Ground Truth Measurements Research Articles

Unoccupied aerial system enabled functional modeling of maize height reveals dynamic expression of loci.

Unoccupied aerial systems (UAS) were used to phenotype growth trajectories of inbred maize populations under field conditions. Three recombinant inbred line populations were surveyed on a weekly basis collecting RGB images across two irrigation regimens (irrigated and non‐irrigated/rain fed). Plant height, estimated by the 95th percentile (P95) height from UAS generated 3D point clouds, exceeded 70% correlation (r) to manual ground truth measurements and 51% of experimental variance was explained by genetics. The Weibull sigmoidal function accurately modeled plant growth (R 2: >99%; RMSE: <4 cm) from P95 genetic means. The mean asymptote was strongly correlated (r 2 = 0.66–0.77) with terminal plant height. Maximum absolute growth rates (mm/day) were weakly correlated with height and flowering time. The average inflection point ranged from 57 to 60 days after sowing (DAS) and was correlated with flowering time (r 2 = 0.45–0.68). Functional growth parameters (asymptote, inflection point, growth rate) alone identified 34 genetic loci, each explaining 3–15% of total genetic variation. Plant height was estimated at one‐day intervals to 85 DAS, identifying 58 unique temporal quantitative trait loci (QTL) locations. Genomic hotspots on chromosomes 1 and 3 indicated chromosomal regions associated with functional growth trajectories influencing flowering time, growth rate, and terminal growth. Temporal QTL demonstrated unique dynamic expression patterns not previously observable, and no QTL were significantly expressed throughout the entire growing season. UAS technologies improved phenotypic selection accuracy and permitted monitoring traits on a temporal scale previously infeasible using manual measurements, furthering understanding of crop development and biological trajectories.

Simultaneous Motion and Distortion Correction Using Dual-Echo Diffusion-Weighted MRI.

Geometric distortions resulting from large pose changes reduce the accuracy of motion measurements and interfere with the ability to generate artifact-free information. Our goal is to develop an algorithm and pulse sequence to enable motion-compensated, geometric distortion compensated diffusion-weighted MRI, and to evaluate its efficacy in correcting for the field inhomogeneity and position changes, induced by large and frequent head motions. Dual echo planar imaging (EPI) with a blip-reversed phase encoding distortion correction technique was evaluated in five volunteers in two separate experiments and compared with static field map distortion correction. In the first experiment, dual-echo EPI images were acquired in two head positions designed to induce a large field inhomogeneity change. A field map and a distortion-free structural image were acquired at each position to assess the ability of dual-echo EPI to generate reliable field maps and enable geometric distortion correction in both positions. In the second experiment, volunteers were asked to move to multiple random positions during a diffusion scan. Images were reconstructed using the dual-echo correction and a slice-to-volume registration (SVR) registration algorithm. The accuracy of SVR motion estimates was compared to externally measured ground truth motion parameters. Our results show that dual-echo EPI can produce slice-level field maps with comparable quality to field maps generated by the reference gold standard method. We also show that slice-level distortion correction improves the accuracy of SVR algorithms as slices acquired at different orientations have different levels of distortion, which can create errors in the registration process. Dual-echo acquisitions with blip-reversed phase encoding can be used to generate slice-level distortion-free images, which is critical for motion-robust slice to volume registration. The distortion corrected images not only result in better motion estimates, but they also enable a more accurate final diffusion image reconstruction.

Retrieval of Snow Depth and Snow Water Equivalent Using Dual Polarization SAR Data

