This paper provides a comprehensive review of deep transfer learning (DTL) and domain adaptation (DA) techniques for advancing the understanding of 3D point clouds (3DPCs). The authors highlight the challenges faced by traditional machine learning (ML) and deep learning (DL) methods, particularly in handling large-scale datasets, computational resources, and data distribution issues. DTL is introduced as a solution to these challenges by leveraging knowledge from a source dataset to improve performance on a target dataset. The paper covers various applications of DTL and DA in 3DPC tasks, including object detection, semantic labeling, segmentation, classification, registration, downsampling/upsampling, and denoising. It presents a well-defined taxonomy of DTL algorithms, discusses existing datasets and evaluation metrics, and identifies open challenges and future research directions. The review also includes a detailed analysis of popular deep learning models for 3DPCs, such as PointNet, PointNet++, SplatNet, SGPN, and FoldingNet, and their contributions to different tasks. Additionally, the paper explores the use of pre-trained 2D and 3D models for 3DPC understanding, emphasizing the importance of feature extraction and geometric detail enhancement. Finally, it discusses various evaluation metrics used to assess the performance of DTL approaches, including F1 measures, overall accuracy (OA), intersection over union (IoU), and mean IoU.This paper provides a comprehensive review of deep transfer learning (DTL) and domain adaptation (DA) techniques for advancing the understanding of 3D point clouds (3DPCs). The authors highlight the challenges faced by traditional machine learning (ML) and deep learning (DL) methods, particularly in handling large-scale datasets, computational resources, and data distribution issues. DTL is introduced as a solution to these challenges by leveraging knowledge from a source dataset to improve performance on a target dataset. The paper covers various applications of DTL and DA in 3DPC tasks, including object detection, semantic labeling, segmentation, classification, registration, downsampling/upsampling, and denoising. It presents a well-defined taxonomy of DTL algorithms, discusses existing datasets and evaluation metrics, and identifies open challenges and future research directions. The review also includes a detailed analysis of popular deep learning models for 3DPCs, such as PointNet, PointNet++, SplatNet, SGPN, and FoldingNet, and their contributions to different tasks. Additionally, the paper explores the use of pre-trained 2D and 3D models for 3DPC understanding, emphasizing the importance of feature extraction and geometric detail enhancement. Finally, it discusses various evaluation metrics used to assess the performance of DTL approaches, including F1 measures, overall accuracy (OA), intersection over union (IoU), and mean IoU.