Point Sampling Method Research Articles

Unmanned Aerial Vehicle-Neural Radiance Field (UAV-NeRF): Learning Multiview Drone Three-Dimensional Reconstruction with Neural Radiance Field

In traditional 3D reconstruction using UAV images, only radiance information, which is treated as a geometric constraint, is used in feature matching, allowing for the restoration of the scene’s structure. After introducing radiance supervision, NeRF can adjust the geometry in the fixed-ray direction, resulting in a smaller search space and higher robustness. Considering the lack of NeRF construction methods for aerial scenarios, we propose a new NeRF point sampling method, which is generated using a UAV imaging model, compatible with a global geographic coordinate system, and suitable for a UAV view. We found that NeRF is optimized entirely based on the radiance while ignoring the direct geometry constraint. Therefore, we designed a radiance correction strategy that considers the incidence angle. Our method can complete point sampling in a UAV imaging scene, as well as simultaneously perform digital surface model construction and ground radiance information recovery. When tested on self-acquired datasets, the NeRF variant proposed in this paper achieved better reconstruction accuracy than the original NeRF-based methods. It also reached a level of precision comparable to that of traditional photogrammetry methods, and it is capable of outputting a surface albedo that includes shadow information.

A many-objective evolutionary algorithm combining simplified hypervolume and a method for reference point sampling based on angular relationship

Determining the path for an evolutionary algorithm is crucial for its performance. Current methods for sampling reference points guiding evolutionary algorithms are inadequate for dealing with convex and concave Pareto fronts,and the uniformity of sampling results decreases significantly in high-dimensional spaces. In this paper,we propose a reference point sampling method based on angular relationships to tackle these issue. And we propose the concept of optimally distributed individuals based on the IGD indicator to ensure the distribution of the evolutionary process and prevent the algorithm from converging to local optima. Additionally,we introduce a novel method for calculating individuals’ fitness within the population,ensuring convergence,uniformity,and distribution of the evolutionary algorithm,thereby enhancing selection pressure among non-dominated individuals. Experimental results on diverse benchmark test problems demonstrate that the proposed algorithm competes favorably with six advanced evolutionary algorithms for many-objective optimization.

An Investigation into the Pole–Slot Ratio and Optimization of a Low-Speed and High-Torque Permanent Magnet Motor

At present, low-speed high-torque permanent magnet motors are widely used in the sampling industry, manufacturing industry and energy industry. However, the research on low-speed high-torque permanent magnet motors is far from enough. The primary difficulty in the initial design of low-speed high-torque permanent magnet motors is the selection of pole–slot ratio. The pole–slot ratio has a great influence on the electromagnetic performance such as torque ripple and the maximum output torque of low-speed motors. Choosing the appropriate pole–slot ratio scheme can make the design of a low-speed motor more efficient. In addition, the optimization design of the motor is also a necessary process. At present, there are many studies on optimization algorithms. However, the research on sample point sampling and surrogate model fitting is not enough. Choosing the appropriate sample point sampling method and surrogate model fitting method can help one obtain a more accurate surrogate model, which lays a foundation for the optimization of the motor. Based on the above analysis, this paper first selects four representative pole–slot ratio schemes for comprehensive comparison of their electromagnetic performances. Secondly, two sample point sampling methods and three surrogate model fitting methods are combined to obtain six surrogate models, and the accuracy of the six surrogate models is compared and analyzed. Finally, a 37kW,160rpm prototype is made, and the comparison of the surrogate model optimization prediction results, the finite element simulation calculation results and the measured results is carried out to further prove the accuracy of the selected surrogate model. The work performed in this paper provides a certain reference value for the initial design and optimization experiment design of low-speed high-torque permanent magnet motor.

Research on Lightweight Method of Segment Beam Point Cloud Based on Edge Detection Optimization

In order to reduce the loss of laser point cloud appearance contours by point cloud lightweighting, this paper takes the laser point cloud data of the segment beam of the expressway viaduct as a sample. After comparing the downsampling algorithm from many aspects and angles, the voxel grid method is selected as the basic theory of the research. By combining the characteristics of the normal vector data of the laser point cloud, the top surface point cloud edge data are extracted and the voxel grid method is fused to establish an optimized point cloud lightweighting algorithm. The research in this paper shows that the voxel grid method performs better than the furthest point sampling method and the curvature downsampling method in retaining the top surface data, reducing the calculation time and optimizing the edge contour. Moreover, the average offset of the geometric contour is reduced from 2.235 mm to 0.664 mm by the edge-optimized voxel grid method, which has a higher retention. In summary, the edge-optimized voxel grid method has a better effect than the existing methods in point cloud lightweighting.