This paper deals with the retrieval of snow depth (SD) and snow water equivalent (SWE) using dual-polarization (HH-VV) synthetic aperture radar (SAR) data. The effect of different snowpack conditions on the SD and SWE inversion accuracy was demonstrated by using three TerraSAR-X acquisitions. The algorithm is based on the relationship between the SD, the co-polar phase difference (CPD), and particle anisotropy. The Dhundi observatory in the Indian Himalaya was selected as a validation test site where a field campaign was conducted for ground truth measurements in January 2016. Using the field measured values of the snow parameters, the particle anisotropy has been optimized and provided as an input to the SD retrieval algorithm. A spatially variable snow density ( ρ s ) was used for the estimation of the SWE, and a temporal resolution of 90 m was achieved in the inversion process. When the retrieval accuracy was tested for different snowpack conditions, it was found that the proposed algorithm shows good accuracy for recrystallized dry snowpack without distinct layering and low wetness (w). The statistical indices, namely, the root mean square error (RMSE), the mean absolute difference (MAD), and percentage error (PE), were used for the accuracy assessment. The algorithm was able to retrieve SD with an average MAE and RMSE of 6.83 cm and 7.88 cm, respectively. The average MAE and RMSE values for SWE were 17.32 mm and 21.41 mm, respectively. The best case PE in the SD and the SWE retrieval were 8.22 cm and 18.85 mm, respectively.

Quantitative Estimation of Soil Salinization in an Arid Region of the Keriya Oasis Based on Multidimensional Modeling

Soil salinity is one of the major factors causing land degradation and desertification on earth, especially its important damage to farming activities and land-use management in arid and semiarid regions. The salt-affected land is predominant in the Keriya River area of Northwestern China. Then, there is an urgent need for rapid, accurate, and economical monitoring in the salt-affected land. In this study, we used the electrical conductivity (EC) of 353 ground-truth measurements and predictive capability parameters of WorldView-2 (WV-2), such as satellite band reflectance and newly optimum spectral indices (OSI) based on two dimensional and three-dimensional data. The features of spectral bands were extracted and tested, and different new OSI and soil salinity indices using reflectance of wavebands were built, in which spectral data was pre-processed (based on First Derivative (R-FD), Second Derivative (R-SD), Square data (R-SQ), Reciprocal inverse (1/R), and Reciprocal First Derivative (1/R-FD)), utilizing the partial least-squares regression (PLSR) method to construct estimation models and mapping the regional soil-affected land. The results of this study are the following: (a) the new OSI had a higher relevance to EC than one-dimensional data, and (b) the cross-validation of established PLSR models indicated that the β-PLSR model based on the optimal three-band index with different process algorithm performed the best result with R2V = 0.79, Root Mean Square Errors (RMSEV) = 1.51 dS·m−1, and Relative Percent Deviation (RPD) = 2.01 and was used to map the soil salinity over the study site. The results of the study will be helpful for the study of salt-affected land monitoring and evaluation in similar environmental conditions.

Insight into the fundamental trade-offs of diffusion MRI from polarization-sensitive optical coherence tomography in ex vivo human brain

In the first study comparing high angular resolution diffusion MRI (dMRI) in the human brain to axonal orientation measurements from polarization-sensitive optical coherence tomography (PSOCT), we compare the accuracy of orientation estimates from various dMRI sampling schemes and reconstruction methods. We find that, if the reconstruction approach is chosen carefully, single-shell dMRI data can yield the same accuracy as multi-shell data, and only moderately lower accuracy than a full Cartesian-grid sampling scheme. Our results suggest that current dMRI reconstruction approaches do not benefit substantially from ultra-high b-values or from very large numbers of diffusion-encoding directions. We also show that accuracy remains stable across dMRI voxel sizes of 1 ​mm or smaller but degrades at 2 ​mm, particularly in areas of complex white-matter architecture. We also show that, as the spatial resolution is reduced, axonal configurations in a dMRI voxel can no longer be modeled as a small set of distinct axon populations, violating an assumption that is sometimes made by dMRI reconstruction techniques. Our findings have implications for in vivo studies and illustrate the value of PSOCT as a source of ground-truth measurements of white-matter organization that does not suffer from the distortions typical of histological techniques.

