Point Cloud Clustering Research Articles

Structural Changes in Boreal Forests Can Be Quantified Using Terrestrial Laser Scanning

Terrestrial laser scanning (TLS) has been adopted as a feasible technique to digitize trees and forest stands, providing accurate information on tree and forest structural attributes. However, there is limited understanding on how a variety of forest structural changes can be quantified using TLS in boreal forest conditions. In this study, we assessed the accuracy and feasibility of TLS in quantifying changes in the structure of boreal forests. We collected TLS data and field reference from 37 sample plots in 2014 (T1) and 2019 (T2). Tree stems typically have planar, vertical, and cylindrical characteristics in a point cloud, and thus we applied surface normal filtering, point cloud clustering, and RANSAC-cylinder filtering to identify these geometries and to characterize trees and forest stands at both time points. The results strengthened the existing knowledge that TLS has the capacity to characterize trees and forest stands in space and showed that TLS could characterize structural changes in time in boreal forest conditions. Root-mean-square-errors (RMSEs) in the estimates for changes in the tree attributes were 0.99–1.22 cm for diameter at breast height (Δdbh), 44.14–55.49 cm2 for basal area (Δg), and 1.91–4.85 m for tree height (Δh). In general, tree attributes were estimated more accurately for Scots pine trees, followed by Norway spruce and broadleaved trees. At the forest stand level, an RMSE of 0.60–1.13 cm was recorded for changes in basal area-weighted mean diameter (ΔDg), 0.81–2.26 m for changes in basal area-weighted mean height (ΔHg), 1.40–2.34 m2/ha for changes in mean basal area (ΔG), and 74–193 n/ha for changes in the number of trees per hectare (ΔTPH). The plot-level accuracy was higher in Scots pine-dominated sample plots than in Norway spruce-dominated and mixed-species sample plots. TLS-derived tree and forest structural attributes at time points T1 and T2 differed significantly from each other (p < 0.05). If there was an increase or decrease in dbh, g, h, height of the crown base, crown ratio, Dg, Hg, or G recorded in the field, a similar outcome was achieved by using TLS. Our results provided new information on the feasibility of TLS for the purposes of forest ecosystem growth monitoring.

Point cloud clustering and outlier detection based on spatial neighbor connected region labeling

Clustering analysis is one of the most important techniques in point cloud processing, such as registration, segmentation, and outlier detection. However, most of the existing clustering algorithms exhibit a low computational efficiency with the high demand for computational resources, especially for large data processing. Sometimes, clusters and outliers are inseparable, especially for those point clouds with outliers. Most of the cluster-based algorithms can well identify cluster outliers but sparse outliers. We develop a novel clustering method, called spatial neighborhood connected region labeling. The method defines spatial connectivity criterion, finds points connections based on the connectivity criterion among the k-nearest neighborhood region and classifies connected points to the same cluster. Our method can accurately and quickly classify datasets using only one parameter k. Comparing with K-means, hierarchical clustering and density-based spatial clustering of applications with noise methods, our method provides better accuracy using less computational time for data clustering. For applications in the outlier detection of the point cloud, our method can identify not only cluster outliers, but also sparse outliers. More accurate detection results are achieved compared to the state-of-art outlier detection methods, such as local outlier factor and density-based spatial clustering of applications with noise.

Using color and 3D geometry features to segment fruit point cloud and improve fruit recognition accuracy

Fruit image segmentation is an essential step to distinguish fruits from the background. In order to improve the fruits recognition accuracy for harvesting robots in three-dimensional (3D) space, a method with the fusion of color and 3D geometry features for fruit point cloud segmentation was proposed in this study. The local descriptor was applied to obtain the candidate regions, and the global descriptor was used to obtain the final segmented results. Firstly, the hue, saturation, value (HSV) color features and normal orientation features of pixels were fused to obtain the preliminary segmentation results. Then, the pre-processed color image and depth image were converted to a point cloud, and it was clustered into multiple regions by the Euclidean clustering algorithm. Finally, we utilized the viewpoint feature histogram (VFH) of each point cloud cluster to remove the remaining non-fruit regions. The experiments showed that the segmentation accuracy of the proposed method was 98.99%, and the precision was 80.09%, which are both superior to the traditional color segmentation methods. In addition, a fruit detection method based on shape analysis showed that it is more effective in improving fruit recognition rate and reducing false detection rate than the color segmentation methods.

