Search

Tree species classification is a ubiquitous task in the forest inventory field. Only directly measured feature vectors have been applied to most existing methods that use LiDAR technology for tree species classification. As a result, it is difficult to obtain a satisfactory tree species classification performance. To solve this challenge, the authors of this paper developed two new kinds of feature vectors, including fractal geometry-based feature vectors and quantitative structural model (QSM)-based feature vectors. In terms of fractal geometry, both two fractal parameters were extracted as feature vectors for reflecting how tree architecture is distributed in three-dimensional space. In terms of QSM, the ratio of length change and the ratio of radius change of different branches were extracted as feature vectors. To reduce the feature vector dimensionality and explore valuable feature vectors, feature vector dimension reduction was conducted using the classification and regression tree (CART). Five hundred and sixty-eight individual trees with five tree species were selected for evaluating the performance of the developed feature vectors. The experimental results indicate that the tree species of Fagus sylvatica achieved the highest overall accuracy, which is 98.06%, while Quercus petraea obtained the lowest overall accuracy, which is 96.65%. Four other classical supervised learning methods were adopted for comparison. The comparison result indicates that the proposed method outperformed the other four supervised learning methods no matter which accuracy indicator was adopted. In comparison with the relevant method, the eight feature vectors developed in this paper also performed much better. This indicates that the fractal geometry-based feature vectors and QSM-based feature vectors developed in this paper can effectively improve the performance of tree species classification.

To improve the accuracy of wood and leaf points classification for individual tree, this paper proposed a separation method based on mode points evolution from terrestrial LiDAR point clouds. In the proposed method, the Mean Shift method was used to first acquire the mode points, which were then adopted as nodes to build a network graph for the individual tree. By path retracing and calculating the visiting frequency of each node, the wood seed nodes were detected. To obtain more wood nodes, the wood seed nodes were evolved based on three constraints, namely the shortest path length of the evolved nodes to the base node should be smaller, the evolved nodes should not belong to the leaf nodes that have been detected by path retracing and the verticality of the evolved nodes should be similar as the wood seed nodes. After wood nodes evolution, the segments corresponding to each wood seed node were merged together to obtain the final wood points. The proposed method has been evaluated using nine tree samples with seven different tree species. Experimental results showed that the proposed method can achieve an average wood and leaf classification accuracy of 0.892. The average F1 score for wood was 0.871, while the average F1 score for leaf was 0.900. Compared to two other famous wood and leaf classification methods, the proposed method can achieve better classification results.

Extracting street trees from mobile Light Detection and Ranging (LiDAR) point clouds is still encountering challenges, such as low extraction accuracy and poor robustness in complex urban environment, and difficulty in the segmentation of overlapping trees. To solve these problems, this paper proposed a street tree extraction and segmentation method based on spatial geometric features of object primitives. In this paper, mobile LiDAR point clouds were first segmented into object primitives based on the proposed graph segmentation method, which can release the computation burden effectively. According to the spatial geometric features of the segmented object primitives, stem points were extracted. In doing so, the robustness and accuracy for stem detecting can be improved. Furthermore, voxel connectivity analysis and individual tree optimization were combined successively. In doing so, the neighboring trees could be separated successfully. Four datasets located in Henan Polytechnic University, China, were used for validating the performance of the proposed method. The four mobile LiDAR point clouds contained 106, 45, 76, and 46 trees, respectively. The experimental results showed that the proposed method can achieve the performance of individual tree separation in all the four testing plots. Compared to the other three methods, the proposed method can make a good balance between the commission and omission errors and achieved the highest average F1 scores.

Unman aerial vehicle (UAV) LiDAR has been widely used in the field of forestry. Individual tree extraction is a key step for forest inventory. Although many individual tree extraction methods have been proposed, the individual tree extraction accuracy is still low due to the complex forest environments. Moreover, many parameters in these methods generally need to be set. Thus, the degree of automation of the methods is generally low. To solve these problems, this paper proposed an automatic mean shift segmentation method, in which the kernel bandwidths can be calculated self-adaptively. Meanwhile, a hierarchy mean shift segmentation technique was proposed to extract individual tree gradually. A plot-level UAV LiDAR tree dataset was adopted for testing the performance of the proposed method. Experimental results showed that the proposed method can achieve better individual tree extraction result without any parameter setting. Compared with the traditional mean shift segmentation method, both the completeness and mean accuracy of the proposed method are higher.

Terrestrial Laser scanner has been widely used in the field of forestry. Wood-leaf separation is the fundamental step to most applications of forestry. This paper presented a robust supervised learning method for wood and leaf classification by developing four new feature vectors. Fractal dimension is first calculated to indicate the difference of regularity or roughness between wood and leaf. Zenith angle and variation are presented to distinguish trunks or branches from leaves. The adaptive axis direction of cylinder is adopted to calculate the local point density precisely. Experimental results show that the supervised learning method using the four feature vectors presented in this paper can achieve a good classification performance. Both accuracy and F1 score are higher than the ones of the method using eigen value based feature vectors.