Evaluation of the point sampling method and inter-comparison of remote emission sensing systems for screening real-world car emissions

Emissions from internal combustion vehicles are currently not properly monitored throughout their life cycle. Remote emission sensing (RES) is a technology that can measure emissions under real driving conditions without contact. Current light extinction based RES systems are capable of providing emission factors for various gases, but lack accuracy for particulate matter (PM). Point Sampling (PS) is an extraction-based RES technique that can measure gases as well as various particle metrics such as black carbon or particle number. In this work, we evaluated the performance of a recently developed PS system and the state-of-the-art light extinction based remote sensing devices EDAR (HEAT) and ORSD (OPUS RSE) during co-location measurements. Validation measurements with portable emission measurement systems and emissions screening of several thousand cars in three European cities provide detailed insights into system's performance. Meteorological evaluations showed that the PS capture rate is strongly influenced by wind, but no other weather influences were found. Both light extinction based systems are unable to measure during rain. We found that all three systems tested were capable of screening NOx emissions from pre-Euro 6 diesel cars. Measurement results show the ability of the PS system to quantify high and low PM emitters equally well. The open-path RES systems (EDAR, ORSD) are capable of estimating PM emissions from pre-Euro 5 diesel cars. However, deficiencies of open-path RES systems are evident in the quantification of PM emissions from newer engine technologies (diesel Euro 5 and beyond) and from petrol cars. The PS system has a 2 to 5 times lower capture rate than open-path RES systems, but the PS measurement results are more accurate (more than 5 times for PM and more than 1.35 times for NOx). The good accuracy of individual measurements makes PS a powerful tool for reliable high emitter identification.

Soil Improvement after Reclamation of Rare Earth Tailings in Northeast Guangdong, China

In order to explore the soil improvement effect of guest soil reclamation, the main physical and chemical properties of the soil in-situ extracted rare earth tailings after reclamation in Pingyuan County, Meizhou city were tested. Using 5 point sampling method, soil samples were taken from a 0-20 cm soil layer to detect NH4+ residues, pH value, content of organic matter, available N, available K, available P, and heavy metal in the soil. The results were shown that the NH4+ content in the soil decreased from 96.67 to 15.29 (mg·kg−1) after reclamation, the pH value increased from 4.58 to 5.40, and the organic matter increased from 2.12 to 2.75 (g·kg−1). However, the available N, P, and K decreased, and the soil fertility index also decreased from 0.55 to 0.49. The reclamation did not significantly improve soil fertility. The Nemero comprehensive index of non-rare earth heavy metal elements decreased from 0.58 to 0.44, but the pollution of non-rare earth heavy metals in the soil before and after reclamation was at a clean level, and there was no heavy metal pollution. The total RE (rare earth) content in tailings soil, reclaimed soil, and control soil was higher than the soil environmental background value in Guangdong Province. Heavy RE content in tailings soil exceeded 50% of total RE content, while light RE was mainly in the reclaimed soil. The pollution levels of total RE metals in soil before and after reclamation were non-pollution to moderate pollution level and non-pollution level, respectively. The control soil was the most severely polluted, reaching a strong pollution level. The three RE metals, Y, Sm and Ce, accounted for the highest proportion of the total soil REs, and the pollution levels of the latter two elements were above strong pollution in control soil. After 4 years of land reclamation, the NH4+-N pollution, soil acidification, and heavy metal pollution in the soil of tailings could be significantly improved, but the soil fertility was still low and required a long remediation process. In the future reclamation work, in addition to continuing to pay attention to the improvement of soil environment, the improvement of soil fertility and the control of RE pollution should also be strengthened, and when using foreign soil for reclamation and improvement, the quality of foreign soil, especially fertility, should also be considered.

