Point Cloud Processing with Machine Learning

Dear Colleagues,

LiDAR (Light Detection and Ranging) and photogrammetry systems have been recognized as two of the most suitable means of remote sensing, acquiring large-scale three-dimensional (3D; x, y, z; a trio) geo-referenced points known as point clouds. On the one hand, recent advances in LiDAR technologies produce dense point clouds that can digitize objects and environments with a millimeter-level of accuracy. Aerial photogrammetry, on the other hand, provides a cost-effective solution for surveying a large terrain of interest with a centimeter-level of accuracy.  Regardless, point clouds are one of the most promising data types which have been used for 3D object representation, classification, segmentation, and modeling in recent years. They provide detailed 3D geometry and spatial information of objects. One of the major challenges of automatizing complex and non-repetitive tasks is data processing. Processing point clouds with complex topology is not trivial as the 3D points are usually incomplete, sparse, unorganized, and have variable point density. Additionally, the presence of outliers and noise is a common phenomenon that makes processing more challenging. Point clouds are often accompanied with huge volumes; hence, they require automatic and efficient processing for reliable–robust analysis and for precise decision making. 

Machine learning (ML) techniques (usually classified as supervised, unsupervised, and reinforcement learning), classical (e.g., regression and support vector machine) to advanced (e.g., ensemble learning; LightGBM; and XGBoost), have been used for point clouds processing. In recent years, a branch of the most advanced ML methods, deep learning (DL), has shown unprecedented success for objects segmentation, classification, and detection in point clouds using different forms of Artificial Neural Networks (ANNs). For this Special Issue, we encourage potential authors to submit novel/improved methods and/or algorithms of ML and DL to benefit the related communities of scientists, researchers, and industry.

This Special Issue aims to show the advantages and limitations of different ML algorithms (including deep learning) in point cloud processing (e.g., objects classification, segmentation, detection, visualization, and modeling) for various fields of applications, such as object modeling, visualization, feature extraction, digital twins solutions, scan-to-BIM, infrastructure (e.g., building, transportation, road-corridor) monitoring, robotics, autonomous driving, forest monitoring, environment, and smart agriculture.

• Object classification, segmentation, detection, monitoring, and change detection in road environment, transportation (e.g., tunnels and bridges), and buildings. 
• Object detection, classification, and scene perception for autonomous vehicles and robots.
• Estimation of metrics of forest inventories, such as individual tree height, diameter of breast height, and stem and canopy modeling.
• Feature extraction for point cloud processing in various applications of city modeling, as well as environmental and agricultural monitoring.
• Multiple sensors (LiDAR, optical sensor, IMU, etc.,) modeling and cross-modality integration.
• Deep learning-based methods for SLAM systems.

Dr. Abdul Awal Md Nurunnabi
Dr. Meida Chen
Dr. Yan Xia
Prof. Dr. Felicia Norma Rebecca Teferle
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Remote Sensing is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2700 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• classification
• deep learning
• feature extraction
• machine learning
• pattern recognition
• point cloud
• remote sensing
• object detection
• segmentation
• visualization
• scene understanding
• SLAM systems