Improving forest type classification using the vegetation local difference index

A spatial feature extraction method was applied to increase the accuracy of land-cover classification of forest type information extraction. Traditional spatial feature extraction applications use high-resolution images. However, improving the classification accuracy is difficult when using medium-resolution images, such as a 30 m resolution Enhanced Thematic Mapper Plus (ETM+) image. In this study, we demonstrated a novel method that used the vegetation local difference index (VLDI) derived from the normalized difference vegetation index (NDVI), which were calculated based on the topographically corrected ETM+ image, to delineate spatial features. A simple maximum likelihood classifier and two different ways to use spatial information were introduced in this study as the frameworks to incorporate both spectral and spatial information for analysis. The results of the experiments, where Landsat ETM+ and digital elevation model (DEM) images, together with ground truth data acquired in the study area were used, show that combining the spatial information extracted from medium-resolution images and spectral information improved both classification accuracy and visual qualities. Moreover, the use of spatial information extracted through the proposed method greatly improved the classification performance of particular forest types, such as sparse woodlands.     

Feasibility of multi-temporal interferometric SAR data for stand-level estimation of boreal forest stem volume

The feasibility of interferometric SAR (INSAR) coherence observations for stem volume (biomass) retrieval is investigated by applying coherence data determined from 14 ERS-1 and ERS-2 C-band SAR image pairs. The image set covers a single forested test area in Finland, and both summer (snow-free) and winter conditions are represented. The data set enabled (a) the study of stem volume retrieval performance under varying conditions, (b) the analysis of the seasonal behavior of interferometric coherence, and (c) the determination of the accuracy characteristics of empirical (nonlinear) coherence modeling. Additionally, a new technique to estimate forest stem volume from INSAR data was developed based on constrained iterative inversion of the applied empirical model. The results indicate that the usability of winter images with snow-covered terrain is superior to that of images obtained under summer conditions. The applied empirical model appears to be adequate for describing the stand-wise coherence of boreal forest. Hence, a practical stem volume estimation method can be established based on it. The highest correlation coefficient between the estimated stem volume and the ground truth stem volume showed values as high as r=0.9 and a relative RMSE level of 48% was obtained, respectively.     

Combining and updating of local estimates and regional maps along sets of one-dimensional tracks

In this paper we consider the problem of combining and updating estimates that may have been generated in a distributed fashion or may represent estimates, generated at different times, of the same process sample path. The first of these cases has applications in decentralized estimation, while the second has applications in updating maps of spatially-distributed random quantities given measurements along several tracks. The method of solution for the second problem uses the result of the first, and the similarity in the formulation and solution of these problems emphasizes the conceptual similarity between many problems in decentralized control and in the analysis of random fields.     

Wetland mapping by fusion of airborne laser scanning and multi-temporal multispectral satellite imagery

Wetlands play a major role in Europe’s biodiversity. Despite their importance, wetlands are suffering from constant degradation and loss, therefore, they require constant monitoring. This article presents an automatic method for the mapping and monitoring of wetlands based on the fused processing of laser scans and multispectral satellite imagery, with validations and evaluations performed over an area of Lake Balaton in Hungary. Markov Random Field models have already been shown to successfully integrate various image properties in several remote sensing applications. In this article, we propose the multi-layer fusion Markov Random Field model for classifying wetland areas, built into an automatic classification process that combines multi-temporal multispectral images with a wetland classification reference map derived from airborne laser scanning (ALS) data acquired in an earlier year. Using an ALS-based wetland classification map that relied on a limited amount of ground truthing proved to improve the discrimination of land-cover classes with similar spectral characteristics. Based on the produced classifications, we also present an unsupervised method to track temporal changes of wetland areas by comparing the class labellings of different time layers. During the evaluations, the classification model is validated against manually interpreted independent aerial orthoimages. The results show that the proposed fusion model performs better than solely image-based processing, producing a non-supervised/semi-supervised wetland classification accuracy of 81–93% observed over different years.     

Improving object orientation estimates by considering multiple viewpoints

This article describes a probabilistic approach for improving the accuracy of general object pose estimation algorithms. We propose a histogram filter variant that uses the exploration capabilities of robots, and supports active perception through a next-best-view proposal algorithm. For the histogram-based fusion method we focus on the orientation of the 6 degrees of freedom (DoF) pose, since the position can be processed with common filtering techniques. The detected orientations of the object, estimated with a pose estimator, are used to update the hypothesis of its actual orientation. We discuss the design of experiments to estimate the error model of a detection method, and describe a suitable representation of the orientation histograms. This allows us to consider priors about likely object poses or symmetries, and use information gain measures for view selection. The method is validated and compared to alternatives, based on the outputs of different 6 DoF pose estimators, using real-world depth images acquired using different sensors, and on a large synthetic dataset.     

