Chinsu Lin

A Multi-level Morphological Active Contour Algorithm for Delineating Tree Crowns in Mountainous Forest

This paper introduces a multi-level morphological active contour (MMAC) algorithm to identify and delineate tree crowns in mountainous forest based on rasterized airborne lidar data. MMAC is a generalized tree crown mapping algorithm which can accommodate multiple heads in a crown as well as overlapping crowns. The MMAC algorithm comprises three steps: bottom up erosion (BUE) which identifies stand candidates, top down dilation (TDD) which estimates the crown periphery, and an active contour model (ACM) which delineates crown contours. Three sample plots were selected in Alishan National Scenic Area, Taiwan, (predominantly alder, sugi, and red cypress) for evaluation of the algorithm. When compared with ground survey data, the algorithm achieved an average detection accuracy of 24 percent omission error and 13 percent commission error in identifying individual trees in mountainous forest stands. Detection accuracy is potentially related to stand density.

Growth-Competition-Based Stem Diameter and Volume Modeling for Tree-Level Forest Inventory Using Airborne LiDAR Data

An individual tree within a forest stand will have its height and diameter growth restricted by the influence of neighboring trees. This is because trees in close proximity compete for resources and space to enable growth. In this paper, the position of trees, tree height (LH), tree crown radius (LCR), and growth competition index (LCI) were extracted from a light-detection-and-ranging (LiDAR)-based rasterized canopy height model using the multilevel morphological active-contour algorithm. The diameter and volume of individual trees are tested and validated to be an exponential function of those LiDAR-derived tree parameters. The best LiDAR-based diameter estimation model and volume estimation model were tested as significant with an R2 value of 0.84 and 0.9 and evaluated with an estimation bias of 8.7 cm and 0.91 m3, respectively. Results also showed that LH and LCR are positively related to the LiDAR-derived diameter at breast height (DBH) and the LiDAR-derived volume of individual trees in a forest stand, whereas LCI is negatively related. The proposed algorithm of individual tree volume estimation was further applied to predict the volume of three sample plots in mountainous forest stands. It was found that the LVM could be used to predict an acceptable volume estimate of old-aged forest stands. The estimation bias, i.e., percentage RMSE (RMSE%), is averaged at around 4% using the LiDAR metrics lnLH, LCI, and LCR, whereas the RMSE% increases to 50% if only lnLH is applied. Results suggest that LCI is an important regulation factor in the estimation of forest volume stocks using LiDAR remote sensing.

Progressive Band Dimensionality Expansion and Reduction Via Band Prioritization for Hyperspectral Imagery

Processing enormous hyperspectral data results in high computational complexity. Band selection (BS) is one common practice to accomplish this goal. However, determining the number of bands to be selected and finding appropriate bands for BS are very challenging since it requires an exhaustive search. Instead of directly dealing with these two issues, this paper introduces a new approach, called progressive band dimensionality process (PBDP) which performs progressive band dimensionality expansion and reduction via band prioritization (BP) which prioritizes the hyperspectral bands according to their priority scores calculated by a specific BP criterion. Two dual processes, referred to as forward PBDP (FPBDP) which performs band expansion in a forward manner and backward PBDP (BPBDP) which performs band dimensionality reduction in a backward manner. By virtue of its progressive nature the PBDP can be implemented by high computing performance while avoiding excessive computing time required by finding an optimal subset from all possible band subset combinations out of full bands. As a consequence, PBDP provides band selection with an advantage of not being trapped in high computational complexity resulting from solving combinatorial mathematics problems. A key to success in PBDP is how to design BP criteria to meet various applications. To address this need, four categories of BP are derived from different designing rationales, second order statistics, higher-order statistics, classification, and band correlation/dependence.

Spectral Feature Probabilistic Coding for Hyperspectral Signatures

Spectral signature coding (SSC) is generally performed by encoding spectral values of a signature across its spectral coverage followed by the Hamming distance to measure signature similarity. The effectiveness of such an SSC largely relies on how well the Hamming distance can capture spectral variations that characterize a signature. Unfortunately, in most cases, this Hamming-distance-based SSC does not provide sufficient discriminatory information for signature analysis because the Hamming distance does not take into account the band-to-band variation, in which case the Hamming distance can be considered as a memoryless distance. This paper extends the Hamming-distance-based SSC to an approach, referred to as spectral feature probabilistic coding (SFPC), which introduces a new concept into SSC that uses a criterion with memory to measure spectral similarity. It implements the well-known arithmetic coding (AC) in two ways to encode a signature in a probabilistic manner, called circular SFPC and split SFPC. The values resulting from the AC is then used to measure the distance between two spectral signatures. In order to demonstrate advantages of using AC-based SSC in signature analysis, a comparative analysis is also conducted against spectral binary coding.

PPI-SVM-Iterative FLDA Approach to Unsupervised Multispectral Image Classification

This paper presents a new approach to unsupervised classification for multispectral imagery. It first implements the pixel purity index (PPI) which is commonly used in hyperspectral imaging for endmember extraction to find seed samples without prior knowledge, then uses the PPI-found samples as support vectors for a kernel-based support vector machine (SVM) to generate a set of initial training samples. In order to mitigate randomness caused by PPI and sensitivity of support vectors used by SVM it further develops an iterative Fishers linear discriminate analysis (IFLDA) that performs FLDA classification iteratively to produce a final set of training samples that will be used to perform a follow-up supervised classification. However, when the image is very large, which is usually the case in multispectral imagery, the computational complexity will be very high for PPI to process the entire image. To resolve this issue a Gaussian pyramid image processing is introduced to reduce image size. The experimental results show the proposed approach has great promise in unsupervised multispectral classification.

