Sarah Jensen MPH, REHS, CPFS

The current health pandemic has precipitated a significant change in vehicular traffic due to quarantine procedures enacted by local and state governments. Such changes have a significant effect on internal transportation affecting such issues as road maintenance and expansion, traffic signal structure, and parking, and traffic fluidity which are all crucial to the movement of resources throughout the state. With predicted modulation in traffic flow density over the next year, there is a need… Show more

The current health pandemic has precipitated a significant change in vehicular traffic due to quarantine procedures enacted by local and state governments. Such changes have a significant effect on internal transportation affecting such issues as road maintenance and expansion, traffic signal structure, and parking, and traffic fluidity which are all crucial to the movement of resources throughout the state. With predicted modulation in traffic flow density over the next year, there is a need for simple and effective ways to estimate the mean number of vehicles on the road passing per unit time. In an effort to find inexpensive but reliable methods for vehicular detection and counting, off-the-shelf panchromatic camera technology and global and local eigenface decompositions based on manifold learning are used to detect the number of cars passing per unit time on a semi-busy road. Such a method represents a different form of vehicular detection based on facial recognition rather than on the usual mask-based machine learning algorithms. Preliminary results suggest that car anomaly detection is possible using eigenface methods based on principal component analysis, Laplacian eigenmaps, and non negative matrix factorization. Mean changes in imagery lighting, however, are shown to cause false positives which impact algorithmic accuracy. Continued research work is aimed at long term quantification of single position, temporal vehicular flow for the purpose of estimating the probability density function of detected vehicles over weekly time scales. Show less

Viola da gambas are extremely resonant instruments whose tone and timbral character provides a unique sound for early music ensembles specializing in medieval, Renaissance, and Baroque music. Construction of the instrument is complex with many different facets influencing the resulting sound including the type of wood used, the glue used for sealant, the varnish applied on the body, the placement of the bridge, as well as the distention of the belly. Wolf tones are a particular annoying… Show more

Viola da gambas are extremely resonant instruments whose tone and timbral character provides a unique sound for early music ensembles specializing in medieval, Renaissance, and Baroque music. Construction of the instrument is complex with many different facets influencing the resulting sound including the type of wood used, the glue used for sealant, the varnish applied on the body, the placement of the bridge, as well as the distention of the belly. Wolf tones are a particular annoying anomalous tonal and timbral phenomenon in the instrument that can have a serious influence on the overall delivery of a musical piece forcing players to adopt different techniques and playing styles that inhibit their potency. An experimental study is conducted to understand the effect of the playing style and string weight suppressors on the structure of wolf tone and timbre existing in d octave scale in a suite of six viola da gambas. Fourier analytical methods and machine learning analytical techniques such as k-means clustering and self organized mapping are used to provide an understanding of similitude and variability of various modulated sounds as different notes are played under different conditions. The results of this work have implications for the mathematical understanding of the quality of viola da gamba sound as well as providing insight into the underlying aesthetics. The works also provides insight into variability of wolf tonal and timbral structure as it exists in variety of viola da gamba created at different times using different methods. Show less

Social media and the entertainment industry have made great strides in popularizing the field of
forensic science, inspiring the next generation of scientists to consider forensic analytical careers and supporting areas as viable areas of study. With the inundation of this information, it is important to provide students with a realistic view of the math and science components of this field. This is to ensure a balanced view of what the applied scientific career entails rather than to… Show more

Social media and the entertainment industry have made great strides in popularizing the field of
forensic science, inspiring the next generation of scientists to consider forensic analytical careers and supporting areas as viable areas of study. With the inundation of this information, it is important to provide students with a realistic view of the math and science components of this field. This is to ensure a balanced view of what the applied scientific career entails rather than to discourage those who are interested. Bayesian statistical analysis is a logical and analytical tool commonly used in such areas as geo-intelligence, food safety, and epidemiology, as well as being of great utility in forensic science. A tri-part educational tutorial based upon the backgrounds of the authors whose areas of expertise incorporate this type of analysis is proposed. The purpose of the tutorial is twofold: provide clarity into the fundamental mechanics of Bayes' rule and the practical utility of how it can be applied to solve analytical problems. The first part of the tutorial centers on the mechanics of Bayes' rule through the use of a toy example, focusing on fundamental ideas which often generate confusion with respect to evidential-based state prediction. The second part seeks to illustrate how the theorem can be used to solve a practical problem that is based on geo-intelligence tracking, and which is also highly relevant to solving spatiotemporal state prediction forensic problems. The third part focuses on explicating how mathematical forensic intelligence skills can be used in very nontraditional ways to address issues involving both physics and search and retrieval of criminal evidence. This educational tutorial should be useful in dispelling myopic views of what elementary, mathematically-based forensic intelligence analysis entails, and in facilitating the right mindset and perspective in the next generation of burgeoning forensic scientists. Show less