IFM calculator: An algorithm for interfragmentary motion calculation in finite element analysis

BackgroundInterfragmentary motion (IFM) is a complex state that significantly impacts the healing process of fractures following implant placement. It is crucial to fully consider the IFM state after implantation in the design and biomechanical testing of implants. However, current finite element analysis software lacks direct tools for calculating IFM, and existing IFM tools do not offer a comprehensive solution in terms of accuracy, functionality, and visualization. MethodsIn our study, we developed a Python-based algorithm for calculating IFM that addresses limitations. Our algorithm automatically calculated IFM distances, sliding distances, gaps, as well as the angles and rotation of the two fracture surfaces using all nodes on both sides of the fracture ends. Researchers could input data and selected desired parameters in the interface. The algorithm then performed the necessary calculations and presented the results in a clear and concise manner. The algorithm also provided comprehensive data export capabilities, allowing researchers to customize analyses based on specific needs.To provide a more intuitive demonstration of the calculation process and usage of IFM-Cal, we conducted simulations in Ansys using two rectangular blocks to compare the accuracy and function of three different methods (Point based method, contact tool and IFM-Cal). ResultsThe point-based method and the contact tool could not accurately calculate IFA, while IFM-Cal could provide a comprehensive evaluation of IFA. In simulation 1, the IFM distances calculated using the point sampling method, contact tool, and IFM-Cal were 2.00 mm, 3.15 mm, and 2.00 mm, respectively. In simulation 2, both the point sampling method and contact tool failed to calculate the interfragmentary angle (IFA), while the IFM-Cal algorithm estimated an angle of -7.87°, which had a small error compared to the ground-truth value of 7.9°. ConclusionWe have developed an algorithm for computing IFM which can be utilized in finite element analysis and biomechanical experiments. By conducting comparative simulations with other existing algorithms, we have demonstrated the superior accuracy and expanded evaluation capabilities of our algorithm.

Real-Time 3D Object Detection on Crowded Pedestrians.

In the field of autonomous driving, object detection under point clouds is indispensable for environmental perception. In order to achieve the goal of reducing blind spots in perception, many autonomous driving schemes have added low-cost blind-filling LiDAR on the side of the vehicle. Unlike point cloud target detection based on high-performance LiDAR, the blind-filling LiDARs have low vertical angular resolution and are mounted on the side of the vehicle, resulting in easily mixed point clouds of pedestrian targets in close proximity to each other. These characteristics are harmful for target detection. Currently, many research works focus on target detection under high-density LiDAR. These methods cannot effectively deal with the high sparsity of the point clouds, and the recall and detection accuracy of crowded pedestrian targets tend to be low. To overcome these problems, we propose a real-time detection model for crowded pedestrian targets, namely RTCP. To improve computational efficiency, we utilize an attention-based point sampling method to reduce the redundancy of the point clouds, then we obtain new feature tensors by the quantization of the point cloud space and neighborhood fusion in polar coordinates. In order to make it easier for the model to focus on the center position of the target, we propose an object alignment attention module (OAA) for position alignment, and we utilize an additional branch of the targets' location occupied heatmap to guide the training of the OAA module. These methods improve the model's robustness against the occlusion of crowded pedestrian targets. Finally, we evaluate the detector on KITTI, JRDB, and our own blind-filling LiDAR dataset, and our algorithm achieved the best trade-off of detection accuracy against runtime efficiency.

Probability propagation for faster and efficient point cloud segmentation using a neural network

Neural networks (NN) have shown promising performance in point cloud segmentation (PCS). However, the measured points are too numerous to be used as model input at once. It results in a long inference time and high computational cost due to iterative sampling and inference. This study proposes Probability Propagation (PP) as a stochastic upsampling method. PP propagates the predicted probability of a sampled part of a point cloud into the other unpredicted points by considering proximity. By replacing the iterative inference of NN with PP, large point clouds can be dealt with quickly and efficiently. We investigated the effectiveness of PP using the ShapeNet benchmark on various settings: sampling methods (random, farthest point, and Poisson disk sampling) with sampling ratios (5%, 10%, 20%, 39%, and 78%) for NN and the stochastic mapping conditions (uniform, linear, cosine, Gaussian, and exponential distributions) for PP. Using NN with PP achieved higher performance and faster inference speed than when using NN alone. For the farthest point sampling method of 5% sampling ratio, NN+PP improved the instance mIoU by 2.457%p with 102 times faster speed compared to that when using NN alone. The result indicates that PP can significantly contribute to the improvement of performance and efficiency in PCS when used in edge AI systems.

