Gender discriminating models from facial surface normals

In this paper, we show how to use facial shape information to construct discriminating models for gender classification. We represent facial shapes using 2.5D fields of facial surface normals, and investigate three different methods to improve the gender discriminating capacity of the model constructed using the standard eigenspace method. The three methods are novel variants of principal geodesic analysis (PGA) namely (a) weighted PGA, (b) supervised weighted PGA, and (c) supervised PGA. Our starting point is to define a weight map over the facial surface that indicates the importance of different locations in discriminating gender. We show how to compute the relevant weights and how to incorporate the weights into the 2.5D model construction. We evaluate the performance of the alternative methods using facial surface normals extracted from 3D range images or recovered from brightness images. Experimental results demonstrate the effectiveness of our methods. Moreover, the classification accuracy, which is as high as 97%, demonstrates the effectiveness of using facial shape information for gender classification.     

Supervised relevance maps for increasing the distinctiveness of facial images

This paper shows how to improve holistic face analysis by assigning importance factors to different facial regions (termed as face relevance maps). We propose a novel supervised learning algorithm for generating face relevance maps to improve the discriminating capability of existing methods. We have successfully applied the developed technique to face identification based on the Eigenfaces and Fisherfaces methods, and also to gender classification based on principal geodesic analysis (PGA). We demonstrate how to iteratively learn the face relevance map using labelled data. Experimental results confirm the effectiveness of the developed approach.     

Characterizing spatial patterns of invasive species using sub-pixel classifications

Invasive species disrupt landscape patterns and compromise the functionality of ecosystem processes. Non-native saltcedar poses significant threats to native vegetation and groundwater resources in the southwestern U.S. and Mexico, and quantifying spatial and temporal distribution patterns is essential for monitoring its spread. Considerable research focuses on determining the accuracy of various remote sensing techniques for distinguishing saltcedar from native woody riparian vegetation through sub-pixel, or soft classifications. However, there is a lack of research quantifying spatial distribution patterns from these classifications, mainly because landscape metrics, which are commonly used to statistically assess these patterns, require bounded classes and cannot be applied directly to soft classifications. This study tests a new method for discretizing sub-pixel data to generate landscape metrics using a continuum of fractional cover thresholds. The developed approach transforms sub-pixel classifications into discrete maps compliant with metric terms and computes and interprets metric results in the context of the region to explain patterns in the extent, distribution, and connectivity of saltcedar in the Rio Grande basin. Results indicate that landscape metrics are sensitive to sub-pixel values and can vary greatly with fractional cover. Therefore spectral unmixing should be performed prior to metric calculations. Analysis of metric trends provides evidence that saltcedar has expanded away from the immediate riparian zones and is displacing native vegetation. This information, coupled with control management strategies, can be used to target remediation activities along the Rio Grande.     

Land cover classification with coarse spatial resolution data to derive continuous and discrete maps for complex regions

Regularly updated land cover information at continental or national scales is a requirement for various land management applications as well as biogeochemical and climate modeling exercises. However, monitoring or updating of map products with sufficient spatial detail is currently not widely practiced due to inadequate time-series coverage for most regions of the Earth. Classifications of coarser spatial resolution data can be automatically generated on an annual or finer time scale. However, discrete land cover classifications of such data cannot sufficiently quantify land surface heterogeneity or change. This study presents a methodology for continuous and discrete land cover mapping using moderate spatial resolution time series data sets. The method automatically selects sample data from higher spatial resolution maps and generates multiple decision trees. The leaves of decision trees are interpreted considering the sample distribution of all classes yielding class membership maps, which can be used as estimates for the diversity of classes in a coarse resolution cell. Results are demonstrated for the heterogeneous, small-patch landscape of Germany and the bio-climatically varying landscape of South Africa. Results have overall classification accuracies of 80%. A sensitivity analysis of individual modules of the classification process indicates the importance of appropriately chosen features, sample data balanced among classes, and an appropriate method to combine individual classifications. The comparison of classification results over several years not only indicates the method's consistency, but also its potential to detect land cover changes.     

Object-based analysis and change detection of major wetland cover types and their classification uncertainty during the low water period at Poyang Lake, China

Productive wetland systems at land-water interfaces that provide unique ecosystem services are challenging to study because of water dynamics, complex surface cover and constrained field access. We applied object-based image analysis and supervised classification to four 32-m Beijing-1 microsatellite images to examine broad-scale surface cover composition and its change during November 2007-March 2008 low water season at Poyang Lake, the largest freshwater lake-wetland system in China (> 4000 km²). We proposed a novel method for semi-automated selection of training objects in this heterogeneous landscape using extreme values of spectral indices (SIs) estimated from satellite data. Dynamics of the major wetland cover types (Water, Mudflat, Vegetation and Sand) were investigated both as transitions among primary classes based on maximum membership value, and as changes in memberships to all classes even under no change in a primary class. Fuzzy classification accuracy was evaluated as match frequencies between classification outcome and a) the best reference candidate class (MAX function) and b) any acceptable reference class (RIGHT function). MAX-based accuracy was relatively high for Vegetation (≥ 90%), Water (≥ 82%), Mudflat (≥ 76%) and the smallest-area Sand (≥ 75%) in all scenes; these scores improved with the RIGHT function to 87-100%. Classification uncertainty assessed as the proportion of fuzzy object area within a class at a given fuzzy threshold value was the highest for all classes in November 2007, and consistently higher for Mudflat than for other classes in all scenes. Vegetation was the dominant class in all scenes, occupying 41.2-49.3% of the study area. Object memberships to Vegetation mostly declined from November 2007 to February 2008 and increased substantially only in February-March 2008, possibly reflecting growing season conditions and grazing. Spatial extent of Water both declined and increased during the study period, reflecting precipitation and hydrological events. The “fuzziest” Mudflat class was involved in major detected transitions among classes and declined in classification accuracy by March 2008, representing a key target for finer-scale research. Future work should introduce Vegetation sub-classes reflecting differences in phenology and alternative methods to discriminate Mudflat from other classes. Results can be used to guide field sampling and top-down landscape analyses in this wetland.     

