Symposium on the Application of Geophysics to Engineering and Environmental Problems 1998, 1998
A study has recently been conducted to assess the extent of hydrocarbon impacts to groundwater an... more A study has recently been conducted to assess the extent of hydrocarbon impacts to groundwater and soil resources at a petroleum refinery site in Billings, Montana. To accomplish the study, forty-six groundwater monitoring wells were installed at the site. Data collected from the wells included detailed lithologic descriptions from split-spoon samples, cutting returns from air rotary drilling, and suites of geophysical well logs. Because the quality of the lithologic descriptions from the borings was erratic, our approach was to produce lithofacies interpretations based on gamma ray logs input into a neural network classifier system. The type of neural network used was a self-organizing map. This type of network does not require user interpretations, instead, the network categorizes each input vector into a class based on similarity to other input vectors. The number of output classes is determined by the user. The output classifications were then plotted as ‘pseudo-logs’ and correlation performed using these pseudo-logs. Cross sections constructed using conventional well log interpretation and the neural network classifications show good, general agreement. A significant advantage of the neural network approach over a conventional interpretation approach is that all of the well log data are analyzed concurrently preventing inconsistencies that frequently occur with conventional methods. Another major benefit to the neural network approach is the choice of the number of classes which correlates with the level of lithologic detail that can be resolved.
This paper presents a novel hyperspectral feature-extraction toolkit that provides a simple, auto... more This paper presents a novel hyperspectral feature-extraction toolkit that provides a simple, automated, and accurate approach to materials classification from hyperspectral imagery (HSI). The proposed toolkit is built as an extension to the state-of-the-art technology in automated feature extraction, the Feature Analyst software suite. Feature Analyst uses, along with spectral information, feature characteristics such as spatial association, size, shape, texture, pattern, and shadow in its generic feature extraction process. While current HSI techniques, such as spectral endmember classification, can provide effective materials classification, these methods are slow (or manual), cumbersome, complex for analysts, and are limited to materials classification only. Feature Analyst, on the other hand, has a simple workflow of (a) an analyst providing a few examples, and (b) an advanced software agent classifying the rest of the imagery; however, Feature Analyst does not have effective pr...
Airborne LIDAR systems provide useful information from a top-down perspective; however, such sens... more Airborne LIDAR systems provide useful information from a top-down perspective; however, such sensors do not provide information about critical urban features (windows, doors, signs, etc) that are located below rooftops and under tree canopies. Vehicle-mounted terrestrial LIDAR sensors, on the other hand, provide the capability to capture highly accurate 3D measurements of the urban environment with spatial resolutions on the order of 5 centimeters or less. The 3D imaging capability of these collection systems is negated, however, by a lack of software tools and approaches capable of exploiting terrestrial LIDAR datasets with a high degree of automation. Current approaches for creating high-resolution 3D urban models are expensive, requiring for even a small scene thousands of man-hours to digitize feature geometries, assign textures to features, and then attribute features. In this paper we describe a new software system and architecture that provides the following benefits: (a) aut...
Page 1. FINAL REPORT Machine Learning Approach for Target Selection and Threat Classification of ... more Page 1. FINAL REPORT Machine Learning Approach for Target Selection and Threat Classification of Wide Area Survey Data SERDP Project MM-1570 DECEMBER 2007 Jim R. McDonald Science Applications International Corporation ...
Symposium on the Application of Geophysics to Engineering and Environmental Problems 1998, 1998
A study has recently been conducted to assess the extent of hydrocarbon impacts to groundwater an... more A study has recently been conducted to assess the extent of hydrocarbon impacts to groundwater and soil resources at a petroleum refinery site in Billings, Montana. To accomplish the study, forty-six groundwater monitoring wells were installed at the site. Data collected from the wells included detailed lithologic descriptions from split-spoon samples, cutting returns from air rotary drilling, and suites of geophysical well logs. Because the quality of the lithologic descriptions from the borings was erratic, our approach was to produce lithofacies interpretations based on gamma ray logs input into a neural network classifier system. The type of neural network used was a self-organizing map. This type of network does not require user interpretations, instead, the network categorizes each input vector into a class based on similarity to other input vectors. The number of output classes is determined by the user. The output classifications were then plotted as ‘pseudo-logs’ and correlation performed using these pseudo-logs. Cross sections constructed using conventional well log interpretation and the neural network classifications show good, general agreement. A significant advantage of the neural network approach over a conventional interpretation approach is that all of the well log data are analyzed concurrently preventing inconsistencies that frequently occur with conventional methods. Another major benefit to the neural network approach is the choice of the number of classes which correlates with the level of lithologic detail that can be resolved.
This paper presents a novel hyperspectral feature-extraction toolkit that provides a simple, auto... more This paper presents a novel hyperspectral feature-extraction toolkit that provides a simple, automated, and accurate approach to materials classification from hyperspectral imagery (HSI). The proposed toolkit is built as an extension to the state-of-the-art technology in automated feature extraction, the Feature Analyst software suite. Feature Analyst uses, along with spectral information, feature characteristics such as spatial association, size, shape, texture, pattern, and shadow in its generic feature extraction process. While current HSI techniques, such as spectral endmember classification, can provide effective materials classification, these methods are slow (or manual), cumbersome, complex for analysts, and are limited to materials classification only. Feature Analyst, on the other hand, has a simple workflow of (a) an analyst providing a few examples, and (b) an advanced software agent classifying the rest of the imagery; however, Feature Analyst does not have effective pr...
Airborne LIDAR systems provide useful information from a top-down perspective; however, such sens... more Airborne LIDAR systems provide useful information from a top-down perspective; however, such sensors do not provide information about critical urban features (windows, doors, signs, etc) that are located below rooftops and under tree canopies. Vehicle-mounted terrestrial LIDAR sensors, on the other hand, provide the capability to capture highly accurate 3D measurements of the urban environment with spatial resolutions on the order of 5 centimeters or less. The 3D imaging capability of these collection systems is negated, however, by a lack of software tools and approaches capable of exploiting terrestrial LIDAR datasets with a high degree of automation. Current approaches for creating high-resolution 3D urban models are expensive, requiring for even a small scene thousands of man-hours to digitize feature geometries, assign textures to features, and then attribute features. In this paper we describe a new software system and architecture that provides the following benefits: (a) aut...
Page 1. FINAL REPORT Machine Learning Approach for Target Selection and Threat Classification of ... more Page 1. FINAL REPORT Machine Learning Approach for Target Selection and Threat Classification of Wide Area Survey Data SERDP Project MM-1570 DECEMBER 2007 Jim R. McDonald Science Applications International Corporation ...
Uploads
Papers by Stuart Blundell