GPU-based discrete element model of realistic non-convex aggregates: Mesoscopic insights into ITZ volume fraction and diffusivity of concrete

Aggregates with realistic non-convex morphologies and their surrounding interfacial transition zones (ITZs) significantly affect the diffusivity of concrete. Herein, we initially propose a mathematical-controllable parameterized method to devise three-dimensional (3D) non-convex aggregates. Then, we develop a GPU-based discrete element modelling (DEM) to generate random close packing of polydisperse non-convex aggregates. Next, we establish three-phase concrete mesostructures containing non-convex aggregates and ITZs and obtain the ITZ volume fraction by a Monte Carlo random point sampling method. Finally, we develop a dual-probability-Brownian motion (DP-BM) scheme to obtain the effective diffusivity of concrete. Moreover, the effects of volume fraction, size polydispersity and shape of aggregates as well as ITZ thickness on the ITZ volume fraction and concrete diffusivity are researched to probe rigorous component-structure-property relations for concrete. This work provides a comprehensive numerical framework for mesostructure characterization and diffusivity evaluation for concrete, which can be a valuable tool for durability assessment of concrete.

Digitizing the coral reef: Machine learning of underwater spectral images enables dense taxonomic mapping of benthic habitats

Abstract Coral reefs are the most biodiverse marine ecosystems, and host a wide range of taxonomic diversity in a complex spatial community structure. Existing coral reef survey methods struggle to accurately capture the taxonomic detail within the complex spatial structure of benthic communities. We propose a workflow to leverage underwater hyperspectral image transects and two machine learning algorithms to produce dense habitat maps of 1150 m2 of reefs across the Curaçao coastline. Our multi‐method workflow labelled all 500+ million pixels with one of 43 classes at taxonomic family, genus or species level for corals, algae, sponges, or to substrate labels such as sediment, turf algae and cyanobacterial mats. With low annotation effort (only 2% of pixels) and no external data, our workflow enables accurate (Fbeta of 87%) survey‐scale mapping, with unprecedented thematic detail and with fine spatial resolution (2.5 cm/pixel). Our assessments of the composition and configuration of the benthic communities of 23 image transects showed high consistency. Digitizing the reef habitat and community structure enables validation and novel analysis of pattern and scale in coral reef ecology. Our dense habitat maps reveal the inadequacies of point sampling methods to accurately describe reef benthic communities.

A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks

Physics-informed neural networks (PINNs) have shown to be effective tools for solving both forward and inverse problems of partial differential equations (PDEs). PINNs embed the PDEs into the loss of the neural network using automatic differentiation, and this PDE loss is evaluated at a set of scattered spatio-temporal points (called residual points). The location and distribution of these residual points are highly important to the performance of PINNs. However, in the existing studies on PINNs, only a few simple residual point sampling methods have mainly been used. Here, we present a comprehensive study of two categories of sampling for PINNs: non-adaptive uniform sampling and adaptive nonuniform sampling. We consider six uniform sampling methods, including (1) equispaced uniform grid, (2) uniformly random sampling, (3) Latin hypercube sampling, (4) Halton sequence, (5) Hammersley sequence, and (6) Sobol sequence. We also consider a resampling strategy for uniform sampling. To improve the sampling efficiency and the accuracy of PINNs, we propose two new residual-based adaptive sampling methods: residual-based adaptive distribution (RAD) and residual-based adaptive refinement with distribution (RAR-D), which dynamically improve the distribution of residual points based on the PDE residuals during training. Hence, we have considered a total of 10 different sampling methods, including six non-adaptive uniform sampling, uniform sampling with resampling, two proposed adaptive sampling, and an existing adaptive sampling. We extensively tested the performance of these sampling methods for four forward problems and two inverse problems in many setups. Our numerical results presented in this study are summarized from more than 6000 simulations of PINNs. We show that the proposed adaptive sampling methods of RAD and RAR-D significantly improve the accuracy of PINNs with fewer residual points for both forward and inverse problems. The results obtained in this study can also be used as a practical guideline in choosing sampling methods.

