a geographic environment into independent geographic themes. The reductionist approach makes the spatiotemporal characteristics of geo-features explicit, but neglects the holistic nature of the environment, such as the hierarchical structure and interactions among environmental elements. To fill this gap, we integrate the concept geographic scenario with the fundamental principles of General System Theory to realize the environmental complexity in GIS. With the integration, a geographic scenario constitutes a hierarchy of spatiotemporal frameworks for organizing environmental elements and subserving the exploration of their relationships. Furthermore, we propose geo-characterization with ontological commitments to both static and dynamic properties of a geographic scenario and prescribe spatial, temporal, semantic, interactive, and causal relationships among environmental elements. We have tested the utility of the proposed representation in OWL and the associated reasoning process in Semantic Web Rule Language (SWRL) rules in a case study in Nanjing, China. The case study represents Nanjing and the Nanjing presidential palace to demonstrate the connections among environmental elements in different scenarios and the support for information queries, evolution process simulation, and semantic inferences. The proposed representation encodes geographic knowledge of the environment, makes the interactions among environmental elements explicit, supports geographic process simulation, opens opportunities for deep knowledge mining, and grounds a foundation for GeoAI to discover geographic complexity and dynamics beyond the support of conventional theme-centric inquiries in GIS.

2018

1. Analysis of parameter estimation and optimization application of ant colony algorithm in vehicle routing problem

   Quan-Li Xu, Yu-Wei Cao, and Kun Yang

   In MIPPR 2017: Parallel Processing of Images and Optimization Techniques; and Medical Imaging, Mar 2018

   DOI