Incremental Instance-Oriented 3D Semantic Mapping via RGB-D Cameras for Unknown Indoor Scene

Scene parsing plays a crucial role when accomplishing human-robot interaction tasks. As the “eye” of the robot, RGB-D camera is one of the most important components for collecting multiview images to construct instance-oriented 3D environment semantic maps, especially in unknown indoor scenes. Although there are plenty of studies developing accurate object-level mapping systems with different types of cameras, these methods either process the instance segmentation problem in completed mapping or suffer from a critical real-time issue due to heavy computation processing required. In this paper, we propose a novel method to incrementally build instance-oriented 3D semantic maps directly from images acquired by the RGB-D camera. To ensure an efficient reconstruction of 3D objects with semantic and instance IDs, the input RGB images are operated by a real-time deep-learned object detector. To obtain accurate point cloud cluster, we adopt the Gaussian mixture model as an optimizer after processing 2D to 3D projection. Next, we present a data association strategy to update class probabilities across the frames. Finally, a map integration strategy fuses information about their 3D shapes, locations, and instance IDs in a faster way. We evaluate our system on different indoor scenes including offices, bedrooms, and living rooms from the SceneNN dataset, and the results show that our method not only builds the instance-oriented semantic map efficiently but also enhances the accuracy of the individual instance in the scene.

A Fast Spatial Clustering Method for Sparse LiDAR Point Clouds Using GPU Programming.

Fast and accurate obstacle detection is essential for accurate perception of mobile vehicles’ environment. Because point clouds sensed by light detection and ranging (LiDAR) sensors are sparse and unstructured, traditional obstacle clustering on raw point clouds are inaccurate and time consuming. Thus, to achieve fast obstacle clustering in an unknown terrain, this paper proposes an elevation-reference connected component labeling (ER-CCL) algorithm using graphic processing unit (GPU) programing. LiDAR points are first projected onto a rasterized x–z plane so that sparse points are mapped into a series of regularly arranged small cells. Based on the height distribution of the LiDAR point, the ground cells are filtered out and a flag map is generated. Next, the ER-CCL algorithm is implemented on the label map generated from the flag map to mark individual clusters with unique labels. Finally, obstacle labeling results are inverse transformed from the x–z plane to 3D points to provide clustering results. For real-time 3D point cloud clustering, ER-CCL is accelerated by running it in parallel with the aid of GPU programming technology.

Indoor Point Cloud Segmentation Using Iterative Gaussian Mapping and Improved Model Fitting

Indoor scene segmentation based on 3-D laser point cloud is important for rebuilding and classification, especially for permanent building structure. However, the existing segmentation methods mainly focus on the large-scale planar structures but ignore the other sharp structures and details, which would cause accuracy degradation in scene reconstruction. To handle this issue, an iterative Gaussian mapping-based segmentation strategy has been proposed in this article, which goes from rough segmentation to refined one iteratively to decompose the indoor scene into detectable point cloud clusters layer by layer. An improved model fitting algorithm based on the maximum likelihood estimation sampling consensus (MLESAC) algorithm is proposed for refined segmentation, which is called the Prior-MLESAC algorithm, to deal with the extraction of both vertical and nonvertical planar and cylindrical structures. The experimental results demonstrate that planar and cylindrical structures are segmented more completely by the proposed strategy, and more details of the indoor structure are restored than other existing methods.

Point cloud registration for agriculture and forestry crops based on calibration balls using Kinect V2

For the process of point cloud registration, and the problem of inaccurate registration due to errors in correspondence between keypoints. In this paper, a registration method based on calibration balls was proposed, the trunk, branch, and crown were selected as experimental objects, and three calibration balls were randomly placed around the experimental objects to ensure different distances between two ball centers. Using the Kinect V2 depth camera to collect the point cloud of the experimental scene from four different viewpoints, the PassThrough filter algorithm was used for point cloud filtering in each view of the experimental scenes. The Euclidean cluster extraction algorithm was employed for point cloud clustering and segmentation to extract the experimental object and the calibration ball. The random sample consensus (RANSAC) algorithm was applied to fit the point cloud of a ball and calculate the coordinates of the ball center so that the distance between two ball centers under different viewpoints can be obtained by using the coordinates of the ball center. Comparing the distance between the ball centers from different viewpoints to determine the corresponding relationship between the ball centers from different viewpoints, and then using the singular value decomposition (SVD) method, the initial registration matrix was obtained. Finally, Iterative Closest Point (ICP) and its improved algorithm were used for accurate registration. The experimental results showed that the method of point cloud registration based on calibration balls can solve the problem of corresponding error of keypoints, and can register point clouds from different viewpoints of the same object. The registration method was evaluated by using the registration running time and the fitness score. The final registration running time of different experimental objects was not more than 6.5 s. The minimum fitness score of the trunk was approximately 0.0001, the minimum fitness score of the branch was approximately 0.0001, and the minimum fitness score of the crown was approximately 0.0006. Keywords: point cloud registration, calibration balls, Kinect V2, ICP DOI: 10.25165/j.ijabe.20201301.5077 Citation: Zhou S Z, Kang F, Li W B, Kan J M, Zheng Y J. Point cloud registration for agriculture and forestry crops based on calibration balls using Kinect V2. Int J Agric & Biol Eng, 2020; 13(1): 198–205.