Triple-frequency multi-GNSS reflectometry snow depth retrieval by using clustering and normalization algorithm to compensate terrain variation

Snow is an important water resource and plays a critical role in the hydrologic cycle. Accurate measurements of snow depth are needed by scientists to set up a more refined meteorology–hydrology model. Recently, the Global Navigation Satellite System Reflectometry (GNSS-R) has been developed and applied for snow depth monitoring, with low cost and high resolution. We propose an improved snow depth retrieval method using a combination of GNSS triple-frequency carrier phase. The topographic feature of the reflecting surface is considered for estimating the snow depth by using the density-based spatial clustering of applications with noise algorithm and normalization method. Observables from the GNSS station in Alaska, USA, are used to monitor snow depth and compared with the ground-truth measurements. Compared with the traditional triple-frequency snow depth retrieval method, the new approach has better performance for Galileo and BDS. The RMSE of the snow depth estimate reduces by nearly 40%, and the correlation coefficient increases from 0.93 to 0.97 for Galileo and from 0.91 to 0.95 for BDS, respectively. The research findings show no notable deviations on snow depth average estimation between Galileo and BDS observations compared to the GPS ones. Moreover, the solution with the proposed method results in improving spatial resolution due to the increasing number of satellites and better azimuth coverage.

Abstract WP395: Detection of Hemorrhagic Expansion With Ai

Background and Purpose: Intracerebral hemorrhage (ICH) expansion is an independent predictor of mortality and functional outcome with each milliliter of expansion increasing the chance of functional dependence by up to 7%. Unfortunately, detection of ICH expansion is often subjective, inaccurate, and may misguide treatment pathways. Artificial intelligence with convolutional neural networks (CNNs) represents a powerful new technology for image analysis and quantification. This study compares the accuracy, sensitivity, and specificity between a CNN optimized for ICH volume quantification and a traditional ABC/2 method. Materials and Methods: We performed a retrospective analysis of ICH patients who have had at least one follow-up non-contrast head CT (NCCT) within 24 hours. ICH expansion was defined as &gt;33% volume of expansion, corresponding to a 10% increase in diameter. Each ICH was manually segmented, which served as ground truth measurements. Comparison of ICH expansion was made using (1) a traditional ABC/2 estimative approach and (2) a previously validated hybrid 3D/2D mask ROI-based CNN for ICH evaluation, which was trained previously on over 10,000 patients. Accuracy, sensitivity, and specificity of the CNN and ABC/2 approaches were then compared. Results: A total of 230 patients were included for a total of 460 NCCTs. The average ICH volume was 44.8 mL. The average ICH volume for the CNN was 45.3 mL (Pearson 0.99) and for ABC/2 was 60.4 mL (Pearson 0.81). Accuracy, sensitivity, and specificity for ICH expansion detection was 100%, 100%, and 100% for the CNN and 93.0%, 74.2%, and 96.0% for ABC/2. On visual inspection, cases of false positives by ABC/2 approaches tended to demonstrate eccentric expansion (Figure 1). Conclusions: A customized deep learning tool is highly accurate in the detection of ICH expansion. This may have important implications clinically for management and surveillance as well as in a clinical trial setting.

Development and evaluation of a self-propelled electric platform for high-throughput field phenotyping in wheat breeding trials

The use of high-throughput phenotyping systems in crop research offers a powerful alternative to traditional methods for understanding plant behaviours. These systems provide a rapid, consistent, repeatable, non-destructive and objective sampling method to quantify complex and previously unobtainable traits at relatively fine resolutions. In this study, a field-based high-throughput phenotyping solution for wheat was developed using a sensor suite mounted on a self-propelled electric platform. A 2D LiDAR was used to scan wheat plots from overhead, while an odometry system was used as a local navigation system to determine the precise plot/plant/scan location. Accurate 3D models of the scanned wheat plots were reconstructed based on the recorded LiDAR and odometry data. Seven plots of different wheat cultivars were scanned to calculate the canopy height using LiDAR data, and these results were compared with manual ground truth measurements. Additionally, in each of these seven plots, the NDVI and PRI spectral indices were calculated using low-cost spectral reflectance sensors (SRSs) and an expensive visible/near-infrared (VIS/NIR) spectral analysis system used for reference purposes. The results of the validation showed good agreement between the LiDAR and manual wheat plant height measurements with an R2 of 0.73 and RMSE = 2.63 cm for three days of campaign measurements. A statistically significant linear correlation was observed between the NDVI values obtained with the reference spectrometer and the low-cost SRS; the coefficients of determination were R2 = 0.69 for day 1 and R2 = 0.81 for day 2, suggesting a similar degree of accuracy among both sensing systems. The developed platform and the obtained wheat phenotyping results demonstrated the suitability of the system for acquiring reliable data under field conditions while maintaining a constant low speed and stability during field deployment. The adaptability of the platform to the structure of the crop and the repeatability of data collection throughout the growing season make the system suitable for integration into commercial breeding programmes.

