Recently, researchers have realized a number of achievements involving deep-learning-based neural networks for the tasks of segmentation and detection based on 2D images, 3D point clouds, etc. Using 2D and 3D information fusion for the advantages of compensation and accuracy improvement has become a hot research topic. However, there are no critical reviews focusing on the fusion strategies of 2D and 3D information integration based on various data for segmentation and detection, which are the basic tasks of computer vision. To boost the development of this research domain, the existing representative fusion strategies are collected, introduced, categorized, and summarized in this paper. In addition, the general structures of different kinds of fusion strategies were firstly abstracted and categorized, which may inspire researchers. Moreover, according to the methods included in this paper, the 2D information and 3D information of different methods come from various kinds of data. Furthermore, suitable datasets are introduced and comparatively summarized to support the relative research. Last but not least, we put forward some open challenges and promising directions for future research.

To bridge the modality gap between camera images and LiDAR point clouds in autonomous driving systems—a critical challenge exacerbated by current fusion methods’ inability to effectively integrate cross-modal features—we propose the Cross-Modal Fusion (CMF) framework. This attention-driven architecture enables hierarchical multi-sensor data fusion, achieving state-of-the-art performance in semantic segmentation tasks.The CMF framework first projects point clouds onto the camera coordinates through the use of perspective projection to provide spatio-depth information for RGB images. Then, a two-stream feature extraction network is proposed to extract features from the two modalities separately, and multilevel fusion of the two modalities is realized by a residual fusion module (RCF) with cross-modal attention. Finally, we design a perceptual alignment loss that integrates cross-entropy with feature matching terms, effectively minimizing the semantic discrepancy between camera and LiDAR representations during fusion. The experimental results based on the SemanticKITTI and nuScenes benchmark datasets demonstrate that the CMF method achieves mean intersection over union (mIoU) scores of 64.2% and 79.3%, respectively, outperforming existing state-of-the-art methods in regard to accuracy and exhibiting enhanced robustness in regard to complex scenarios. The results of the ablation studies further validate that enhancing the feature interaction and fusion capabilities in semantic segmentation models through cross-modal attention and perceptually guided cross-entropy loss (Pgce) is effective in regard to improving segmentation accuracy and robustness.

Identifying the driving factors of urban land surface temperatures (U-LSTs) is critical in improving urban thermal environments and in supporting the sustainable development of cities. Previous studies have demonstrated that two- and three-dimensional (2D and 3D) urban structure parameters (USPs) largely influence seasonal U-LSTs. However, the effects of 2D and 3D USPs on seasonal U-LSTs at different spatial scales still await a general explanation. In this study, we used very-high-resolution remotely sensed data to investigate how 2D and 3D USPs impact seasonal U-LSTs at different spatial scales (including pixel and city block scales). In addition, the influences of various functional zones on U-LSTs were analyzed. The results show that, (1) generally, the links between USPs and U-LSTs at the city block scale were more obvious than those at the pixel scale, e.g., the Pearson correlation coefficient (r) between U-LST and the mean building height at the city block scale (summer: r = −0.156) was higher than that at the pixel scale (summer: r = −0.081). Tree percentage yielded a considerable cooling effect on summer U-LSTs on both the pixel (r = −0.199) and city block (r = −0.369) scales, and the effect was more obvious in regions with tall trees. (2) The independently total explained variances (R²) of 3D USPs on seasonal U-LSTs were considerably higher than those of 2D USPs in most urban functional zones (UFZs), suggesting the distinctive roles of 3D USPs in U-LST regulation at the local scale. Three-dimensional USPs (R² value = 0.66) yielded more decisive influences on summer U-LSTs than 2D USPs did (R² value = 0.48). (3) Manufacturing zones yielded the highest U-LST, followed by residential and commercial zones. Notably, it is found that the explained variances of the total study area for seasonal U-LSTs were significantly lower than those of each UFZ, suggesting the different roles of 2D and 3D USPs played in various UFZs and that it is critical to explain U-LST variations by using UFZs.

The measurement accuracy of the Mobile Mapping System (MMS) is the main problem, which restricts its development and application, so how to calibrate the MMS to improve its measurement accuracy has always been a research hotspot in the industry. This paper proposes a position and attitude calibration method with error correction based on the combination of the feature point and feature surface. First, the initial value of the spatial position relationship between each sensor of MMS is obtained by close-range photogrammetry. Second, the optimal solution for error correction is calculated by feature points in global coordinates jointly measured with International GNSS Service (IGS) stations. Then, the final transformation parameters are solved by combining the initial values obtained originally, thereby realizing the rapid calibration of the MMS. Finally, it analyzed the RMSE of MMS point cloud after calibration, and the results demonstrate the feasibility of the calibration approach proposed by this method. Under the condition of a single measurement sensor accuracy is low, the plane and elevation absolute accuracy of the point cloud after calibration can reach 0.043 m and 0.072 m, respectively, and the relative accuracy is smaller than 0.02 m. It meets the precision requirements of data acquisition for MMS. It is of great significance for promoting the development of MMS technology and the application of some novel techniques in the future, such as autonomous driving, digital twin city, urban brain et al.

Segmentation is an important step in point cloud data feature extraction and three‐dimensional modelling. Currently, it is also a challenging problem in point cloud processing. There are some disadvantages of the DBSCAN method, such as requiring the manual definition of parameters and low efficiency when it is used for large amounts of calculation. This paper proposes the AQ‐DBSCAN algorithm, which is a density clustering segmentation method combined with Gaussian mapping. The algorithm improves upon the DBSCAN algorithm by solving the problem of automatic estimation of the parameter neighborhood radius. The improved algorithm can carry out density clustering processing quickly by reducing the amount of computation required.