Global ETD Search

191	Incorporating Scene Depth in Discriminative Correlation Filters for Visual Tracking Stynsberg, John January 2018 (has links) Visual tracking is a computer vision problem where the task is to follow a targetthrough a video sequence. Tracking has many important real-world applications in several fields such as autonomous vehicles and robot-vision. Since visual tracking does not assume any prior knowledge about the target, it faces different challenges such occlusion, appearance change, background clutter and scale change. In this thesis we try to improve the capabilities of tracking frameworks using discriminative correlation filters by incorporating scene depth information. We utilize scene depth information on three main levels. First, we use raw depth information to segment the target from its surroundings enabling occlusion detection and scale estimation. Second, we investigate different visual features calculated from depth data to decide which features are good at encoding geometric information available solely in depth data. Third, we investigate handling missing data in the depth maps using a modified version of the normalized convolution framework. Finally, we introduce a novel approach for parameter search using genetic algorithms to find the best hyperparameters for our tracking framework. Experiments show that depth data can be used to estimate scale changes and handle occlusions. In addition, visual features calculated from depth are more representative if they were combined with color features. It is also shown that utilizing normalized convolution improves the overall performance in some cases. Lastly, the usage of genetic algorithms for hyperparameter search leads to accuracy gains as well as some insights on the performance of different components within the framework. Tracking Visual Deep Learning Machine Learning CNN Convolutional Neural Network Unsupervised Learning Clustering Genetic Algorithms Features Visual featues Channel Coding RGBD Scene Depth Map Kinect Discriminative Correlation Filters SRDCF DCF Spatial Spatially Regularized Hyperparameter Search Occlusion Detection Handling Kalman Filters Normalized Convolution Bayesian Gaussian Mixture Scale Estimation Conjugate Gradient Linkoping Sweden Visuell Följning Särdrag Djupa Faltningsnätverk Maskininlärning Djup Inlärning Genetiska Algoritmer Klustring Djup RGBD Linköping Sverige
192	Multispectral imaging of Sphagnum canopies: measuring the spectral response of three indicator species to a fluctuating water table at Burns Bog Elves, Andrew 02 May 2022 (has links) Northern Canadian peatlands contain vast deposits of carbon. It is with growing urgency that we seek a better understanding of their assimilative capacity. Assimilative capacity and peat accumulation in raised bogs are linked to primary productivity of resident Sphagnum species. Understanding moisture-mediated photosynthesis of Sphagnum spp. is central to understanding peat production rates. The relationship between depth to water table fluctuation and spectral reflectance of Sphagnum moss was investigated using multispectral imaging at a recovering raised bog on the southwest coast of British Columbia, Canada. Burns Bog is a temperate oceanic ombrotrophic bog. Three ecohydrological indicator species of moss were chosen for monitoring: S. capillifolium, S. papillosum, and S. cuspidatum. Three spectral vegetation indices (SVIs) were used to characterize Sphagnum productivity: the normalized difference vegetation index 660, the chlorophyll index, and the photochemical reflectance index. In terms of spectral sensitivity and the appropriateness of SVIs to species and field setting, we found better performance for the normalized difference vegetation index 660 in the discrimination of moisture mediated species-specific reflectance signals. The role that spatiotemporal scale and spectral mixing can have on reflectance signal fidelity was tested. We were specifically interested in the relationship between changes in the local water table and Sphagnum reflectance response, and whether shifting between close spatial scales can affect the statistical strength of this relationship. We found a loss of statistical significance when shifting from the species-specific cm2 scale to the spectrally mixed dm2 scale. This spatiospectral uncoupling of the moisture mediated reflectance signal has implications for the accuracy and reliability of upscaling from plot based measurements. In terms of species-specific moisture mediated reflectance signals, we were able to effectively discriminate between the three indicator species of Sphagnum along the hummock-to-hollow gradient. We were also able to confirm Sphagnum productivity and growth outside of the vascular growing season, establishing clear patterns of reflectance correlated with changes in the local moisture regime. The strongest relationships for moisture mediated Sphagnum productivity were found in the hummock forming species S. capillifolium. Each indicator Sphagnum spp. of peat has distinct functional traits adapted to its preferred position along the ecohydrological gradient. We also discovered moisture mediated and species-specific reflectance phenologies. These phenospectral characteristics of Sphagnum can inform future monitoring work, including the creation of a regionally specific phenospectral library. It’s recommended that further close scale multispectral monitoring be carried out incorporating more species of moss, as well as invasive and upland species of concern. Pervasive vascular reflectance bias in remote sensing products has implications for the reliability of peatland modelling. Avoiding vascular bias, targeted spectral monitoring of Sphagnum indicator species provides a more reliable measure for the modelling of peatland productivity and carbon assimilation estimates. / Graduate peatland peatland restoration peatland monitoring peatland modelling peat peat moss multispectral multispectral imaging multicamera array multitemporal multiple linear regression linear mixed effects Sphagnum sphagna Sphagnum capillifolium Sphagnum papillosum Sphagnum cuspidatum spatiospectral uncoupling near-sensing remote sensing phenospectral phenology phenologies phenospectral bias phenospectral libraries carbon assimilation carbon dioxide carbon dynamics carbon emissions carbon geographies carbon mineralization carbon sequestration carbon source carbon sink carbon storage irrecoverable carbon climate forcing methane Ecological Restoration restoration ecology ecohydrology ecohydrological restoration hydrology ecohydrological gradients ecological integrity reflectance nature based solutions ombrotrophic moss raised peatland mires mires photochemical reflectance index normalized difference vegetation index chlorophyll index climate change climate forcing climate envelope climatope nonvascular vascular bias plant functional type plant functional types structure equation models spectral vegetation indices hummock hollow acrotelm catotelm diplotelmic biocrusts organic soil periodicity moisture deficit senescence rejuvenescence productivity photosynthesis light use efficiency photosynthetic function photosynthetically active radiation water table hydrology rewetting depth to water table landscape change landscape degradation peat subsidence Burns Bog

Search results

Incorporating Scene Depth in Discriminative Correlation Filters for Visual Tracking

Multispectral imaging of Sphagnum canopies: measuring the spectral response of three indicator species to a fluctuating water table at Burns Bog