RANdom SAmple Consensus (RANSAC) is a widely adopted method for LiDAR point cloud segmentation because of its robustness to noise and outliers. However, RANSAC has a tendency to generate false segments consisting of points from several nearly coplanar surfaces. To address this problem, we formulate the weighted RANSAC approach for the purpose of point cloud segmentation. In our proposed solution, the hard threshold voting function which considers both the point-plane distance and the normal vector consistency is transformed into a soft threshold voting function based on two weight functions. To improve weighted RANSAC’s ability to distinguish planes, we designed the weight functions according to the difference in the error distribution between the proper and improper plane hypotheses, based on which an outlier suppression ratio was also defined. Using the ratio, a thorough comparison was conducted between these different weight functions to determine the best performing function. The selected weight function was then compared to the existing weighted RANSAC methods, the original RANSAC, and a representative region growing (RG) method. Experiments with two airborne LiDAR datasets of varying densities show that the various weighted methods can improve the segmentation quality differently, but the dedicated designed weight functions can significantly improve the segmentation accuracy and the topology correctness. Moreover, its robustness is much better when compared to the RG method.

Regular inspection of transmission lines is an essential work, which has been implemented by either labor intensive or very expensive approaches. 3D reconstruction could be an alternative solution to satisfy the need for accurate and low cost inspection. This paper exploits the use of an unmanned aerial vehicle (UAV) for outdoor data acquisition and conducts accuracy assessment tests to explore potential usage for offsite inspection of transmission lines. Firstly, an oblique photogrammetric system, integrating with a cheap double-camera imaging system, an onboard dual-frequency GNSS (Global Navigation Satellite System) receiver and a ground master GNSS station in fixed position, is designed to acquire images with ground resolutions better than 3 cm. Secondly, an image orientation method, considering oblique imaging geometry of the dual-camera system, is applied to detect enough tie-points to construct stable image connection in both along-track and across-track directions. To achieve the best geo-referencing accuracy and evaluate model measurement precision, signalized ground control points (GCPs) and model key points have been surveyed. Finally, accuracy assessment tests, including absolute orientation precision and relative model precision, have been conducted with different GCP configurations. Experiments show that images captured by the designed photogrammetric system contain enough information of power pylons from different viewpoints. Quantitative assessment demonstrates that, with fewer GCPs for image orientation, the absolute and relative accuracies of image orientation and model measurement are better than 0.3 and 0.2 m, respectively. For regular inspection of transmission lines, the proposed solution can to some extent be an alternative method with competitive accuracy, lower operational complexity and considerable gains in economic cost.

3D semantic labeling is a fundamental task in airborne laser scanning (ALS) point clouds processing. The complexity of observed scenes and the irregularity of point distributions make this task quite challenging. Existing methods rely on a large number of features for the LiDAR points and the interaction of neighboring points, but cannot exploit the potential of them. In this paper, a convolutional neural network (CNN) based method that extracts the high-level representation of features is used. A point-based feature image-generation method is proposed that transforms the 3D neighborhood features of a point into a 2D image. First, for each point in the ALS data, the local geometric features, global geometric features and full-waveform features of its neighboring points within a window are extracted and transformed into an image. Then, the feature images are treated as the input of a CNN model for a 3D semantic labeling task. Finally, to allow performance comparisons with existing approaches, we evaluate our framework on the publicly available datasets provided by the International Society for Photogrammetry and Remote Sensing Working Groups II/4 (ISPRS WG II/4) benchmark tests on 3D labeling. The experiment results achieve 82.3% overall accuracy, which is the best among all considered methods.