Left ventricle landmark localization and identification in cardiac MRI by deep metric learning-assisted CNN regression

Accurate left ventricle landmark localization in cardiac MRI plays a vital role in computer-aided diagnosis of heart disease. Typical classification models hardly deal with artifacts and low discrimination of landmark regions by using only local image information, while regression models suffer from ambiguity between random samples and label, also the imbalance of samples. To overcome this limitation, this paper proposes a left ventricle landmark localization and identification method in cardiac MRI based on deep distance metric learning and CNN (convolutional neural network) regression. The method includes sample generation and regression stages, where super-pixel over-segmentation and unsupervised deep metric learning are integrated to extract the embedding information of local images, then a dual-channel salient sample mining module is designed by integrating specific super-pixel patches and grid patches extracted by the embedded triplet network. The weights of this triplet network are fed to build the CNN regression model, and the salient samples are employed to predict the landmark coordinate point cloud clusters. Furthermore, the centroid of each point cloud cluster is refined as landmark by Mean Shift iteration. The proposed method and close related methods were thoroughly evaluated on the CAP (cardiac atlas project) data set, and experimental results show that the proposed method achieves the competitive accuracy and outperforms the state-of-the-art classification and regression-based models.

Constant Size Point Cloud Clustering: A Compact, Non-Overlapping Solution

Point clouds have recently become a popular 3D representation model for many application domains, notably virtual and augmented reality. Since point cloud data is often very large, processing a point cloud may require that it be segmented into smaller clusters. For example, the input to deep learning-based methods like auto-encoders should be constant size point cloud clusters, which are ideally compact and non-overlapping. However, given the unorganized nature of point clouds, defining the specific data segments to code is not always trivial. This paper proposes a point cloud clustering algorithm which targets five main goals: i) clusters with a constant number of points; ii) compact clusters, i.e., with low dispersion; iii) non-overlapping clusters, i.e., not intersecting each other; iv) ability to scale with the number of points; and v) low complexity. After appropriate initialization, the proposed algorithm transfers points between neighboring clusters as a propagation wave, filling or emptying clusters until they achieve the same size. The proposed algorithm is unique since there is no other point cloud clustering method available in the literature offering the same clustering features for large point clouds at such low complexity.

High-Speed Phase-Shifting 3D Profilometry on Human Face Assisted by Statistical Model

Phase-shifting 3D profilometry is of high precision, but its performance could be far below expectations due to its long exposure time in the measurement of dynamic objects, especially human faces. In this article, a new idea is proposed to improve the measuring speed by providing reference derived from the statistical characteristics of human faces for the reconstruction of faces measured. During the measurement, no Gray code or low-frequency fringe is required, which reduces the number of patterns projected and sharply shortens the measuring time. Assisted by face detection and point cloud clustering, several candidate faces of different orders can be generated from the wrapped phase. Afterwards, the most probable candidate is selected according to the probability model of human face to obtain the correct result. Furthermore, the applicable conditions of our method have been analyzed, and the validity has been proved by experiments on actual human faces as well.