Denoising is a key pre-processing step for many airborne LiDAR point cloud applications. However, the previous algorithms have a number of problems, which affect the quality of point cloud post-processing, such as DTM generation. In this paper, a novel automated denoising algorithm is proposed based on empirical mode decomposition to remove outliers from airborne LiDAR point cloud. Comparing with traditional point cloud denoising algorithms, the proposed method can detect outliers from a signal processing perspective. Firstly, airborne LiDAR point clouds are decomposed into a series of intrinsic mode functions with the help of morphological operations, which would significantly decrease the computational complexity. By applying OTSU algorithm to these intrinsic mode functions, noise-dominant components can be detected and filtered. Finally, outliers are detected automatically by comparing observed elevations and reconstructed elevations. Three datasets located at three different cities in China were used to verify the validity and robustness of the proposed method. The experimental results demonstrate that the proposed method removes both high and low outliers effectively with various terrain features while preserving useful ground details.

Filtering of ground points is a key step for most applications of airborne LiDAR point clouds. Although many filtering algorithms have been proposed in recent years, most of them suffer from parameter setting or thresholds fine-tuning. This is most often time-consuming and reduces the degree of automation of the applied algorithm. To overcome such problems, this paper proposes a threshold-free filtering algorithm based on expectation–maximization (EM). The filter is developed based on the assumption that point clouds are seen as a mixture of Gaussian models. Thus, the separation of ground points and non-ground points from point clouds is partitioning of the point clouds by a mixed Gaussian model that is used for screening ground points. EM is applied to realize the separation, which calculates the maximum likelihood estimates of the mixture parameters. Using the estimated parameters, the likelihoods of each point belonging to ground or non-ground are computed. Noticeably, point clouds are labeled as the component with a larger likelihood. The proposed method has been tested using the standard filtering datasets provided by the ISPRS. Experimental results showed that the proposed method performed the best in comparison with the classic progressive triangulated irregular network densification (PTD) and segment-based PTD methods in terms of omission error. The average omission error of the proposed method was 52.81% and 16.78% lower than the classic PTD method and the segment-based PTD method, respectively. Moreover, the proposed method was able to reduce its average total error by 31.95% compared to the classic PTD method.

Airborne Light Detection and Ranging (LiDAR) is a popular active remote sensing technology that has been developing very rapidly in recent years. To solve the problems of low filtering accuracy of airborne LiDAR point clouds in complex terrain environments and avoiding too much human intervention, this paper proposes a point cloud filtering method based on active learning. In the proposed method, the initial training samples are acquired and marked automatically by multi-scale morphological operations. In so doing, no training samples are selected and labeled manually, i.e., the training samples are added gradually according to the oracle used in active learning. In this paper, the oracle is set to a sigmoid function of residuals from the points to the fitted surface. Subsequently, the training model is revised progressively using the updated training samples. Finally, the classification results are further optimized by a slope-based method. Three datasets with different filtering challenges provided by the International Society for Photogrammetry and Remote Sensing (ISPRS) were used to test the proposed method. Comparing with the other ten famous filtering methods, the proposed method can achieve the smallest average total error (5.51%). Thus, it can be concluded that the proposed method performs very well toward different terrain environments.

The all-embracing inspection of geometry structures of revetments along urban rivers using the conventional field visual inspection is technically complex and time-consuming. In this study, an approach using dense point clouds derived from low-cost unmanned aerial vehicle (UAV) photogrammetry is proposed to automatically and efficiently recognize the signatures of revetment damage. To quickly and accurately recover the finely detailed surface of a revetment, an object space-based dense matching approach, that is, region growing coupled with semi-global matching, is exploited to generate pixel-by-pixel dense point clouds for characterizing the signatures of revetment damage. Then, damage recognition is conducted using a proposed operator, that is, a self-adaptive and multiscale gradient operator, which is designed to extract the damaged regions with different sizes in the slope intensity image of the revetment. A revetment with slope protection along urban rivers is selected to evaluate the performance of damage recognition. Results indicate that the proposed approach can be considered an effective alternative to field visual inspection for revetment damage recognition along urban rivers because our method not only recovers the finely detailed surface of the revetment but also remarkably improves the accuracy of revetment damage recognition.