Data fusion and multicue data matching by diffusion maps

Data fusion and multicue data matching are fundamental tasks of high-dimensional data analysis. In this paper, we apply the recently introduced diffusion framework to address these tasks. Our contribution is three-fold: first, we present the Laplace-Beltrami approach for computing density invariant embeddings which are essential for integrating different sources of data. Second, we describe a refinement of the Nystrom extension algorithm called "geometric harmonics." We also explain how to use this tool for data assimilation. Finally, we introduce a multicue data matching scheme based on nonlinear spectral graphs alignment. The effectiveness of the presented schemes is validated by applying it to the problems of lipreading and image sequence alignment     

Improving acoustic vehicle classification by information fusion

We present an information fusion approach for ground vehicle classification based on the emitted acoustic signal. Many acoustic factors can contribute to the classification accuracy of working ground vehicles. Classification relying on a single feature set may lose some useful information if its underlying sound production model is not comprehensive. To improve classification accuracy, we consider an information fusion diagram, in which various aspects of an acoustic signature are taken into account and emphasized separately by two different feature extraction methods. The first set of features aims to represent internal sound production, and a number of harmonic components are extracted to characterize the factors related to the vehicle’s resonance. The second set of features is extracted based on a computationally effective discriminatory analysis, and a group of key frequency components are selected by mutual information, accounting for the sound production from the vehicle’s exterior parts. In correspondence with this structure, we further put forward a modified Bayesian fusion algorithm, which takes advantage of matching each specific feature set with its favored classifier. To assess the proposed approach, experiments are carried out based on a data set containing acoustic signals from different types of vehicles. Results indicate that the fusion approach can effectively increase classification accuracy compared to that achieved using each individual features set alone. The Bayesian-based decision level fusion is found to be improved than a feature level fusion approach.     

Spatially explicit characterization of boreal forest gap dynamics using multi-temporal lidar data

Understanding a disturbance regime such as gap dynamics requires that we study its spatial and temporal characteristics. However, it is still difficult to observe and measure canopy gaps extensively in both space and time using field measurements or bi-dimensional remote sensing images, particularly in open and patchy boreal forests. In this study, we investigated the feasibility of using small footprint lidar to map boreal canopy gaps of sizes ranging from a few square meters to several hectares. Two co-registered canopy height models (CHMs) of optimal resolution were created from lidar datasets acquired respectively in 1998 and 2003. Canopy gaps were automatically delineated using an object-based technique with an accuracy of 96%. Further, combinatorics was applied on the two CHMs and the delineated gaps to provide information on the area of old and new gaps, gap expansions, new random gap openings, gap closure due to lateral growth of adjacent vegetation or due to vertical growth of regeneration. The results obtained establish lidar as an excellent tool for rapidly acquiring detailed and spatially extensive short-term dynamics of canopy gaps.     

Indecomposable local maps of tessellation automata

Indecomposable local maps of one-dimensional tessellation automata are studied. The main results of this paper are the following. (1) For any alphabet containing two or more symbols and for anyn 1, there exist indecomposable scope-n local maps over . (2) If is a finite field of prime order, then a linear scope-n local map over is indecomposable if and only if its associated polynomial is an irreducible polynomial of degreen – 1 over , except for a trivial case. (3) Result (2) is no longer true if is a finite field whose order is not prime.     

Speckle reduction for the forest mapping analysis of multi-temporal Radarsat-1 images

As the number of satellite-borne synthetic aperture radar (SAR) systems increases, both the availability and the length of multi-temporal (MT) sequences of SAR images have also increased. Previous research on MT SAR sequences suggests that they increase the classification accuracy for all applications over single date images. Yet the presence of speckle noise remains a problem and all images in the sequence must be speckle filtered before acceptable classification accuracy can be attained. Several speckle filters designed specifically for MT sequences have been reported in the literature. Filtering in the spatial domain, as is usually done, reduces the effective spatial resolution of the filtered image. MT speckle filters operate in both the spatial and temporal dimensions, thus the reduction in resolution is not likely to be as severe (although a comparison between MT and spatial filters has not been reported). While this advantage may be useful when extracting spatial features from the image sequence, it is not quite as apparent for classification applications. This research explores the relative performance of spatial and MT speckle filtering for a particular classification application: mapping boreal forest types. We report filter performance using the radiometric resolution as measured by the equivalent number of looks (NL), and classification performance as measured by the classification accuracy. We chose representative spatial and MT filters and found that spatial speckle filters offer the advantage of higher radiometric resolution and higher classification accuracy with lower algorithm complexity. Thus, we confirm that MT filtering offers no advantage for classification applications; spatial speckle filters yield higher overall performance.     