Thematic accuracy of the National Land Cover Database (NLCD) 2001 land cover for Alaska

The National Land Cover Database (NLCD) 2001 Alaska land cover classification is the first 30-m resolution land cover product available covering the entire state of Alaska. The accuracy assessment of the NLCD 2001 Alaska land cover classification employed a geographically stratified three-stage sampling design to select the reference sample of pixels. Reference land cover class labels were determined via fixed wing aircraft, as the high resolution imagery used for determining the reference land cover classification in the conterminous U.S. was not available for most of Alaska. Overall thematic accuracy for the Alaska NLCD was 76.2% (s.e. 2.8%) at Level II (12 classes evaluated) and 83.9% (s.e. 2.1%) at Level I (6 classes evaluated) when agreement was defined as a match between the map class and either the primary or alternate reference class label. When agreement was defined as a match between the map class and primary reference label only, overall accuracy was 59.4% at Level II and 69.3% at Level I. The majority of classification errors occurred at Level I of the classification hierarchy (i.e., misclassifications were generally to a different Level I class, not to a Level II class within the same Level I class). Classification accuracy was higher for more abundant land cover classes and for pixels located in the interior of homogeneous land cover patches.     

Mapping wildfire and clearcut harvest disturbances in boreal forests with Landsat time series data

Information regarding the extent, timing and magnitude of forest disturbance are key inputs required for accurate estimation of the terrestrial carbon balance. Equally important for studying carbon dynamics is the ability to distinguish the cause or type of forest disturbance occurring on the landscape. Wildfire and timber harvesting are common disturbances occurring in boreal forests, with each having differing carbon consequences (i.e., biomass removed, recovery rates). Development of methods to not only map, but distinguish these types of disturbance with satellite data will depend upon an improved understanding of their distinctive spectral properties. In this study, we mapped wildfires and clearcut harvests occurring in a Landsat time series (LTS) acquired in the boreal plains of Saskatchewan, Canada. This highly accurate reference map (kappa = 0.91) depicting the year and cause of historical disturbances was used to determine the spectral and temporal properties needed to accurately classify fire and clearcut disturbances. The results showed that spectral data from the short-wave infrared (SWIR; e.g., Landsat band 5) portion of the electromagnetic spectrum was most effective at separating fires and clearcut harvests possibly due to differences in structure, shadowing, and amounts of exposed soil left behind by the two disturbance types. Although SWIR data acquired 1 year after disturbance enabled the most accurate discrimination of fires and clearcut harvests, good separation (e.g., kappa ≥ 0.80) could still be achieved with Landsat band 5 and other SWIR-based indices 3 to 4 years after disturbance. Conversely, minimal disturbance responses in near infrared-based indices associated with green leaf area (e.g., NDVI) led to unreliably low classification accuracies regardless of time since disturbance. In addition to exploring the spectral and temporal manifestation of forest disturbance types, we also demonstrate how Landsat change maps which attribute cause of disturbance can be used to help elucidate the social, ecological and carbon consequences associated with wildfire and clearcut harvesting in Canadian boreal forests.     

Multi-class supervised classification of electrical borehole wall images using texture features

Electrical borehole wall images represent micro-resistivity measurements at the borehole wall. The lithology reconstruction is often based on visual interpretation done by geologists. This analysis is very time-consuming and subjective. Different geologists may interpret the data differently. In this work, linear discriminant analysis (LDA) in combination with texture features is used for an automated lithology reconstruction of ODP (Ocean Drilling Program) borehole 1203A drilled during Leg 197. Six rock groups are identified by their textural properties in resistivity data obtained by a Formation MircoScanner (FMS). Although discriminant analysis can be used for multi-class classification, non-optimal decision criteria for certain groups could emerge. For this reason, we use a combination of 2-class (binary) classifiers to increase the overall classification accuracy. The generalization ability of the combined classifiers is evaluated and optimized on a testing dataset where a classification rate of more than 80% for each of the six rock groups is achieved. The combined, trained classifiers are then applied on the whole dataset obtaining a statistical reconstruction of the logged formation. Compared to a single multi-class classifier the combined binary classifiers show better classification results for certain rock groups and more stable results in larger intervals of equal rock type.