Polarimetric synthetic aperture radar is an important technology for Earth observation. Generally, polarimetric calibration (PolCAL) should be performed before any quantitative applications. The reflection symmetric targets can relieve the dependence of PolCAL on calibrators. In this work, we use a helix-ratio feature to extract these targets. Through analysis, we find that the helix-ratio feature has a high tolerance to polarimetric distortions and is suitable for extracting samples in un-calibrated images for PolCAL. An adaptive segmentation algorithm is also developed to extract reflection symmetric targets automatically. Experiments with C- and X-band airborne images show that the extraction of reflection symmetric targets can improve the calibration accuracy of crosstalk. In addition, the proposed method can maintain the polarimetric signature of the targets in the calibration and can improve calibration performance when compared with other reflection symmetric target extraction methods.

In recent years, many airborne point clouds filtering methods have been developed. However, it is still challenging for distinguishing ground and non-ground points in forested areas due to the rugged terrains, dense vegetation canopy and low-level penetration of laser pulses. To derive satisfactory filtering results, this paper proposed a mean shift segmentation morphological filter. In this method, the mean shift segmentation is used for acquiring object primitives to determine filtering window sizes automatically. The point clouds detrending is proposed for improving the adaptability towards sloped terrains. A point cloud shifting in x and y directions technique is developed to acquire more ground seeds for generating a more accurate trending surface. Finally, the filtered ground points by the progressive morphological filter are recovered by adopting the surface-based filtering strategy. The proposed method is tested and validated using 14 samples with different forested environments. Experimental results show that the proposed method can achieve the average total error of 1.11%. The kappa coefficients of all these 14 samples are larger than 90% and the average kappa coefficient is 96.43%. The average root mean square error (RMSE) of the proposed method is 0.63. All these indicators are the best when compared to some other famous filtering methods.

Road information acquisition is an important part of city informatization construction. Airborne LiDAR provides a new means of acquiring road information. However, the existing road extraction methods using LiDAR point clouds always decide the road intensity threshold based on experience, which cannot obtain the optimal threshold to extract a road point cloud. Moreover, these existing methods are deficient in removing the interference of narrow roads and several attached areas (e.g., parking lot and bare ground) to main roads extraction, thereby imparting low completeness and correctness to the city road network extraction result. Aiming at resolving the key technical issues of road extraction from airborne LiDAR point clouds, this paper proposes a novel method to extract road centerlines from airborne LiDAR point clouds. The proposed approach is mainly composed of three key algorithms, namely, Skewness balancing, Rotating neighborhood, and Hierarchical fusion and optimization (SRH). The skewness balancing algorithm used for the filtering was adopted as a new method for obtaining an optimal intensity threshold such that the “pure” road point cloud can be obtained. The rotating neighborhood algorithm on the other hand was developed to remove narrow roads (corridors leading to parking lots or sidewalks), which are not the main roads to be extracted. The proposed hierarchical fusion and optimization algorithm caused the road centerlines to be unaffected by certain attached areas and ensured the road integrity as much as possible. The proposed method was tested using the Vaihingen dataset. The results demonstrated that the proposed method can effectively extract road centerlines in a complex urban environment with 91.4% correctness and 80.4% completeness.

Filtering is a key step for most applications of airborne LiDAR point clouds. Although lots of filtering algorithms have been put forward in recent years, most of them suffer from parameters setting or thresholds adjusting, which will be time-consuming and reduce the degree of automation of the algorithm. To overcome this problem, this paper proposed a threshold-free filtering algorithm based on expectation-maximization. The proposed algorithm is developed based on an assumption that point clouds are seen as a mixture of Gaussian models. The separation of ground points and non-ground points from point clouds can be replaced as a separation of a mixed Gaussian model. Expectation-maximization (EM) is applied for realizing the separation. EM is used to calculate maximum likelihood estimates of the mixture parameters. Using the estimated parameters, the likelihoods of each point belonging to ground or object can be computed. After several iterations, point clouds can be labelled as the component with a larger likelihood. Furthermore, intensity information was also utilized to optimize the filtering results acquired using the EM method. The proposed algorithm was tested using two different datasets used in practice. Experimental results showed that the proposed method can filter non-ground points effectively. To quantitatively evaluate the proposed method, this paper adopted the dataset provided by the ISPRS for the test. The proposed algorithm can obtain a 4.48 % total error which is much lower than most of the eight classical filtering algorithms reported by the ISPRS.