Detection and Characterization of Rumble Strips in Roadway Video Logs

State departments of transportation often maintain extensive “video logs” of their roadways that include signs, lane markings, as well as non-image-based information such as grade, curvature, etc. In this work we use the Roadway Information Database (RID), developed for the Second Strategic Highway Research Program, as a surrogate for a video log to design and test algorithms to detect rumble strips in the roadway images. Rumble strips are grooved patterns at the lane extremities designed to produce an audible queue to drivers who are in danger of lane departure. The RID contains 6,203,576 images of roads in six locations across the United States with extensive ground truth information and measurements, but the rumble strip measurements (length and spacing) were not recorded. We use an image correction process along with automated feature extraction and convolutional neural networks to detect rumble strip locations and measure their length and pitch. Based on independent measurements, we estimate our true positive rate to be 93% and false positive rate to be 10% with errors in length and spacing on the order of 0.09 meters RMS and 0.04 meters RMS. Our results illustrate the feasibility of this approach to add value to video logs after initial capture as well as identify potential methods for autonomous navigation.

Retrieval of tree branch architecture attributes from terrestrial laser scan data using a Laplacian algorithm

Tree architecture, defined as the three-dimensional arrangement of tree above-ground elements, directly influences the biological and physical processes of vegetation such as photosynthesis and evapotranspiration. Accurate description of tree architecture is of central importance to understand the above biophysical processes. Terrestrial laser scanning (TLS) has been proved to be a promising tool to quantitatively describe tree architecture parameters. However, previous studies using TLS usually focused on architectural parameter measurements at individual tree, crown scale and leaf scales. Very few studies have achieved a comprehensive quantitative description of branch architecture (including angle, diameter, length and volume). In this study, we improved the Laplacian-Based Contraction skeletonization algorithm using the Dijkstra algorithm, developed a new path discrimination method to identify and encode branch orders, and retrieved branch architecture parameters based on branch order and topology information. To assess the influence of branching complexity and branching pattern on the estimation accuracy, we scanned 15 different sized magnolia trees without a leading stem and simulated 10 different sized trees with a leading stem. Results showed the overall branch order identification and parameters retrieval accuracy of trees with a leading stem was obviously higher than trees without a leading stem. The identification accuracy of branch order decreased with the increase in the number of branch and tree branching complexity. The estimated branch architecture parameters agreed well with ground truth measurements (R2 up to 0.99), except for the second- and third-order branch volume. Compared with branch angle and diameter, branch length showed the best correlations with manually measured values (0.14 vs 0.002, 8.48 in RMSE; 0.99 vs 0.99, 0.78 in R2). The second-and third-order branch volume estimations were highly underestimated compared with the ground truth values (R2 = 0.53, RMSE = 0.0239 and R2 = 0.70, RMSE = 0.0257 respectively). This study demonstrated that TLS was an effective way to retrieve branch architecture parameters and provided a useful tool for comprehensive studies of biophysical processes and metabolic theories in ecology.