The classification of point clouds is a basic task in airborne laser scanning (ALS) point cloud processing. It is quite a challenge when facing complex observed scenes and irregular point distributions. In order to reduce the computational burden of the point-based classification method and improve the classification accuracy, we present a segmentation and multi-scale convolutional neural network-based classification method. Firstly, a three-step region-growing segmentation method was proposed to reduce both under-segmentation and over-segmentation. Then, a feature image generation method was used to transform the 3D neighborhood features of a point into a 2D image. Finally, feature images were treated as the input of a multi-scale convolutional neural network for training and testing tasks. In order to obtain performance comparisons with existing approaches, we evaluated our framework using the International Society for Photogrammetry and Remote Sensing Working Groups II/4 (ISPRS WG II/4) 3D labeling benchmark tests. The experiment result, which achieved 84.9% overall accuracy and 69.2% of average F1 scores, has a satisfactory performance over all participating approaches analyzed.

Stereo matching of high-resolution satellite images (HRSI) is still a fundamental but challenging task in the field of photogrammetry and remote sensing. Recently, deep learning (DL) methods have demonstrated the potential for stereo matching on public benchmark datasets. Currently, mainstream stereo matching models depend on quantities of training data with ground truth. However, datasets for stereo matching of satellite images are scarce, which profoundly blocks the application of DL in this field. To facilitate further research, this paper publishes a large-scale dataset, termed WHU-Stereo, for stereo matching DL network training and testing. This dataset is created by using airborne LiDAR point clouds and high-resolution stereo imageries taken from the Chinese GaoFen-7 satellite (GF-7). Occlusions should be seriously considered in the generation of ground-truth disparities. We propose an occlusion removal technique, which can adapt to point clouds with different densities and shows high potential in training data preparation. The WHU-Stereo dataset contains more than 1700 epipolar rectified image pairs, which cover six areas in China and includes various kinds of landscapes. We have assessed the accuracy of ground-truth disparity maps, and it is shown that our dataset achieves comparable precision compared with existing state-of-the-art stereo matching datasets. To verify its feasibility, in experiments, the hand-crafted SGM algorithm and recent DL networks have been tested on the dataset. Experimental results show that the WHU-Stereo dataset can serve as a challenging benchmark for stereo matching of high-resolution satellite images and performance evaluation of deep learning models. Our dataset is available at https://github.com/Sheng029/WHU-Stereo.

Building height is one of the basic geographic information for planning and analysis in urban construction. It is still very challenging to estimate the accurate height of complex buildings from satellite images, especially for buildings with podium. This paper proposes a solution for building height estimation from GF-7 satellite images by using a roof contour constrained stereo matching algorithm and DSM (Digital Surface Model) based bottom elevation estimation. First, an object-oriented roof matching algorithm is proposed based on building contour to extract accurate building roof elevation from GF-7 stereo image, and DSM generated from the GF-7 stereo images is then used to obtain building bottom elevation. Second, roof contour constrained stereo matching is conducted between backward and forward image blocks, in which the difference of standard deviation maps is used for the similarity measure. To deal with the multi-height problem of podium buildings, the gray difference image is adopted to segment podium buildings, and re-matching is conducted to find out their actual heights. Third, the building height is obtained through the elevation difference between the building top and bottom, in which the evaluation of the building bottom is calculated according to the elevation histogram statistics of the building buffer in DSM. Finally, two GF-7 stereo satellite images, collected in Yingde, Guangzhou, and Xi’an, Shanxi, are used for performance evaluation. Besides, the aerial LiDAR point cloud is used for absolute accuracy evaluation. The results demonstrate that compared with other methods, our solution obviously improves the accuracy of height estimation of high-rise buildings. The MAE (Mean Absolute Error) of the estimated building heights in Yingde is 2.31 m, and the MAE of the estimated elevation of building top and bottom is approximately 1.57 m and 1.91 m, respectively. Then the RMSE (Root Mean Square Error) of building top and bottom is 2.01 m and 2.57 m. As for the Xi’an dataset with 7 buildings with podium out of 40 buildings, the MAE of the estimated building height is 1.69 m and the RMSE is 2.34 m. The proposed method can be an effective solution for building height extraction from GF-7 satellite images.