Improving protein secondary structure predictions by prediction fusion

Protein secondary structure prediction is still a challenging problem at today. Even if a number of prediction methods have been presented in the literature, the various prediction tools that are available on-line produce results whose quality is not always fully satisfactory. Therefore, a user has to know which predictor to use for a given protein to be analyzed. In this paper, we propose a server implementing a method to improve the accuracy in protein secondary structure prediction. The method is based on integrating the prediction results computed by some available on-line prediction tools to obtain a combined prediction of higher quality. Given an input protein p whose secondary structure has to be predicted, and a group of proteins F, whose secondary structures are known, the server currently works according to a two phase approach: (i) it selects a set of predictors good at predicting the secondary structure of proteins in F (and, therefore, supposedly, that of p as well), and (ii) it integrates the prediction results delivered for p by the selected team of prediction tools. Therefore, by exploiting our system, the user is relieved of the burden of selecting the most appropriate predictor for the given input protein being, at the same time, assumed that a prediction result at least as good as the best available one will be delivered. The correctness of the resulting prediction is measured referring to EVA accuracy parameters used in several editions of CASP.     

Extension of dynamic link matching by introducing local linear maps.

It is well known that dynamic link matching (DLM) is a flexible pattern matching model tolerant of deformation or nonlinear transformation. However, previous models cannot treat severely deformed data pattern in which local features do not have their counterparts in a template pattern. We extend DLM by introducing local linear maps (LLMs). Our model has a reference vector and an LLM for each lattice point of a data pattern. The reference vector maps the lattice point into a template pattern and the LLM carries the information regarding how the local neighborhood is mapped. Our model transforms local features by LLMs in a data pattern and then matches them with their counterparts in a template pattern. Therefore, our model is adaptable to larger transformations. For simplicity, we restricted LLMs to rotations. Neighboring LLMs are diffusionally coupled with each other. The model is numerically demonstrated to be very flexible in dealing with deformation and rotation compared to previous models. The framework of our model can be easily extended to models with more general LLMs (expansion, contraction, and so on).     

融合局部奇异值特性的人脸识别方法*

针对仅在整幅人脸图像上进行奇异值分解无法得到人脸识别所需的足够信息的不足,提出了一种利用人脸图像的局部奇异值和灰色关联分析进行人脸识别的方法。该方法的关键是不在整幅人脸图像上进行,而是在人脸的不同区域进行奇异值分解以提取更丰富的信息和克服小样本效应。在识别阶段,对待识别人脸,计算其与各人脸样本的隶属度,最后作出判别。该方法与传统方法在ORL与AR人脸库上进行的对比实验结果表明,该方法不仅提高了识别率,且对人脸姿态变化与部分遮挡也具有一定的鲁棒性。     

Conservative interpolation between volume meshes by local Galerkin projection

The problem of interpolating between discrete fields arises frequently in computational physics. The obvious approach, consistent interpolation, has several drawbacks such as suboptimality, non-conservation, and unsuitability for use with discontinuous discretisations. An alternative, Galerkin projection, remedies these deficiencies; however, its implementation has proven very challenging. This paper presents an algorithm for the local implementation of Galerkin projection of discrete fields between meshes. This algorithm extends naturally to three dimensions and is very efficient.     

Combination of multiple classifiers using local accuracy estimates

This paper presents a method for combining classifiers that uses estimates of each individual classifier's local accuracy in small regions of feature space surrounding an unknown test sample. An empirical evaluation using five real data sets confirms the validity of our approach compared to some other combination of multiple classifiers algorithms. We also suggest a methodology for determining the best mix of individual classifiers     

附着贴图的3维角色模型融合方法

现有的3维模型融合均是立足于模型的几何网格信息,需要大量的顶点来提升融合效果。然而在目前的游戏与动画中,角色3维模型一般使用低精度模型,而以贴图来表现模型细节。因此针对低模角色网格顶点稀疏以及贴图特征,提出简单调和映射和均平面的思想简化网格连接,提高网格融合效率;其次采用基于仿射变换的贴图合成方法完成网格模型融合后的新贴图生成问题,进而完成角色模型的无缝融合。针对本文的方法以及3维模型设计制作的业务流程,建立人机交互的3维模型融合应用系统,实现高效率的3维模型融合与贴图合成。实验结果表明,该方法能够快速有效地完成动画和游戏中角色模型的融合,形成逼真度较高的新角色,为动画和游戏应用增添乐趣。     