Soil Salinity Detection in Semi-Arid Region Using Spectral Unmixing, Remote Sensing and Ground Truth Measurements

Soil salinity is one of the serious environmental problems ravaging the soils of arid and semi-arid region, thereby affecting crop productivity, livestock, increase level of poverty and land degradation. Hyperspectral remote sensing is one of the important techniques to monitor, analyze and estimate the extent and severity of soil salt at regional to local scale. In this study we develop a model for the detection of salt-affected soils in arid and semi-arid regions and in our case it’s Ghannouch, Gabes. We used fourteen spectral indices and six spectral bands extracted from the Hyperion data. Linear Spectral Unmixing technique (LSU) was used in this study to improve the correlation between electrical conductivity and spectral indices and then improve the prediction of soil salinity as well as the reliability of the model. To build the model a multiple linear regression analysis was applied using the best correlated indices. The standard error of the estimate is about 1.57 mS/cm. The results of this study show that hyperion data is accurate and suitable for differentiating between categories of salt affected soils. The generated model can be used for management strategies in the future.

Evaluation of Backscattering Models and Support Vector Machine for the Retrieval of Bare Soil Moisture from Sentinel-1 Data

The main objective of this work was to retrieve surface soil moisture (SSM) by using scattering models and a support vector machine (SVM) technique driven by backscattering coefficients obtained from Sentinel-1 satellite images acquired over bare agricultural soil in the Tensfit basin of Morocco. Two backscattering models were selected in this study due to their wide use in inversion procedures: the theoretical integral equation model (IEM) and the semi-empirical model (Oh). To this end, the sensitivity of the SAR backscattering coefficients at V V ( σ v v ∘ ) and V H ( σ v h ∘ ) polarizations to in situ soil moisture data were analyzed first. As expected, the results showed that over bare soil the σ v v ∘ was well correlated with SSM compared to the σ v h ∘ , which showed more dispersion with correlation coefficients values (r) of about 0.84 and 0.61 for the V V and V H polarizations, respectively. Afterwards, these values of σ v v ∘ were compared to those simulated by the backscatter models. It was found that IEM driven by the measured length correlation L slightly underestimated SAR backscatter coefficients compared to the Oh model with a bias of about − 0.7 dB and − 1.2 dB and a root mean square (RMSE) of about 1.1 dB and 1.5 dB for Oh and IEM models, respectively. However, the use of an optimal value of L significantly improved the bias of IEM, which became near to zero, and the RMSE decreased to 0.9 dB. Then, a classical inversion approach of σ v v ∘ observations based on backscattering model is compared to a data driven retrieval technic (SVM). By comparing the retrieved soil moisture against ground truth measurements, it was found that results of SVM were very encouraging and were close to those obtained by IEM model. The bias and RMSE were about 0.28 vol.% and 2.77 vol.% and − 0.13 vol.% and 2.71 vol.% for SVM and IEM, respectively. However, by taking into account the difficultly of obtaining roughness parameter at large scale, it was concluded that SVM is still a useful tool to retrieve soil moisture, and therefore, can be fairly used to generate maps at such scales.

ESTIMATION OF FOREST COVERAGE IN NORTHERN REGION OF MONGOLIA USING SENTINEL AND LANDSAT DATA

Abstract. This paper aims to apply Forest Index (FI) and to determine forest coverage in the study area. The study area (49° 15ʹ to 49° 10ʹ N and 104° 05ʹ to 104° 15ʹ E) is located in the northern region of Mongolia and consist of mixed forest. Larch forest (86.12%) is dominating in the study area. The Sentinel-2 satellite data for the years 2015–2019 were used in the research. The land surface temperature (LST) was produced from Landsat-8 OL. FI methodology was applied for the Sentinel data in order to estimate larch forest coverage. The output map of forest coverage was compared with ground truth measurements and thematic map. The agreement between FI map and ground measurement was 85%. LST from Landsat and FI from Sentinel were sampled in to same size. The relationship between LST (Landsat-8) and FI (Sentinel-2) was reasonable (R = 0.5). FI index and LST is applicable for different forest type in the region.

