Real-Time 3D Head Pose Tracking Through 2.5D Constrained Local Models with Local Neural Fields

Tracking the head in a video stream is a common thread seen within computer vision literature, supplying the research community with a large number of challenging and interesting problems. Head pose estimation from monocular cameras is often considered an extended application after the face tracking task has already been performed. This often involves passing the resultant 2D data through a simpler algorithm that best fits the data to a static 3D model to determine the 3D pose estimate. This work describes the 2.5D constrained local model, combining a deformable 3D shape point model with 2D texture information to provide direct estimation of the pose parameters, avoiding the need for additional optimization strategies. It achieves this through an analytical derivation of a Jacobian matrix describing how changes in the parameters of the model create changes in the shape within the image through a full-perspective camera model. In addition, the model has very low computational complexity and can run in real-time on modern mobile devices such as tablets and laptops. The point distribution model of the face is built in a unique way, so as to minimize the effect of changes in facial expressions on the estimated head pose and hence make the solution more robust. Finally, the texture information is trained via local neural fields—a deep learning approach that utilizes small discriminative patches to exploit spatial relationships between the pixels and provide strong peaks at the optimal locations.

     

Effects of feeding Lunaria oil rich in nervonic and erucic acids on the fatty acid compositions of sphingomyelins from erythrocytes, liver, and brain of the quaking mouse mutant

Feeding an oil from Lunaria biennis rich in 22∶1n−9 and 24∶1n−9 to homozygous quaking (qk.qk) mice caused a large increase in the percentage of 24∶1n−9 and corresponding decreases in the percentage of 24∶0 and 22∶0 in sphingomyelins from liver, erythrocytes, and milk. Brain sphingomyelin from 2-wk-old qk.qk pups born to qk.qk mothers maintained on the Lunaria oil had essentially normal percentage of 24∶1n−9 and 18∶0, in contrast to pups born to mothers maintained on a control oil rich in 18∶1n−9 whose brain sphingomyelin had a markedly reduced percentage of 24∶1n−9 and an increased percentage of 18∶0. After 2 wk and up to and beyond weaning, the qk.qk pups from Lunaria-fed mothers weaned on to the Lunaria diet had a markedly decreased percentage of 24∶1n−9 in their brain sphingomyelin, accompanied by an increased percentage of 18∶0, as compared to heterozygous quaking mice. However, the percentage of 24∶1n−9 in brain sphingomyelin in qk.qk pups weaned on to the Lunaria diet continued throughout this period (2–8 wk postbirth) to be significantly higher than in qk.qk pups weaned on to the control diet. We conclude that dietary 24∶1n−9 influences the fatty acid composition of brain sphingomyelin in qk.qk mice, but only via the mother in pre- or early postnatal animals. We further consider that the dietary effects may be elicited mainly in the sphingomyelin of nonmyelinated brain cells, and that the nervonic acid in myelin sphingomyelin may be formed mainly by chain elongation in oligodendrocytes from shorter chain fatty acid precursors.     

The Effect of Emulsifier Type and Droplet Size on Phase Transitions in Emulsified Even-Numbered <Emphasis Type="Italic">n</Emphasis>-Alkanes

Phase transitions in emulsified even-numbered n-alkanes (C₁₆, C₁₈, and C₂₀) are studied as a function of droplet size (0.15–3.45 μm) and surfactant type (polyoxyethylene sorbitan monolaurate or caseinate) using microcalorimetry (DSC) and ultrasonic attenuation measurements (2.25 MHz). Two DSC exothermic peaks were observed during the heating of C₁₈ and C₂₀ emulsions stabilized by Tween 20: a minor peak around 15 and 25 °C, respectively, and a major double peak about 10 °C higher. We tentatively attribute the minor peak to crystal-rotator phase transition, and the split major peak to melting of the surface and core of the droplets. The C₁₆ emulsions showed similar behavior for the major melting peak (15 °C), but the minor peak was absent-possibly as the sample was not cooled enough to cause the rotator phase to enter the low temperature crystalline state. For similar sodium caseinate stabilized emulsions of C₁₈ and C₂₀, the minor peak was much less pronounced (~25%), which we attribute to the lack of compatibility between the alkane and protein molecules. There were two ultrasonic attenuation peaks for the melting of C₁₈ and C₂₀ and one for C₁₆ corresponding to the DSC peaks. In all cases, the magnitude of the attenuation decreased with increasing particle size. Using an extended scattering theory approach we were able to relate the changes in ultrasonic attenuation to the changes in the effective physical properties of the alkane molecules during melting.     