Modelling flood susceptibility based on deep learning coupling with ensemble learning models

Modelling flood susceptibility is an indirect way to reduce the loss from flood disaster. Now, flood susceptibility modelling based on data driven model is state-of-the-art method such as ensemble learning and deep learning. However, the effect of deep learning coupling with ensemble learning models in flood susceptibility modelling is still unknown. Therefore, the aim of this paper is to propose three deep learning coupling with ensemble learning models by combining the deep learning (DL) with Filtered Classifier (FC), Rotation forest (RF) and Random Subspace (RSS) and explore the effect of coupling method for modelling flood susceptibility. The key step of this paper is as following: firstly, a Dingnan County which is lied in the Jiangxi Province of China is chosen as a case study, single flood event point and random sampling method was applied to generate the flood and non-flood data, respectively, then frequency ratio was utilized to analyze the relationship between each influencing factor and flood occurrence, based on the value of VIF, Spearman's correlation and One R classifier, the result show that there is no multicollinearity between each influencing factor, ten influencing factors have contribution to the flood occurrence and all of them are applied to construct the coupling model. Finally, the DL, FC-DL, RF-DL and RSS-DL were applied to produce flood susceptibility maps. Then, several statistical indexes such as area under the curve (AUC), Kappa index, accuracy (ACC), and F-measure were used to assess the accomplishment of these coupling models. For the train data, the FC-DL model acquired the highest AUC value (0.996), followed by RF-DL (0.944), RSS-DL (0.934), and DL (0.934). For the validation data, the result showed that all models have a good accomplishment (AUC>0.8). In a word, the deep learning coupling with ensemble learning models demonstrates the more reliable and excellent performance. Hence, the proposed new method will help the government for land use planning and can be applied in other area around the world.

DGPolarNet: Dynamic Graph Convolution Network for LiDAR Point Cloud Semantic Segmentation on Polar BEV

Semantic segmentation in LiDAR point clouds has become an important research topic for autonomous driving systems. This paper proposes a dynamic graph convolution neural network for LiDAR point cloud semantic segmentation using a polar bird’s-eye view, referred to as DGPolarNet. LiDAR point clouds are converted to polar coordinates, which are rasterized into regular grids. The points mapped onto each grid distribute evenly to solve the problem of the sparse distribution and uneven density of LiDAR point clouds. In DGPolarNet, a dynamic feature extraction module is designed to generate edge features of perceptual points of interest sampled by the farthest point sampling and K-nearest neighbor methods. By embedding edge features with the original point cloud, local features are obtained and input into PointNet to quantize the points and predict semantic segmentation results. The system was tested on the Semantic KITTI dataset, and the segmentation accuracy reached 56.5%

2D&3DHNet for 3D Object Classification in LiDAR Point Cloud

Accurate semantic analysis of LiDAR point clouds enables the interaction between intelligent vehicles and the real environment. This paper proposes a hybrid 2D and 3D Hough Net by combining 3D global Hough features and 2D local Hough features with a classification deep learning network. Firstly, the 3D object point clouds are mapped into the 3D Hough space to extract the global Hough features. The generated global Hough features are input into the 3D convolutional neural network for training global features. Furthermore, a multi-scale critical point sampling method is designed to extract critical points in the 2D views projected from the point clouds to reduce the computation of redundant points. To extract local features, a grid-based dynamic nearest neighbors algorithm is designed by searching the neighbors of the critical points. Finally, the two networks are connected to the full connection layer, which is input into fully connected layers for object classification.

Vehicle Delay Analysis at Signalized Intersection

Traffic delay is an important measure to evaluate road operation efficiency and service level, which can reflect the rationality of road design and intersection signal control scheme design. Therefore, delay analysis is of great significance to the evaluation of road service level and intersection signal control scheme design. In this paper, the point sample method and VISSIM simulation software are used to study the traffic delay of Yingbin Road and Century Road intersection in Jiaozuo City, and the traffic condition of the intersection is evaluated according to the calculation results. In this paper, the point sample method and VISSIM simulation software are used to study the traffic delay of Yingbin Road and Century Road intersection in Jiaozuo City, and the traffic condition of the intersection is evaluated according to the calculation results.

Point cloud sampling method based on offset-attention and mutual supervision

Point cloud sampling method based on offset-attention and mutual supervision

THE EFFECT OF THE POINT SAMPLING TO THE RESULT OF COORDINATE MEASURING OF FREE-FORM SURFACE

The coordinate measuring technique appropriates to measure dimensional and geometric properties of a machine part. The result of the measuring is effected by several parameters, like the measuring method, the point sampling technique, and the mathematical processing of the measured coordinates. The current article investigates the effect of the point sampling methods in case of a free-form surface. Two methods are compared: the uniform matrix method, and the Halton-Zaremba quasi-random method. The number of measured points is investigated also. The free-form test surface was produced by ball-end milling, and the radius, the cylindricity and the surface profile error were assessed.