Art is a mirror of an artist’s personal inner creativity and abstract conception of their own individual experience. However, because all artists are human there are latent similarities between works which are not always apparent. Machine learning and photonic technology combined can be used to address the issue of artistic similitude in Venetian Mardi Grad masks. Each mask is made of paper mache’ and is decorated with extravagant and vibrant colors providing artistic variance to a set… Show more

Art is a mirror of an artist’s personal inner creativity and abstract conception of their own individual experience. However, because all artists are human there are latent similarities between works which are not always apparent. Machine learning and photonic technology combined can be used to address the issue of artistic similitude in Venetian Mardi Grad masks. Each mask is made of paper mache’ and is decorated with extravagant and vibrant colors providing artistic variance to a set underlying facial mold. The scientific-artistic question addressed is the degree to which the similarity between different masks can be assessed using machine learning algorithms and criteria. Machine learning algorithms are used to analyze three masks and extract potent latent features which describe them. A fourth mask’s features are then distilled and compared to the previous three masks for the purpose of classification. Such a study has implication for identifying latent artistic styles and assessing artistic similarities in works which on the surface seem quite dissimilar. Show less

Military, law enforcement, and intelligence agencies are often tasked with forensic analysis-based operations in which there is a need to quickly assess the probabilistic presence of the wounded in remote positions. Hyperspectral imagery, a proven technology for the detection of anomalous substances in heterogeneous background environments, has seen some use in the detection of the wounded through its use in imaging blood stained clothing. Blood on clothing often appears in different… Show more

Military, law enforcement, and intelligence agencies are often tasked with forensic analysis-based operations in which there is a need to quickly assess the probabilistic presence of the wounded in remote positions. Hyperspectral imagery, a proven technology for the detection of anomalous substances in heterogeneous background environments, has seen some use in the detection of the wounded through its use in imaging blood stained clothing. Blood on clothing often appears in different configurations which changes due to stresses that smear the fluid along the material. The ability to spectrally detect blood stains on clothing under non-optimal photonic conditions is important in assessing the versatility of hyperspectral imaging technology as a detection tool for surface signatures of the wounded. Blood detection under conditions of smearing is explored for the purpose of investigating the degree to which noisy hyperspectral imagery and machine learning algorithms can detect various configurations of blood stains on camouflage clothing. Basic image segmentation of deer blood stains is demonstrated to be achievable using non-negative matrix factorization, competitive leaky learning, and Laplacian eigenmap-based anomaly detection for three blood smear conditions. Distance metrics based on the Kullback-Leibler divergence and the Bhattacharyya distance for blood stained cloth and clean cloth demonstrate decreasing values with increasing blood smear. These results are consistent with expected image clustering trends. Hyperspectral imagery characterization of blood stains on clothing is the first step toward the creation of inexpensive multispectral imagery technology capable of performing remote sensing-based blood stain image segmentation. Show less

Panchromatic imagery based environmental engineering and military remote sensing platforms are often responsible for monitoring coastal regions in order to safeguard national waters. The inability of remote sensing platforms to penetrate the water column at depth because of turbulence-induced sediment-concentration modulation necessitates using models that dynamically link surface and sub-surface structures. Such structural flow assessment necessitates at least two flow facets. The first is the… Show more

Panchromatic imagery based environmental engineering and military remote sensing platforms are often responsible for monitoring coastal regions in order to safeguard national waters. The inability of remote sensing platforms to penetrate the water column at depth because of turbulence-induced sediment-concentration modulation necessitates using models that dynamically link surface and sub-surface structures. Such structural flow assessment necessitates at least two flow facets. The first is the differentiation and classification of different sub-surface turbulent flow regimes from remotely sensed sub-surface signals so as to definitively understand existing flow states. The second is the ability to determine the sub-surface turbulent state from surface water spatial measurements for flow state assessment and decision-making. Laboratory data from a recent experiment involving panchromatic imagery measurements of a subsurface turbulent bubble water flow was recently analyzed to develop the first part of this signal processor. Sparse reconstruction of sub-surface turbulent flow imagery, a proxy for sonar signal imagery, provides the means for classification of different turbulent bubble flow regimes accomplishing the first part. Modeling results have great implications for satellite and ground based visible spectrum environmental remote sensing of rivers and littoral zones where there is a dire need of operational models to understand internal flow dynamics. Show less