Point cloud registration for agriculture and forestry crops based on calibration balls using Kinect V2

For the process of point cloud registration, and the problem of inaccurate registration due to errors in correspondence between keypoints. In this paper, a registration method based on calibration balls was proposed, the trunk, branch, and crown were selected as experimental objects, and three calibration balls were randomly placed around the experimental objects to ensure different distances between two ball centers. Using the Kinect V2 depth camera to collect the point cloud of the experimental scene from four different viewpoints, the PassThrough filter algorithm was used for point cloud filtering in each view of the experimental scenes. The Euclidean cluster extraction algorithm was employed for point cloud clustering and segmentation to extract the experimental object and the calibration ball. The random sample consensus (RANSAC) algorithm was applied to fit the point cloud of a ball and calculate the coordinates of the ball center so that the distance between two ball centers under different viewpoints can be obtained by using the coordinates of the ball center. Comparing the distance between the ball centers from different viewpoints to determine the corresponding relationship between the ball centers from different viewpoints, and then using the singular value decomposition (SVD) method, the initial registration matrix was obtained. Finally, Iterative Closest Point (ICP) and its improved algorithm were used for accurate registration. The experimental results showed that the method of point cloud registration based on calibration balls can solve the problem of corresponding error of keypoints, and can register point clouds from different viewpoints of the same object. The registration method was evaluated by using the registration running time and the fitness score. The final registration running time of different experimental objects was not more than 6.5 s. The minimum fitness score of the trunk was approximately 0.0001, the minimum fitness score of the branch was approximately 0.0001, and the minimum fitness score of the crown was approximately 0.0006. Keywords: point cloud registration, calibration balls, Kinect V2, ICP DOI: 10.25165/j.ijabe.20201301.5077 Citation: Zhou S Z, Kang F, Li W B, Kan J M, Zheng Y J. Point cloud registration for agriculture and forestry crops based on calibration balls using Kinect V2. Int J Agric & Biol Eng, 2020; 13(1): 198–205.

Three-dimensional photogrammetric mapping of cotton bolls in situ based on point cloud segmentation and clustering

Three-dimensional high throughput plant phenotyping techniques provide an opportunity to measure plant organ-level traits which can be highly useful to plant breeders. The number and locations of cotton bolls, which are the fruit of cotton plants and an important component of fiber yield, are arguably among the most important phenotypic traits but are complex to quantify manually. Hence, there is a need for effective and efficient cotton boll phenotyping solutions to support breeding research and monitor the crop yield leading to better production management systems. We developed a novel methodology for 3D cotton boll mapping within a plot in situ. Point clouds were reconstructed from multi-view images using the structure from motion algorithm. The method used a region-based classification algorithm that successfully accounted for noise due to sunlight. The developed density-based clustering method could estimate boll counts for this situation, in which bolls were in direct contact with other bolls. By applying the method to point clouds from 30 plots of cotton plants, boll counts, boll volume and position data were derived. The average accuracy of boll counting was up to 90% and the R2 values between fiber yield and boll number, as well as fiber yield and boll volume were 0.87 and 0.66, respectively. The 3D boll spatial distribution could also be analyzed using this method. This method, which was low-cost and provided improved site-specific data on cotton bolls, can also be applied to other plant/fruit mapping analysis after some modification.

Multiview infrared target detection and localization

Infrared (IR) images are not affected by factors such as illumination and have the ability to work all day long, which is of great significance for night detection of unmanned platforms. We propose a multiview infrared target detection and localization algorithm (MVIDL), a complete sensory-fusion framework that uses IR images and lidar point cloud to detect and locate infrared targets (pedestrian and vehicle). MVIDL is a two-stage pipeline with an IR camera and three-dimensional lidar information as input. First, we introduce an infrared region proposal method that fuses lidar point cloud cluster results and IR image cluster results to obtain target regions and their position. In the second stage, an aggregate feature is proposed and extracted from the target regions, after which SVM is adopted to classify. Experimental results demonstrate that this algorithm can effectively detect targets and precisely get their position and size.

Seedling-lump integrated non-destructive monitoring for automatic transplanting with Intel RealSense depth camera