Bird Assemblages in Coffee Agroforestry Systems and Other Human Modified Habitats in Indonesia.

Simple SummaryGiven the high degree of deforestation in the tropics due to shifting agriculture, it is a priority for conservation to find sustainable agriculture systems. We assessed bird communities over 1228 plots from 21 sites in the highly populated island of Java, Indonesia. We found that commercial coffee polycultures (i.e., fields comprised of coffee plants, other crops and/or fruit trees, and diverse shade trees) could sustain similar levels of bird abundance, diversity, and richness than coffee systems under natural forests and community managed forests. Commercial coffee polyculture fields host higher bird abundance, diversity, and richness than sun-exposed coffee fields, fields with other crops or fruit trees, and tree farms. We provide evidence that complex commercial agroforestry systems can host similar levels of biodiversity to agroforestry systems under natural forest.Deforestation in the tropics is mainly driven by the need to expand agriculture and forestry land. Tropical cropland has also undergone a process of intensification, particularly evident in regions that are the main exporters of deforestation-driven commodities. Around 25 million people in the world depend on coffee production, which has a profound contribution to global biodiversity loss through agricultural extensification and intensification. Nevertheless, coffee agroforestry systems have been postulated to serve as an alternative refuge for biodiversity across different regions. We aim to compare bird abundance, diversity, and richness in commercial polyculture coffee systems (i.e., the highest degree of habitat complexity that can be achieved in coffee fields after deforestation) with other coffee agroforestry systems and human modified habitats in Java, Indonesia. We collected data in 21 sites (1228 points) on Java from February to August 2021 using the point sampling method. Via generalised additive models, we tested whether the abundance, diversity, and richness of birds were different between different human modified habitats including other potential predictors such as elevation, distance to protected areas, shade tree richness, and plant diversity. Using the non-metric multidimensional scaling, we tested whether there was a difference in terms of the composition of foraging guilds between habitats. Commercial polyculture coffee fields can sustain levels of bird abundance, diversity, and richness comparable to agroforestry systems under natural forest, and higher than sun coffee and shaded monoculture coffee, and of other human modified habitats such as crop/fruit fields and tree farms. Coffee agroforestry systems have a higher proportion of nectarivores, insectivores, and frugivores than other systems that can sustain high diversity and richness of birds such as paddy fields that mainly have granivores and carnivores. Complex polycultures can represent an avenue for the future of sustainable agriculture in conditions where deforestation rates are high and in crops such as coffee, which maintain high yield in the presence of diverse shade.

Classification of Typical Tree Species in Laser Point Cloud Based on Deep Learning

We propose the Point Cloud Tree Species Classification Network (PCTSCN) to overcome challenges in classifying tree species from laser data with deep learning methods. The network is mainly composed of two parts: a sampling component in the early stage and a feature extraction component in the later stage. We used geometric sampling to extract regions with local features from the tree contours since these tend to be species-specific. Then we used an improved Farthest Point Sampling method to extract the features from a global perspective. We input the intensity of the tree point cloud as a dimensional feature and spatial information into the neural network and mapped it to higher dimensions for feature extraction. We used the data obtained by Terrestrial Laser Scanning (TLS) and Unmanned Aerial Vehicle Laser Scanning (UAVLS) to conduct tree species classification experiments of white birch and larch. The experimental results showed that in both the TLS and UAVLS datasets, the input tree point cloud density and the highest feature dimensionality of the mapping had an impact on the classification accuracy of the tree species. When the single tree sample obtained by TLS consisted of 1024 points and the highest dimension of the network mapping was 512, the classification accuracy of the trained model reached 96%. For the individual tree samples obtained by UAVLS, which consisted of 2048 points and had the highest dimension of the network mapping of 1024, the classification accuracy of the trained model reached 92%. TLS data tree species classification accuracy of PCTSCN was improved by 2–9% compared with other models using the same point density, amount of data and highest feature dimension. The classification accuracy of tree species obtained by UAVLS was up to 8% higher. We propose PCTSCN to provide a new strategy for the intelligent classification of forest tree species.