Deriving the Spatiotemporal NPP Pattern in Terrestrial Ecosystems of Mongolia Using MODIS Imagery

Abstract Net primary production (NPP) is a carbon cycle process that is examined within terrestrial ecosystems. Exploring the distribution of nationwide NPP helps to diagnose the response of ecosystems to natural/anthropogenic influences and resource management. Based on the CASA model, the MODIS-NDVI-derived spatiotemporal pattern of Mongolian NPP was analyzed by factorial ANOVA and regression analysis. Results showed that the nationwide distribution of NPP was coincidence with the distribution of terrestrial ecosystems. During the growing season, the monthly-NPP average of every terrestrial ecosystem behaved temporally as an inverse-U shape that peaked in June/July and varied as a power and logarithm function of the monthly average temperature and precipitation respectively. The desert had an insignificant growth of NPP during the growing season, while the forest, grassland, and desert steppe had a significant positive-growth in April/June period and then a significant negative-growth in July/October. Interannual NPP showed insignificantchange during a five-year period.

An IPCC-Compliant Technique for Forest Carbon Stock Assessment Using Airborne LiDAR-Derived Tree Metrics and Competition Index

This study developed an IPCC (Intergovernmental Panel on Climate Change) compliant method for the estimation of above-ground carbon (AGC) in forest stands using remote sensing technology. A multi-level morphological active contour (MMAC) algorithm was employed to obtain tree-level metrics (tree height (LH), crown radius (LCR), competition index (LCI), and stem diameter (LDBH)) from an airborne LiDAR-derived canopy height model. Seven biomass-based AGC models and 13 volume-based AGC models were developed using a training dataset and validated using a separate validation dataset. Four accuracy measures, mean absolute error (MAE), root-mean-square error (RMSE), percentage RMSE (PRMSE), and root-mean-square percentage error (RMSPE) were calculated for each of the 20 models. These measures were transformed into a new index, accuracy improvement percentage (AIP), for post hoc testing of model performance in estimating forest stand AGC stock. Results showed that the tree-level AGC models explained 84% to 91% of the variance in tree-level AGC within the training dataset. Prediction errors (RMSEs) for these models ranged between 15 ton/ha and 210 ton/ha in mature forest stands, which is equal to an error percentage in the range 6% to 86%. At the stand-level, several models achieved accurate and reliable predictions of AGC stock. Some models achieved 90% to 95% accuracy, which was equal to or superior to the R-squared of the tree-level AGC models. The first recommended model was a biomass-based model using the metrics LDBH, LH, and LCI and the others were volume-based models using LH, LCI, and LCR and LDBH and LH. One metric, LCI, played a critical role in upgrading model performance when banded together with LH and LCR or LDBH and LCR. We conclude by proposing an IPCC-compatible method that is suitable for calculating tree-level AGC and predicting AGC stock of forest stands from airborne LiDAR data.

Classification of Tree Species in Overstorey Canopy of Subtropical Forest Using QuickBird Images.

This paper proposes a supervised classification scheme to identify 40 tree species (2 coniferous, 38 broadleaf) belonging to 22 families and 36 genera in high spatial resolution QuickBird multispectral images (HMS). Overall kappa coefficient (OKC) and species conditional kappa coefficients (SCKC) were used to evaluate classification performance in training samples and estimate accuracy and uncertainty in test samples. Baseline classification performance using HMS images and vegetation index (VI) images were evaluated with an OKC value of 0.58 and 0.48 respectively, but performance improved significantly (up to 0.99) when used in combination with an HMS spectral-spatial texture image (SpecTex). One of the 40 species had very high conditional kappa coefficient performance (SCKC ≥ 0.95) using 4-band HMS and 5-band VIs images, but, only five species had lower performance (0.68 ≤ SCKC ≤ 0.94) using the SpecTex images. When SpecTex images were combined with a Visible Atmospherically Resistant Index (VARI), there was a significant improvement in performance in the training samples. The same level of improvement could not be replicated in the test samples indicating that a high degree of uncertainty exists in species classification accuracy which may be due to individual tree crown density, leaf greenness (inter-canopy gaps), and noise in the background environment (intra-canopy gaps). These factors increase uncertainty in the spectral texture features and therefore represent potential problems when using pixel-based classification techniques for multi-species classification.

Comparison of carbon sequestration potential in agricultural and afforestation farming systems.

In the last few decades, many forests have been cut down to make room for cultivation and to increase food or energy crops production in developing countries. In this study, carbon sequestration and wood production were evaluated on afforested farms by integrating the Gaussian diameter distribution model and exponential diameter-height model derived from sample plots of an afforested hardwood forest in Taiwan. The quantity of sequestrated carbon was determined based on aboveground biomass. Through pilot tests run on an age-volume model, an estimation bias was obtained and used to correct predicted volume estimates for a farm forest over a 20-year period. An estimated carbon sequestration of 11,254 t C was observed for a 189ha-hardwood forest which is equivalent to 41,264 t CO2. If this amount of carbon dioxide were exchanged on the Chicago Climate Exchange (CCX) market, the income earned would be 821 US

An Empirical Model-based Method for Signal Restoration of SWIR in ASD Field Spectroradiometry

ha - 1. Carbon sequestration from rice (Oryza sativa) or sugarcane (Saccharum officinarum) production is discharged as a result of straw decomposition in the soil which also improves soil quality. Sugarcane production does not contribute significantly to carbon sequestration, because almost all the cane fiber is used as fuel for sugar mills. As a result of changing the farming systems to hardwood forest in this study area, carbon sequestration and carbon storage have increased at the rate of 2.98 t C ha - 1 year - 1. Net present value of afforestation for a 20-year period of carbon or wood management is estimated at around US