Non-destructive plant growth parameters measurement is an important concern in automatic-seedling transplanting. Recently, several image-based monitoring approaches have been proposed and potentially developed for several agricultural applications. The presented study proposed and developed a RealSense-based machine vision system for the close-shot seedling-lump integrated monitoring. The strategy was based on the close-shot depth information. Further, the point cloud clustering and suitable algorithms were applied to obtain the segmentation of 3D seedling models. In addition, the data processing pipeline was developed to assess the different morphological parameter of 4 different seedling varieties. The experiments were carried out with 4 different seedling varieties (pepper, tomato, cucumber, and lettuce) and trained under different light conditions (light and dark). Moreover, analysis results showed that there was not significantly different (p < 0.05) found towards light and dark environments due to close-shot near-infrared detection. However, the results revealed that the stem diameter relationship between RealSense and the manual method was found for R2 = 0.68 cucumber, R2 = 0.54 tomato, R2 = 0.35 pepper, and R2 = 0.58 lettuce seedlings. Whereas, the seedling height relationship between RealSense and the manual method was found higher than R2 = 0.99, 0.99, 0.99, and 0.99 for pepper, tomato, cucumber, and lettuce, respectively. Based on the experiment results, it was concluded that the RGB-D integrated monitoring system with the purposed method could be practiced for nursery seedlings most promisingly without high labour requirements in terms of ease of use. The system revealed a good sturdiness and relevance for plant growth monitoring. Additionally, it has the perspective for future practical value to real-time vision servo operations for transplanting robots.

Online learning for 3D LiDAR-based human detection: experimental analysis of point cloud clustering and classification methods

This paper presents a system for online learning of human classifiers by mobile service robots using 3D LiDAR sensors, and its experimental evaluation in a large indoor public space. The learning framework requires a minimal set of labelled samples (e.g. one or several samples) to initialise a classifier. The classifier is then retrained iteratively during operation of the robot. New training samples are generated automatically using multi-target tracking and a pair of “experts” to estimate false negatives and false positives. Both classification and tracking utilise an efficient real-time clustering algorithm for segmentation of 3D point cloud data. We also introduce a new feature to improve human classification in sparse, long-range point clouds. We provide an extensive evaluation of our the framework using a 3D LiDAR dataset of people moving in a large indoor public space, which is made available to the research community. The experiments demonstrate the influence of the system components and improved classification of humans compared to the state-of-the-art.

IMAGE-BASED VEHICLE TRACKING FROM ROADSIDE LIDAR DATA

Abstract. Vehicle tracking is of great importance in urban traffic systems, and the adoption of lidar technologies – including on-board and roadside systems – has significant potential for such applications. This research therefore proposes and develops an image-based vehicle-tracking framework from roadside lidar data to track the precise location and speed of a vehicle. Prior to tracking, vehicles are detected in point clouds through a three-step procedure. Cluster tracking then provides initial tracking results. The second tracking stage aims to provide more precise results, in which two strategies are developed and tested: frame-by-frame and model-matching strategies. For each strategy, tracking is implemented through two threads by converting the 3D point cloud clusters into 2D images relating to the plan and side views along the tracked vehicle’s trajectory. During this process, image registration is exploited in order to retrieve the transformation parameters between every image pair. Based on these transformations, vehicle speeds are determined directly based on (a) the locations of the chosen tracking point in the first strategy; (b) a vehicle model is built and tracking point locations can be calculated after matching every frame with the model in the second strategy. In contrast with other existing methods, the proposed method provides improved vehicle tracking via points instead of clusters. Moreover, tracking in a decomposed manner provides an opportunity to cross-validate the results from different views. The effectiveness of this method has been evaluated using roadside lidar data obtained by a Robosense 32-line laser scanner.

An Accurate Skeleton Extraction Approach From 3D Point Clouds of Maize Plants.

Accurate and high-throughput determination of plant morphological traits is essential for phenotyping studies. Nowadays, there are many approaches to acquire high-quality three-dimensional (3D) point clouds of plants. However, it is difficult to estimate phenotyping parameters accurately of the whole growth stages of maize plants using these 3D point clouds. In this paper, an accurate skeleton extraction approach was proposed to bridge the gap between 3D point cloud and phenotyping traits estimation of maize plants. The algorithm first uses point cloud clustering and color difference denoising to reduce the noise of the input point clouds. Next, the Laplacian contraction algorithm is applied to shrink the points. Then the key points representing the skeleton of the plant are selected through adaptive sampling, and neighboring points are connected to form a plant skeleton composed of semantic organs. Finally, deviation skeleton points to the input point cloud are calibrated by building a step forward local coordinate along the tangent direction of the original points. The proposed approach successfully generates accurately extracted skeleton from 3D point cloud and helps to estimate phenotyping parameters with high precision of maize plants. Experimental verification of the skeleton extraction process, tested using three cultivars and different growth stages maize, demonstrates that the extracted matches the input point cloud well. Compared with 3D digitizing data-derived morphological parameters, the NRMSE of leaf length, leaf inclination angle, leaf top length, leaf azimuthal angle, leaf growth height, and plant height, estimated using the extracted plant skeleton, are 5.27, 8.37, 5.12, 4.42, 1.53, and 0.83%, respectively, which could meet the needs of phenotyping analysis. The time required to process a single maize plant is below 100 s. The proposed approach may play an important role in further maize research and applications, such as genotype-to-phenotype study, geometric reconstruction, functional structural maize modeling, and dynamic growth animation.