Concentration Measurement by Acoustic Reflectance

ABSTRACT The speed of sound and the acoustic reflection coefficient of sucrose (0 to 60 wt%), glycerol (0 to 50 wt%), sodium chloride (0 to 27 wt%) solutions and tomato ketchup (0 to 100 wt%) dispersions were measured using a modified pulse-echo technique (2.25 MHz transducer, 20 °C). The density of all solutions is measured using a vibrating tube densitometer. All measured parameters are linear functions of concentration. Ultrasonic reflectance can provide equally precise measurements of concentration as conventional ultrasonic velocity measurements. For all samples except concentrated ketchup (> 50 wt%), measured reflectance is demonstrated to be a function of ultrasonic velocity and density.     

Effect of surface-active stabilizers on the surface properties of coconut milk emulsions

Previously we have demonstrated improved stability of coconut milk emulsions homogenized with various surface-active stabilizers, i.e., 1 wt% sodium caseinate, whey protein isolate (WPI), sodium dodecyl sulfate (SDS), or polyoxyethylene sorbitan monolaurate (Tween 20) [Tangsuphoom, N., & Coupland, J. N. (2008). Effect of surface-active stabilizers on the microstructure and stability of coconut milk emulsions. Food Hydrocolloids, 22(7), 1233–1242]. This study examines the changes in bulk and microstructural properties of those emulsions following thermal treatments normally used to preserve coconut milk products (i.e., −20 °C, −10 °C, 5 °C, 70 °C, 90 °C, and 120 °C). Calorimetric methods were used to determine the destabilization of emulsions and the denaturation of coconut and surface-active proteins. Homogenized coconut milk prepared without additives was destabilized by freeze–thaw, (−20 °C and −10 °C) but not by chilling (5 °C). Samples homogenized with proteins were not affected by low temperature treatments while those prepared with surfactants were stable to chilling but partially or fully coalesced following freeze–thaw. Homogenized coconut milk prepared without additives coalesced and flocculated after being heated at 90 °C or 120 °C for 1 h in due to the denaturation and subsequent aggregation of coconut proteins. Samples emulsified with caseinate were not affected by heat treatments while those prepared with WPI showed extensive coalescence and phase separation after being treated at 90 °C or 120 °C. Samples prepared with SDS were stable to heating but those prepared with Tween 20 completely destabilized by heating at 120 °C.     

Effect of thermal treatments on the properties of coconut milk emulsions prepared with surface-active stabilizers

Previously we have demonstrated improved stability of coconut milk emulsions homogenized with various surface-active stabilizers, i.e., 1 wt% sodium caseinate, whey protein isolate (WPI), sodium dodecyl sulfate (SDS), or polyoxyethylene sorbitan monolaurate (Tween 20) [Tangsuphoom, N., & Coupland, J. N. (2008). Effect of surface-active stabilizers on the microstructure and stability of coconut milk emulsions. Food Hydrocolloids, 22(7), 1233–1242]. This study examines the changes in bulk and microstructural properties of those emulsions following thermal treatments normally used to preserve coconut milk products (i.e., −20 °C, −10 °C, 5 °C, 70 °C, 90 °C, and 120 °C). Calorimetric methods were used to determine the destabilization of emulsions and the denaturation of coconut and surface-active proteins. Homogenized coconut milk prepared without additives was destabilized by freeze–thaw, (−20 °C and −10 °C) but not by chilling (5 °C). Samples homogenized with proteins were not affected by low temperature treatments while those prepared with surfactants were stable to chilling but partially or fully coalesced following freeze–thaw. Homogenized coconut milk prepared without additives coalesced and flocculated after being heated at 90 °C or 120 °C for 1 h in due to the denaturation and subsequent aggregation of coconut proteins. Samples emulsified with caseinate were not affected by heat treatments while those prepared with WPI showed extensive coalescence and phase separation after being treated at 90 °C or 120 °C. Samples prepared with SDS were stable to heating but those prepared with Tween 20 completely destabilized by heating at 120 °C.     

