Foliage Clumping Index (CI) is an important structural parameter within vegetation canopy, and current satellite-borne CI products mainly retrieved by using the linear relationship between the CI and the normalized difference between hotspot and dark spot (NDHD), while there is no directly model to calculate the CI of mixed forest. The objective of this paper is to propose a new method to calculate the mixed forest CI (MFCI) and access the ability to response the spatial heterogeneity of mixed forest pixels. The results show that: (1) The accuracy of MFCI is significantly higher than that of existing MODIS CI products, the average error can be reduced by 6.3% (2) the total sensitivity of MFCI to spatial heterogeneity is high(>0.6), and with the highest sensitivity to bare soil.

The surface anisotropy, usually described as bidirectional reflectance distribution function (BRDF), plays a key role in the quantitative remote sensing. Numerous BRDF studies are focus on sensors onboard polar-orbiting satellites such as POLDER and MODIS based on multi-angle sensors or accumulative observations through multiple days. Notably, sensors onboard geostationary satellites can also obtain multi-angle reflectances benefited from their high revisit frequencies, while by which only a few BRDF studies have been completed. In this study, we aim to evaluate the BRDF information collected from time-series directional observations of the Advanced Himawari Imager (AHI) onboard geostationary satellite Himawari-8. The multi-angle reflectances of 6 × 7 km POLDER and 0.05° AHI at mixed forest area during a whole month were collected. Generally, AHI has a good weight of Determination (WoD) of 0.03, as well as small fit-RMSEs of 0.0055 and 0.0235 in the red and NIR bands based on the kernel-driven Ross-Li BRDF model, which shows promising potential to provide BRDF information with a good quality.

Bidirectional Reflectance Distribution Function (BRDF) is an important component in quantitative remote sensing. In this study, we explored the relationship of sampling distribution and reconstructed BRDF curve for snow surface. In order to get enough observations, we utilize the field-measured data to analyze the different BRDF under different sampling pattern. This work can be very meaningful because it is hard to acquire sufficient measurements in all directions especially for snow in the real life.

Clumping index (CI) characterizes the clumping degree of vegetation canopy foliage relative to the random distribution, which is an important vegetation structure parameter. Digital Hemispherical Photography (DHP) is widely used in ground CI measurement and one of the key steps using this method is to determine the sampling resolution. This study presents a set of sampling way including 30 sampling methods with 17 levels of sampling resolutions. They were applied to four vegetation types and different growth stages of crops to explore the variation of CI with the decrease of sampling resolution of 17 levels. The results show that with the decrease of sampling resolution at 17 levels, the average increase of CI in the four vegetation types was 26%, 29%, 14% and 35%, and in different growth stages of soybean crop, the increase of CI was different which could be up to 60% at the maximum increase.

Clumping Index (CI) is an important vegetation structure parameter, which describes the grouping of leaves relative to the random distribution. Multi-angle data of the POLarization and Directionality of Earth Reflectance (POLDER) sensor (about 6×7 km) and the MODerate resolution Imaging Spectradiometer (MODIS) (500 m) are two main sources for global CI products. To better understand the variability inherent in CIs of such different spatial resolutions and optimize the used of CI products, extensive POLDER CIs and corresponding MODIS CIs were retrieved and compared in this study. Field measurements were conduct in one selected POLDER pixel in Hebei, China. Our results showed that POLDER and MODIS CIs presented relative good consistency (R2=0.65, RMSE=0.07, bias=0.003), and the correlation coefficient can reach 0.98 at the class level. Both POLDER CI (0.67) and MODIS CI (0.62±0.06) showed good consistency with field CIs (0.64±0.10) and POLDER CI was more likely to overestimate than MODIS CI.