Land cover classification using maximum likelihood method (2000 and 2019) at Khandgait valley in Mongolia

Promoting the recovery of forest management has been identified as a key priority by the Government of Mongolia. The objective of this paper is to define land cover classification and land cover change in Khandgait valley between 2000 and 2019. The study area is located in the North central part of Mongolia in Bulgan province. Landsat satellite images with 30m resolution were applied. For the validation, we used ground truth measurements. Maximum-likelihood method was applied in this study. The output map of land cover classification was analyzed and compared with the ground truth measurements. The results showed an overall accuracy of 86.5% and 89.0% for the 2000 and 2019 images, respectively. Land cover changes were quantitatively presented with the results of accuracy assessments between 2000 and 2019. In the future, we need to improve forest monitoring and analyze forest management using satellite images.

Optimizing the Quality of 4D-DSA Temporal Information.

Quantification of blood flow using a 4D-DSA would be useful in the diagnosis and treatment of cerebrovascular diseases. A protocol optimizing identification of density variations in the time-density curves of a 4D-DSA has not been defined. Our purpose was to determine the contrast injection protocol most likely to result in the optimal pulsatility signal strength. Two 3D-printed patient-specific models were used and connected to a pulsatile pump and flow system, which delivered 250-260 mL/min to the model. Contrast medium (Isovue, 370 mg I/mL, 75% dilution) was injected through a 6F catheter positioned upstream from the inlet of the model. 4D-DSA acquisitions were performed for the following injection rates: 1.5, 2.0, 2.5, 3.0 and 3.5 mL/s for 8 seconds. To determine pulsatility, we analyzed the time-density curve at the inlets using the oscillation amplitude and a previously described numeric metric, the sideband ratio. Vascular geometry from 4D-DSA reconstructions was compared with ground truth and micro-CT measurements of the model. Dimensionless numbers that characterize hemodynamics, Reynolds and Craya-Curtet, were calculated for each injection rate. The strongest pulsatility signal occurred with the 2.5 mL/s injections. The largest oscillation amplitudes were found with 2.0- and 2.5-mL/s injections. Geometric accuracy was best preserved with injection rates of >1.5 mL/s. An injection rate of 2.5 mL/s provided the strongest pulsatility signal in the 4D-DSA time-density curve. Geometric accuracy was best preserved with injection rates above 1.5 mL/s. These results may be useful in future in vivo studies of blood flow quantification.

Gas distribution mapping and source localization using a 3D grid of metal oxide semiconductor sensors

The difficulty to obtain ground truth (i.e. empirical evidence) about how a gas disperses in an environment is one of the major hurdles in the field of mobile robotic olfaction (MRO), impairing our ability to develop efficient gas source localization strategies and to validate gas distribution maps produced by autonomous mobile robots. Previous ground truth measurements of gas dispersion have been mostly based on expensive tracer optical methods or 2D chemical sensor grids deployed only at ground level. With the ever-increasing trend towards gas-sensitive aerial robots, 3D measurements of gas dispersion become necessary to characterize the environment these platforms can explore. This paper presents ten different experiments performed with a 3D grid of 27 metal oxide semiconductor (MOX) sensors to visualize the temporal evolution of gas distribution produced by an evaporating ethanol source placed at different locations in an office room, including variations in height, release rate and air flow. We also studied which features of the MOX sensor signals are optimal for predicting the source location, considering different lengths of the measurement window. We found strongly time-varying and counter-intuitive gas distribution patterns that disprove some assumptions commonly held in the MRO field, such as that heavy gases disperse along ground level. Correspondingly, ground-level gas distributions were rarely useful for localizing the gas source and elevated measurements were much more informative. We make the dataset and the code publicly available to enable the community to develop, validate, and compare new approaches related to gas sensing in complex environments.