Geocentric translation model (GTM) in recent times has not gained much popularity in coordinate transformation research due to its attainable accuracy. Accurate transformation of coordinate is a major goal and essential procedure for the solution of a number of important geodetic problems. Therefore, motivated by the successful application of Artificial Intelligence techniques in geodesy, this study developed, tested and compared a novel technique capable of improving the accuracy of GTM. First, GTM based on official parameters (OP) and new parameters determined using the arithmetic mean (AM) were applied to transform coordinate from global WGS84 datum to local Accra datum. On the basis of the results, the new parameters (AM) attained a maximum horizontal position error of 1.99 m compared to the 2.75 m attained by OP. In line with this, artificial neural network technology of backpropagation neural network (BPNN), radial basis function neural network (RBFNN) and generalized regression neural network (GRNN) were then used to compensate for the GTM generated errors based on AM parameters to obtain a new coordinate transformation model. The new implemented models offered significant improvement in the horizontal position error from 1.99 m to 0.93 m. Resumo: O modelo de translação geocêntrica (GTM) atualmente não tem sido muito utilizado nas pesquisas de transformação de coordenadas devido a sua exatidão. Uma transformação acurada de coordenadas é um procedimento essencial na maioria dos problemas geodésicos. Entretanto, motivado pelo sucesso na aplicação de técnicas de Inteligência Artificial na Geodésia, este estudo testou e comparou uma nova técnica capaz de aumentar a exatidão do modelo GTM. Inicialmente, o GTM é definido com parâmetros oficiais (OP) e novos parâmetros determinados usando a média aritmética (AM) foram aplicados na transformação de coordenadas do datum global WGS84 para o datum local Accra. Com base nesses resultados, os novos parâmetros da média aritmética (AM) alcançaram um erro posicional horizontal máximo de 1.99m, em comparação com 2.75m alcançado pelos parâmetros oficiais (OP). De maneira similar, a tecnologia de redes neurais artificiais dos modelos BPNN (BackPropagation Neural Network), RBFNN (Radial Basis Function Neural Network) e GRNN (Generalized Regression Neural Network) foram então utilizadas para compensar os erros do modelo GTM com os parâmetros baseados na média aritmética (AM) para se obter um novo modelo de transformação de coordenadas. Os novos modelos implementados apresentaram uma melhora significativa no erro posicional horizontal de 1.99m para 0.93m

Filtering is a key step for most airborne lidar post-applications in urban areas. To solve the problems of complex parameter settings and low filtering accuracy in complicated urban environments, an automatic morphological filter was proposed. In this paper, the optimal maximum filtering window can be determined automatically by applying a series of morphological top-hat operations. Meanwhile, the thresholds for filtering were calculated adaptively according to the gradient changes. Seven publicly available data sets provided by the International Society for Photogrammetry and Remote Sensing were used to evaluate the performance. Experimental results show that the proposed method achieved an average total error of 4.07% and an average kappa coefficient of 90.90%, which are the best performances when compared to some other filtering methods.

This paper proposes a method by which scattered LIDAR point clouds are converted into a two-dimensional image and then building footprints are extracted through the image processing. Firstly point cloud grid is handled to generate georeferenced feature image, and then image threshold segmentation, morphological close operation, connectivity analysis and contour tracking method is used to obtain the final building footprints. Finally, based on the mapping relationship between georeferenced feature image and scattered point clouds, building outline points are obtained. Experimental result shows that this method could extract building footprints very well in plain area, but due to the adoption of single image segmentation method in the georeferenced feature image, it is not suitable for the building footprints extraction in mountainous area.

Obtaining high-precision filtering results from airborne lidar point clouds in complex environments has always been a hot topic. Mathematical morphology was widely used for filtering, owing to its simplicity and high efficiency. However, the morphology-based algorithms are deficient in preserving terrain details. In order to obtain a better filtering effect, this paper proposed an improved progressive morphological filter based on hierarchical radial basis function interpolation (PMHR) to refine the classical progressive morphological filter. PMHR involved two main improvements, namely, automatic setting of self-adaptive thresholds and terrain details preservation, respectively. The performance of PMHR was evaluated using datasets provided by the International Society for Photogrammetry and Remote Sensing. Experimental results show that PMHR achieved good performance under variant terrain features with an average total error of 4.27% and average Kappa coefficient of 84.57%.

Filtering is one of the core post-processing steps for airborne LiDAR point cloud. In recent years, the morphology-based filtering algorithms have proven to be a powerful and efficient tool for filtering airborne LiDAR point cloud. However, most traditional morphology-based algorithms have difficulties in preserving abrupt terrain features, especially when using larger filtering windows. In order to suppress the omission error caused by protruding terrain features, this paper proposes an improved morphological algorithm based on multi-level kriging interpolation. This algorithm is essentially a combination of progressive morphological filtering algorithm and multi-level interpolation filtering algorithm. The morphological opening operation is performed with filtering window gradually downsizing, while kriging interpolation is conducted at different levels according to the different filtering windows. This process is iterative in a top to down fashion until the filtering window is no longer greater than the preset minimum filtering window. Fifteen samples provided by the ISPRS commission were chosen to test the performance of the proposed algorithm. Experimental results show that the proposed method can achieve promising results not only in flat urban areas but also in rural areas. Comparing with other eight classical filtering methods, the proposed method obtained the lowest omission error, and preserved protruding terrain features better.