A posture measurement approach for an articulated manipulator by RGB-D cameras

The objective of this article aims at the safety problems where robots and operators are highly coupled in a working space. A method to model an articulated robot manipulator by cylindrical geometries based on partial cloud points is proposed in this article. Firstly, images with point cloud data containing the posture of a robot with five resolution links are captured by a pair of RGB-D cameras. Secondly, the process of point cloud clustering and Gaussian noise filtering is applied to the images to separate the point cloud data of three links from the combined images. Thirdly, an ideal cylindrical model fits the processed point cloud data are segmented by the random sample consensus method such that three joint angles corresponding to three major links are computed. The original method for calculating the normal vector of point cloud data is the cylindrical model segmentation method, but the accuracy of posture measurement is low when the point cloud data is incomplete. To solve this problem, a principal axis compensation method is proposed, which is not affected by number of point cloud cluster data. The original method and the proposed method are used to estimate the three joint angular of the manipulator system in experiments. Experimental results show that the average error is reduced by 27.97%, and the sample standard deviation of the error is reduced by 54.21% compared with the original method for posture measurement. The proposed method is 0.971 piece/s slower than the original method in terms of the image processing velocity. But the proposed method is still feasible, and the purpose of posture measurement is achieved.

A 3D Point Cloud Filtering Method for Leaves Based on Manifold Distance and Normal Estimation

Leaves are used extensively as an indicator in research on tree growth. Leaf area, as one of the most important index in leaf morphology, is also a comprehensive growth index for evaluating the effects of environmental factors. When scanning tree surfaces using a 3D laser scanner, the scanned point cloud data usually contain many outliers and noise. These outliers can be clusters or sparse points, whereas the noise is usually non-isolated but exhibits different attributes from valid points. In this study, a 3D point cloud filtering method for leaves based on manifold distance and normal estimation is proposed. First, leaf was extracted from the tree point cloud and initial clustering was performed as the preprocessing step. Second, outlier clusters filtering and outlier points filtering were successively performed using a manifold distance and truncation method. Third, noise points in each cluster were filtered based on the local surface normal estimation. The 3D reconstruction results of leaves after applying the proposed filtering method prove that this method outperforms other classic filtering methods. Comparisons of leaf areas with real values and area assessments of the mean absolute error (MAE) and mean absolute error percent (MAE%) for leaves in different levels were also conducted. The root mean square error (RMSE) for leaf area was 2.49 cm2. The MAE values for small leaves, medium leaves and large leaves were 0.92 cm2, 1.05 cm2 and 3.39 cm2, respectively, with corresponding MAE% values of 10.63, 4.83 and 3.8. These results demonstrate that the method proposed can be used to filter outliers and noise for 3D point clouds of leaves and improve 3D leaf visualization authenticity and leaf area measurement accuracy.

Semi-automatic LiDAR point cloud denoising using a connected-component labelling method

The Smart City concept requires new, fast methods for collection of 3-D data representing features of urban landscape. Laser scanning technology (LiDAR - Light Detection and Ranging) enables such approach producing dense 3-D point clouds of millions of points, which, however, contain noise. Therefore, we developed a new approach allowing for a semi-automatic elimination of data noise resulting from motion of objects within the scanned scene such as persons. We used a connected-component labelling method to filter out the noise points from terrestrial laser scanning point clouds. Our approach was based on a step-by-step object classification with a proper parameterisation. In the first step, all points located close to the predicted terrain were selected. In the second step, the points representing the terrain and floor were classified using the surface filter tool implemented in the RiScan Pro software by RIEGL. The rest of points were classified using point cloud clustering via the connected-component labelling method implemented in the CloudCompare software. In the final step, the operator manually decides whether the point cluster represents the noise. The method was applied to the Cathedral of Saint Elizabeth, a sacral object located in the historical centre of the city of Košice in Slovakia during normal operating hours. We managed to capture approximately 80% of the data noise in total. The method provides a better flexibility in surveying overcrowded city locations using the laser scanning technology.