Interactions between β-lactoglobulin and dextran sulfate at near neutral pH and their effect on thermal stability

The effect of interactions between β-lactoglobulin (β-LG) and dextran sulfate (DS) on thermal stability at near neutral pH was investigated. Samples containing 6% w/w β-LG and DS (M_w = 5–500 kDa) at different biopolymer weight ratios, pH (5.6–6.2), and NaCl concentrations (0–30 mM) were heated at 85 °C for 15 min. Turbidity results showed that the presence of DS at appropriate biopolymer weight ratio and pH significantly lowered the turbidity of heated β-LG. Solutions containing DS:β-LG weight ratios of 0.02 or less showed improved heat stability as indicated by decreased turbidity. Analysis of the unheated mixture by size exclusion chromatography coupled with multi-angle laser light scattering (SEC–MALLS) showed an interaction between β-LG and DS. The size of the aggregates increased as pH decreased. The β-LG–DS aggregates had a greater negative charge as seen from electrophoretic mobility measurement. Addition of 30 mM NaCl inhibited complex formation and the effect of DS on reducing the turbidity of heated β-LG, suggesting that the interaction was electrostatic in nature. Other than charge property, the amount and size of native aggregates appeared to be the major factor in determining how DS altered heat-induced aggregation. The presence of DS decreased denaturation temperature of β-LG, indicating that DS did not improve thermal stability of β-LG by stabilizing its native state but rather by altering its aggregation. The results provide information that will facilitate the application of whey proteins and polysaccharides as functional ingredients in foods and beverages.     

Non-contact ultrasonic measurements in food materials

A novel air-coupled transducer and signal generator is used to measure the speed of sound and thickness (5–15 mm) of samples of various food products (cheese, reduced fat cheese, luncheon meat, reduced fat luncheon meat, and cranberry sauce). Measurements were made with two transducers (1 MHz) placed one either side of the sample at a distance of approximately 30 mm from the food surface. Thicknesses measured by this method were similar to measurements made using calipers for all samples. At a single point the ultrasonic measurements were very precise (s.d. 0.003 mm, n=12) but when several points on the same slice were averaged the precision reflected unevenness in the slicing operation. Ultrasonic velocity measurements were made using non-contact and conventional contact-mode ultrasound. The precision of both methods was similar but the apparent speed of sound measurements made in contact mode were significantly higher, suggesting the soft solids studied were compressed during analysis.     

Essential Oils Against Pathogen and Spoilage Microorganisms of Fruit Juices: Use of Versatile Antimicrobial Delivery Systems

Essential oils (EO) are increasingly used as natural antimicrobial compounds, however the effect of delivery system to enhance their antimicrobial activity has not been widely studied. Limonene (0 to 10 μL/mL) was added to microbial suspensions (～10⁵ CFU/mL) of selected foodborne pathogens (Listeria monocytogenes Scott A, Salmonella enterica Typhimurium, Escherichia coli and Staphylococcus aureus), and spoilage microorganisms (Lactobacillus plantarum, Saccharomyces cerevisiae, and Candida albicans). S. aureus was found to be the most sensitive foodborne pathogen while Salmonella enterica showed continued growth under all concentrations. Stable nanoemulsions and solid lipid nanoparticles (SLN) (d ～ 170 nm) were prepared using an alkane carrier oil (n‐tetradecane and n‐eicosane, respectively). Interfacial effects and homogenous distribution of limonene in nanoemulsions improved its (8 and 12 μL/mL) antimicrobial effect against S. aureus. Higher aqueous concentrations as a result of expulsion from SLN further enhanced the antimicrobial activity pronounced at higher limonene concentrations. Therefore, our findings confirm that the emulsion‐based delivery systems are able to effectively distribute limonene inside a microbial suspension to improve its antimicrobial activity.     

Rotationally invariant pattern recognition by use of linear and nonlinear cascaded filters

We discuss the merits of using single-layer (linear and nonlinear) and multiple-layer (nonlinear) filters for rotationally invariant and noise-tolerant pattern recognition. The capability of each approach is considered with reference to a two-class, rotation-invariant, character recognition problem. The minimum average correlation energy (MACE) filter is a linear filter that is generally accepted to be optimal for detecting signals that are free from noise. Here it is found that an optimized MACE filter cannot differentiate between the characters E and F in a rotation-invariant manner. We have found, however, that this task is possible when a single optimized linear filter is used to achieve the required response when a nonlinear threshold function is included after the filter. We show that this structure can be cascaded to form a multiple-layer, cascaded filter and that the capability of such a system is enhanced by its increased noise tolerance in the character recognition problem. Finally, we show the capability of a two-layer cascade as a means to detect different species of bacteria in images obtained from a phase-contrast microscope.