Keith Pelletier

An object-based system for LiDAR data fusion and feature extraction

In urbanized areas of the developed world, light detection and ranging (LiDAR) exists alongside a wealth of other geospatial information. Despite this bounty, high-resolution land cover is still lacking in many urban areas. This can be attributed to the complexity of many landscapes, the volume of available data and the challenges associated with combining data that were acquired over differing time periods using inconsistent standards. Object-based approaches are ideal for overcoming these limitations. We describe the design, development and deployment of an object-based system that incorporated LiDAR, imagery and vector data sets to develop a comprehensive, multibillion-pixel land-cover data set for the City of Philadelphia. A novel approach using parallel processing allowed us to distribute the feature extraction load to multiple cores, providing massive gains in efficiency and permitting continual modification of the expert system until the accuracy goals of the project were achieved.

Wetland mapping in the Upper Midwest United States: an object-based approach integrating lidar and imagery data: An object-based approach integrating lidar and imagery data

This study investigated the effectiveness of using high resolution data to map wetlands in three ecoregions in Minnesota. High resolution data included multispectral leaf-off aerial imagery and lidar elevation data. These data were integrated using an Object-Based Image Analysis (OBIA) approach. Results for each study area were compared against field and image interpreted reference data using error matrices, accuracy estimates, and the kappa statistic. Producer’s and user’s accuracies were in the range of 92 to 96 percent and 91 to 96 percent, respectively, and overall accuracies ranged from 96-98 percent for wetlands larger than 0.20 ha (0.5 acres). The results of this study may allow for increased accuracy of mapping wetlands efforts over traditional remote sensing methods.

The Effects of Point or Polygon Based Training Data on RandomForest Classification Accuracy of Wetlands

Wetlands are dynamic in space and time, providing varying ecosystem services. Field reference data for both training and assessment of wetland inventories in the State of Minnesota are typically collected as GPS points over wide geographical areas and at infrequent intervals. This status-quo makes it difficult to keep updated maps of wetlands with adequate accuracy, efficiency, and consistency to monitor change. Furthermore, point reference data may not be representative of the prevailing land cover type for an area, due to point location or heterogeneity within the ecosystem of interest. In this research, we present techniques for training a land cover classification for two study sites in different ecoregions by implementing the RandomForest classifier in three ways: (1) field and photo interpreted points; (2) fixed window surrounding the points; and (3) image objects that intersect the points. Additional assessments are made to identify the key input variables. We conclude that the image object area training method is the most accurate and the most important variables include: compound topographic index, summer season green and blue bands, and grid statistics from LiDAR point cloud data, especially those that relate to the height of the return.

Object-based high-resolution land-cover mapping

There has been a marked increase in availability of high-resolution remotely-sensed datasets over the past eight years. The ability to efficiently extract accurate and meaningful land-cover information from these datasets is crucial if the full potential of these datasets is to be harnessed. Land-cover datasets, particularly high-resolution ones, must be statistically accurate and depict a realistic representation of the landscape if they are to be used by decision makers and trusted by the general public. Furthermore, if such datasets are to be accessible and relevant, mechanisms must exist that facilitate cost-effective and timely production. Object-based image analysis (OBIA) techniques offer the greatest potential for generating accurate and meaningful land-cover datasets in an efficient manner. They overcome the limitations of traditional pixel-based classification methods by incorporating not only spectral data but also spatial and contextual information, and they offer substantial efficiency gains compared to manual interpretation. Drawing on our experience in applying OBIA techniques to high-resolution data, we believe any automated approach to land-cover mapping must: 1) effectively replicate the human image analyst; 2) incorporate datasets from multiple sources; and 3) be capable of processing large datasets. To meet this functionality, an operational OBIA system should: 1) employ expert systems; 2) support vector and raster datasets; and 3) leverage enterprise computing architecture.

An object-based image analysis approach for mapping and monitoring flooding and topographic change near Duluth, Minnesota, USA

A short-term precipitation event near Duluth, Minnesota, USA caused flooding, erosion, and deposition that impacted the natural and anthropogenic landscape. This study quantified these impacts with an object-based image analysis approach that integrated multi-temporal lidar and optical data. Flooding inundated 3% of the study area and impacted 28% of the buildings. Topographic change volumes from erosion and deposition ranged from 36.3 m3 to -32.2 m3. Erosion occurred over 21% of the area while deposition occurred over 11% of the area. These results provide decision managers with a spatially-explicit framework for assessing and monitoring inundation and topographic change in response short-term precipitation events.