Unsupervised Building Instance Segmentation of Airborne LiDAR Point Clouds for Parallel Reconstruction Analysis

Efficient building instance segmentation is necessary for many applications such as parallel reconstruction, management and analysis. However, most of the existing instance segmentation methods still suffer from low completeness, low correctness and low quality for building instance segmentation, which are especially obvious for complex building scenes. This paper proposes a novel unsupervised building instance segmentation (UBIS) method of airborne Light Detection and Ranging (LiDAR) point clouds for parallel reconstruction analysis, which combines a clustering algorithm and a novel model consistency evaluation method. The proposed method first divides building point clouds into building instances by the improved kd tree 2D shared nearest neighbor clustering algorithm (Ikd-2DSNN). Then, the geometric feature of the building instance is obtained using the model consistency evaluation method, which is used to determine whether the building instance is a single building instance or a multi-building instance. Finally, for multiple building instances, the improved kd tree 3D shared nearest neighbor clustering algorithm (Ikd-3DSNN) is used to divide multi-building instances again to improve the accuracy of building instance segmentation. Our experimental results demonstrate that the proposed UBIS method obtained good performances for various buildings in different scenes such as high-rise building, podium buildings and a residential area with detached houses. A comparative analysis confirms that the proposed UBIS method performed better than state-of-the-art methods.

HGC: fast hierarchical clustering for large-scale single-cell data

Clustering is a key step in revealing heterogeneities in single-cell data. Cell heterogeneity can be explored at different resolutions and the resulted varying cell states are inherently nested. However, most existing single-cell clustering methods output a fixed number of clusters without the hierarchical information. Classical hierarchical clustering provides dendrogram of cells, but cannot scale to large datasets due to the high computational complexity. We present HGC, a fast Hierarchical Graph-based Clustering method to address both problems. It combines the advantages of graph-based clustering and hierarchical clustering. On the shared nearest neighbor graph of cells, HGC constructs the hierarchical tree with linear time complexity. Experiments showed that HGC enables multiresolution exploration of the biological hierarchy underlying the data, achieves state-of-the-art accuracy on benchmark data, and can scale to large datasets. HGC is freely available for academic use at https://www.github.com/XuegongLab/[email protected], [email protected]

Fast Searching Density Peak Clustering Algorithm Based on Shared Nearest Neighbor and Adaptive Clustering Center

The clustering analysis algorithm is used to reveal the internal relationships among the data without prior knowledge and to further gather some data with common attributes into a group. In order to solve the problem that the existing algorithms always need prior knowledge, we proposed a fast searching density peak clustering algorithm based on the shared nearest neighbor and adaptive clustering center (DPC-SNNACC) algorithm. It can automatically ascertain the number of knee points in the decision graph according to the characteristics of different datasets, and further determine the number of clustering centers without human intervention. First, an improved calculation method of local density based on the symmetric distance matrix was proposed. Then, the position of knee point was obtained by calculating the change in the difference between decision values. Finally, the experimental and comparative evaluation of several datasets from diverse domains established the viability of the DPC-SNNACC algorithm.

Discovery of new stellar groups in the Orion complex

We test the ability of two unsupervised machine learning algorithms, EnLink and Shared Nearest Neighbor (SNN), to identify stellar groupings in the Orion star-forming complex as an application to the 5D astrometric data from Gaia DR2. The algorithms represent two distinct approaches to limiting user bias when selecting parameter values and evaluating the relative weights among astrometric parameters. EnLink adopts a locally adaptive distance metric and eliminates the need for parameter tuning through automation. The original SNN relies only on human input for parameter tuning so we modified SNN to run in two stages. We first ran the original SNN 7000 times, each with a randomly generated sample according to within-source co-variance matrices provided in Gaia DR2 and random parameter values within reasonable ranges. During the second stage, we modified SNN to identify the most repeating stellar groups from the 25 798 we obtained in the first stage. We recovered 22 spatially and kinematically coherent groups in the Orion complex, 12 of which were previously unknown. The groups show a wide distribution of distances extending as far as about 150 pc in front of the star-forming Orion molecular clouds, to about 50 pc beyond them, where we, unexpectedly, find several groups. Our results reveal the wealth of sub-structure in the OB association, within and beyond the classical Blaauw Orion OBI sub-groups. A full characterization of the new groups is essential as it offers the potential to unveil how star formation proceeds globally in large complexes such as Orion.

An Efficient Clustering Method for Hyperspectral Optimal Band Selection via Shared Nearest Neighbor

A hyperspectral image (HSI) has many bands, which leads to high correlation between adjacent bands, so it is necessary to find representative subsets before further analysis. To address this issue, band selection is considered as an effective approach that removes redundant bands for HSI. Recently, many band selection methods have been proposed, but the majority of them have extremely poor accuracy in a small number of bands and require multiple iterations, which does not meet the purpose of band selection. Therefore, we propose an efficient clustering method based on shared nearest neighbor (SNNC) for hyperspectral optimal band selection, claiming the following contributions: (1) the local density of each band is obtained by shared nearest neighbor, which can more accurately reflect the local distribution characteristics; (2) in order to acquire a band subset containing a large amount of information, the information entropy is taken as one of the weight factors; (3) a method for automatically selecting the optimal band subset is designed by the slope change. The experimental results reveal that compared with other methods, the proposed method has competitive computational time and the selected bands achieve higher overall classification accuracy on different data sets, especially when the number of bands is small.