Health agencies and regulators have a continual need and desire for educating the public with respect to the proper method of thawing meat and the potential health risks when inappropriate methods are used. This directive is born from the numerous cases each of year of food-born illnesses which are undeniably due to the creation of thermal conditions which allow for the development of bacteria diseases such as salmonella. Taking advantage of the public’s current interest in artificial… Show more

Health agencies and regulators have a continual need and desire for educating the public with respect to the proper method of thawing meat and the potential health risks when inappropriate methods are used. This directive is born from the numerous cases each of year of food-born illnesses which are undeniably due to the creation of thermal conditions which allow for the development of bacteria diseases such as salmonella. Taking advantage of the public’s current interest in artificial intelligence can be an innovative and alternative method for educating the public as to how computational processing can assist in visualizing thermal dynamics and its relationship to bacterial growth during the meat thawing process. A thermal-biological experiment was performed to illustrate how thermal-temporal variability affects the development of bacteria residing on chicken breasts. The use of machine learning and traditional biological techniques in particular has implications for education via artificial intelligence. Show less

Hundreds of dollars in revenue are lost each year due to food transport loss from kitchen to patrons’ tables in restaurants. This is due to food order mistakes, waiter and waitress theft, and normal food order destruction. Cost saving preemptive measures are needed to account for food products placed on patron’s plates to minimize waste in food transport. Visible spectrum remote sensing combined with machine learning technology, both forms of artificial intelligence, are innovative tools that… Show more

Hundreds of dollars in revenue are lost each year due to food transport loss from kitchen to patrons’ tables in restaurants. This is due to food order mistakes, waiter and waitress theft, and normal food order destruction. Cost saving preemptive measures are needed to account for food products placed on patron’s plates to minimize waste in food transport. Visible spectrum remote sensing combined with machine learning technology, both forms of artificial intelligence, are innovative tools that can efficiently assess food state. In particular, the technology can be exploited for food tracking and accounting to identify food independent of rotated configuration of food products placed on a plate being delivered to a restaurant customer. Rather than store all food images, can the underlying ‘parent’ food template that gives rise to individual food image ‘children’, be estimated? As a first step in the direction of robust food classification, a machine learning experiment and analysis was performed. Show less

Hundreds of dollars in revenue are lost each year due to food transport loss from kitchen to patrons’ tables in restaurants. This is due to food order mistakes, waiter and waitress theft, and normal food order destruction. Cost saving preemptive measures are needed to account for food products placed on patron’s plates to minimize waste in food transport from kitchen to table. Visible spectrum remote sensing combined with machine learning technology, both forms of artificial intelligence, are… Show more

Hundreds of dollars in revenue are lost each year due to food transport loss from kitchen to patrons’ tables in restaurants. This is due to food order mistakes, waiter and waitress theft, and normal food order destruction. Cost saving preemptive measures are needed to account for food products placed on patron’s plates to minimize waste in food transport from kitchen to table. Visible spectrum remote sensing combined with machine learning technology, both forms of artificial intelligence, are innovative tools that can efficiently assess food state where items are positioned in different configurations on a plate relative to an observer. In particular, the technology can be exploited for food tracking and accounting to identify food independent of rotated configuration of food products placed on a plate being delivered to a restaurant customer. Rather than store all food images, can the ‘parent’ food template that gives rise to individual food image ‘children’ which are delivered to customer tables be estimated? As a first step in the direction of robust food classification, a machine learning experiment and analysis was performed. Show less

HSI technology and machine learning has revolutionized meat quality assessment and food safety through the use of a non-invasive technique and information processing allowing for objective comprehension of product structure. A large focus has centered on the NIR and SWIR part of the EM spectrum for meat characterization, but the VNIR also offers promising food structural insight especially for understanding the spectral variability of beef meat as it undergoes the cooking process. Of particular… Show more

HSI technology and machine learning has revolutionized meat quality assessment and food safety through the use of a non-invasive technique and information processing allowing for objective comprehension of product structure. A large focus has centered on the NIR and SWIR part of the EM spectrum for meat characterization, but the VNIR also offers promising food structural insight especially for understanding the spectral variability of beef meat as it undergoes the cooking process. Of particular importance is the development of a fast, efficient remote sensing technology that captures a small subset of spectral features pertinent to parameterizing temperature dependent meat state. This eliminates information redundancy and facilitates quick state assessment of roasted beef for businesses.
State-of-the-art machine learning in combination with HSI is applied to meat at various stages of the cooking process. This illustrates how VNIR HSI technology can facilitate temperature-dependent characterization leading to the development multispectral technology of remotely sensed meat undergoing heat based chemical change. Show less

Studied the presentation of cellular pathology of leukemia samples using electron microscopy.

Studied the water of Buck Creek for the presence of E. Coli in effluent waters as a result of combined sewage overflow discharges during storm events.