Master-CAM: Multi-scale fusion guided by Master map for high-quality class activation maps

Class Activation Map (CAM) is one of the most popular approaches to visually explain the convolutional neural networks (CNNs). To obtain fine-grained saliency maps, some works fuse saliency signals of the same image at larger scales. However, existing methods based on multi-scale fusion cannot effectively remove the noise from larger-scale images. In this paper, we propose Master-CAM, which uses Master map to guide multi-scale fusion process to obtain a high-quality class activation map. Master-CAM utilizes the general localization ability of the Master map to reduce the noise of the maps. We call the one with the general localization ability among the saliency maps from the same image as Master map, which is the saliency map of the original-scale input in the multi-scale scenario. In addition, we also present a simple yet effective fusion strategy, Master-Fusion, which is derived from the fusion operation in Master-CAM. Master-Fusion strategy can be easily attached to some saliency methods to improve the performance of these methods. We show through qualitative and quantitative experiments that the proposed Master-CAM outperforms the state-of-the-art methods in different CNN frameworks and datasets.     

Assessment of forest damage caused by an ice storm using multi-temporal remote-sensing images: a case study from Guangdong Province

In early 2008, forest ecosystems in southern China suffered damage due to a severe ice storm disaster. The area and degree of forest damage caused by the ice storm was assessed using Satellite Pour l’Observation de la Terre (SPOT)-Vegetation images for Guangdong Province acquired between 1999 and 2008. By using the maximum value composition method and image thresholding techniques, the forest vegetation loss, expressed as the change in net primary productivity (NPP) and two indicators (I₁, I₂), was estimated. The damage threshold was determined by comparing the standard deviation of pixels of the undamaged areas in 2008 and other years without any disaster, which was 10%. The area of damaged forest vegetation was 47,670 km², with the northern Guangdong Province most seriously affected. The total loss of NPP for forest vegetation was 50,578,055 t (DW) year^?1, with 52 counties (43.7%) suffering forest vegetation damage. Evergreen coniferous forest was most widely affected, but evergreen broad-leaved forest was the most severely damaged vegetation type. Terrain topography influenced the damage to forest vegetation, which was found to increase with increasing elevation and slope gradient. The range and degree of damaged forest determined by remote-sensing data is consistent with the extent of the ice storm, indicating that this study provides a new approach for rapid assessment of forest disasters at a regional scale.     

Radiometric correction of multi-temporal Landsat data for characterization of early successional forest patterns in western Oregon

Detecting and characterizing continuous changes in early forest succession using multi-temporal satellite imagery requires atmospheric correction procedures that are both operationally reliable, and that result in comparable units (e.g., surface reflectance). This paper presents a comparison of five atmospheric correction methods (2 relative, 3 absolute) used to correct a nearly continuous 20-year Landsat TM/ETM+ image data set (19-images) covering western Oregon (path/row 46/29). In theory, full absolute correction of individual images in a time-series should effectively minimize atmospheric effects resulting in a series of images that appears more similar in spectral response than the same set of uncorrected images. Contradicting this theory, evidence is presented that demonstrates how absolute correction methods such as Second Simulation of the Satellite Signal in the Solar Spectrum (6 s), Modified Dense Dark Vegetation (MDDV), and Dark Object Subtraction (DOS) actually make images in a time-series somewhat less spectrally similar to one another. Since the development of meaningful spectral reflectance trajectories is more dependant on consistent measurement of surface reflectance rather than on accurate estimation of true surface reflectance, correction using image pairs is also tested. The relative methods tested are variants of an approach referred to as “absolute-normalization”, which matches images in a time-series to an atmospherically corrected reference image using pseudo-invariant features and reduced major axis (RMA) regression. An advantage of “absolute-normalization” is that all images in the time-series are converted to units of surface reflectance while simultaneously being corrected for atmospheric effects. Of the two relative correction methods used for “absolute-normalization”, the first employed an automated ordination algorithm called multivariate alteration detection (MAD) to statistically locate pseudo-invariant pixels between each subject and reference image, while the second used analyst selected pseudo-invariant features (PIF) common to the entire image set. Overall, relative correction employed in the “absolute-normalization” context produced the most consistent temporal reflectance response, with the automated MAD algorithm performing equally as well as the handpicked PIFs. Although both relative methods performed nearly equally in terms of observed errors, several reasons emerged for preferring the MAD algorithm. The paper concludes by demonstrating how “absolute-normalization” improves (i.e., reduces scatter in) spectral reflectance trajectory models used for characterizing patterns of early forest succession.