Surface albedo is closely related to the Earth’s energy budget and is usually estimated by integrating remotely sensed bidirectional reflectance distribution function (BRDF) data based on the widely used Ross–Li kernel-driven models. However, for large solar zenith angles (i.e., SZAs > 70°), albedo estimation using the operational algorithm of the Moderate Resolution Imaging Spectroradiometer (MODIS), i.e., RossThick-LiSparseReciprocal (RTLSR), is not recommended because it is reported to somewhat underestimate the black-sky albedo (BSA) at large SZAs based on ground albedo measurements. Recently, various combinations of the Ross–Li BRDF models with improved capabilities have been developed, and the assessments of these models based on worldwide satellite BRDF data with good spatial sampling, particularly at the large view and solar angles, will be important to improve an understanding of their performance in estimating intrinsic albedos. Following previous studies, the objective of this study is to further assess a series of hotspot-corrected Ross–Li models by demonstrating their ability to fit the POLarization and Directionality of the Earth’s Reflectances (POLDER) data sets and estimate albedo, especially at large SZAs, based on selected concurrent POLDER and MODIS data. The hotspot-corrected RTLSR model obtained by combining the RossThickChen and LiSparseReciprocalChen kernels (RTLSR_C) shows the best fitting ability, with a high cumulative frequency of small root-mean-square errors (RMSEs), thus confirming previous conclusions. Model differences mainly appear in albedo estimates, especially BSA estimates at large SZAs. The BSAs estimated by other models are significantly different from the RTLSR_C estimates in the near-infrared (NIR) and red bands as the SZA increases to approximately 60° and 70°, respectively. In this case, RossThinChen-LiSparseReciprocalChen (RTNLSR_C) yields higher BSA estimates than those of RTLSR_C. Comparisons of the MODIS and POLDER albedos estimated with Ross–Li models show that models with the RossThinChen kernel yield higher BSA estimates than those of the RTLSR_C model as the SZA increases. The results indicate that the retrieved albedo is likely to be more accurate with appropriately selected kernels for BRDF models at large SZAs, providing guidance for selecting suitable combinations of multiple kernels.

A series reflectance models for snow have been developed in recent years, but there is no trusted pure snow database to test the models yet. The Polarization and Directionality of Earth Reflectances (POLDER) BRDF data have been widely used in quantitative remote sensing community, but the recent research has revealed that an obvious wrong classification has occurred in snow POLDER BRDF database. In this study, we propose a snow index (SI) to characterize the scattering feature of snow based on RTLSRS model and develop an effective method to identify the pure snow data in POLDER database. Finally, we build a snow database with higher purity, providing a data support for model validation in the future.

To parse large-scale urban scenes using the supervised methods, a large amount of training data that can account for the vast visual and structural variance of urban environment is necessary. Unfortunately, such training data are mostly obtained by tedious and time-consuming manual work. To overcome the drawback, we propose a semisupervised learning framework that combines the margin, cograph, and label constraints into an objective function for point cloud parsing. Mathematically, the margin constraint is presented to learn a novel distance criterion that can effectively recognize points of different classes. The graph regularization is then employed to characterize the intrinsic geometry structure of the data manifold and explore relationships among points. The label consistency regularization is introduced to ensure the category consistency of the clustered points and single point. To classify the out-of-sample data, the framework successfully transforms the semisupervised classification results into the linear classifier by adopting a linear regression. An iterative algorithm is utilized to efficiently and effectively optimize the objective function with characteristics of multiple variables and highly nonlinear. The point clouds of four urban scenes are used to validate our method. The experimental results show that our method outperforms the state-of-the-art algorithms.

The large number of object categories and many overlapping or closely neighboring objects in large-scale urban scenes pose great challenges in point cloud classification. Most works in deep learning have achieved a great success on regular input representations, but they are hard to be directly applied to classify point clouds due to the irregularity and inhomogeneity of the data. In this paper, a deep neural network with spatial pooling (DNNSP) is proposed to classify large-scale point clouds without rasterization. The DNNSP first obtains the point-based feature descriptors of all points in each point cluster. The distance minimum spanning tree-based pooling is then applied in the point feature representation to describe the spatial information among the points in the point clusters. The max pooling is next employed to aggregate the point-based features into the cluster-based features. To assure the DNNSP is invariant to the point permutation and sizes of the point clusters, the point-based feature representation is determined by the multilayer perception (MLP) and the weight sharing for each point is retained, which means that the weight of each point in the same layer is the same. In this way, the DNNSP can learn the features of points scaled from the entire regions to the centers of the point clusters, which makes the point cluster-based feature representations robust and discriminative. Finally, the cluster-based features are input to another MLP for point cloud classification. We have evaluated qualitatively and quantitatively the proposed method using several airborne laser scanning and terrestrial laser scanning point cloud data sets. The experimental results have demonstrated the effectiveness of our method in improving classification accuracy.