Leaf Scanner: A portable and low-cost multispectral corn leaf scanning device for precise phenotyping

During recent years, portable plant phenotyping instruments have become increasingly important to monitor conditions of plant health non-destructively in the greenhouse and field environments. These devices, such as the Soil Plant Analysis Development (SPAD) meter, leverage indices that represent leaf chlorophyll content due to its strong correlation with leaf nitrogen (N) content. However, instruments such as SPAD meters are expensive and measure only individual, restricted leaf regions per data capture. In this publication, we developed a Leaf Scanner device to analyze the chlorophyll content distributions of whole leaves with greater efficiency and precision. The validating samples for this device were top-collared corn leaves grown under high and low N treatments in a greenhouse. For each region of a corn leaf, this device rapidly flashed visible and near infrared (NIR) LEDs to obtain the visible and NIR transmittance images of leaves. These images were summarized pixel-wisely into the Green NDVI index. A sufficient high framerate permitted continuous collection of regional index images. Using image registration techniques, these regional images were stitched together into a ‘leaf panorama’. The total pixel amount of each leaf panorama, thus, served as a substitute for leaf area. A Minolta SPAD-502Plus meter collected ground-truth measurements along the length of each leaf sample. The results showed that there was a strong correlation between the average Leaf Scanner’s measurements and averaged SPAD values (R2 = 0.92), and the Leaf Scanner was able to clearly detect differences between high and low fertilized samples in terms of chlorophyll content and leaf area. Furthermore, this Leaf Scanner device enabled us to study the chlorophyll content distributions on the plants under different treatments.

Developing a visual method to characterize displays

The goal is to develop a display characterization model to include the personal vision characteristics. A two-step model for visually characterizing displays was developed. It was based on the concept of half-toning technique for obtaining gamma factor for each colour channel, and unique hue concept for achieving 3x3 matrix coefficients, respectively. The variation can be presented by the optimized RGB primaries for each observer. The typical difference between the individual and the measured ground truth is 2.2 in terms of CIEDE2000 units.

3D Convolutional Neural Networks for Remote Pulse Rate Measurement and Mapping from Facial Video

Remote pulse rate measurement from facial video has gained particular attention over the last few years. Research exhibits significant advancements and demonstrates that common video cameras correspond to reliable devices that can be employed to measure a large set of biomedical parameters without any contact with the subject. A new framework for measuring and mapping pulse rate from video is presented in this pilot study. The method, which relies on convolutional 3D networks, is fully automatic and does not require any special image preprocessing. In addition, the network ensures concurrent mapping by producing a prediction for each local group of pixels. A particular training procedure that employs only synthetic data is proposed. Preliminary results demonstrate that this convolutional 3D network can effectively extract pulse rate from video without the need for any processing of frames. The trained model was compared with other state-of-the-art methods on public data. Results exhibit significant agreement between estimated and ground-truth measurements: the root mean square error computed from pulse rate values assessed with the convolutional 3D network is equal to 8.64 bpm, which is superior to 10 bpm for the other state-of-the-art methods. The robustness of the method to natural motion and increases in performance correspond to the two main avenues that will be considered in future works.

Ensemble Neural Networks for Modeling DEM Error

Digital elevation models (DEMs) have become ubiquitous and remarkably effective in the field of earth sciences as a tool to characterize surface topography. All DEMs have a degree of inherent error and uncertainty that is propagated to subsequent models and analyses, which can lead to misinterpretation and inaccurate estimates. A new method was developed to estimate local DEM errors and implement corrections while quantifying the uncertainties of the implemented corrections. The method is based on the flexibility and ability to model complex problems with ensemble neural networks (ENNs). The method was developed to be applied to any DEM created from a corresponding set of elevation points (point cloud) and a set of ground truth measurements. The method was developed and tested using hyperspatial resolution terrestrial laser scanning (TLS) data (sub-centimeter point spacing) collected from a marsh site located along the southern portion of the Texas Gulf Coast, USA. ENNs improve the overall DEM accuracy in the study area by 68% for six model inputs and by 75% for 12 model inputs corresponding to root mean square errors (RMSEs) of 0.056 and 0.045 m, respectively. The 12-input model provides more accurate tolerance interval estimates, particularly for vegetated areas. The accuracy of the method is confirmed based on an independent data set. Although the method still underestimates the 95% tolerance interval, 8% below the 95% target, results show that it is able to quantify the spatial variability in uncertainties due to a relationship between vegetation/land cover and accuracy of the DEM for the study area. There are still opportunities and challenges in improving and confirming the applicability of this method for different study sites and data sets.