Processing large remote sensing image data sets on Beowulf clusters

Classification and retrieval on macroinvertebrate image databases

Data conversion From raster to vector (R2V) is a key function in Geographic Information Systems (GIS) and remote sensing (RS) image processing for integrating GIS and RS data. The R2V module is available in commercial RS software packages, but there is still room to improve the computation efficiency. This paper presents an efficient R2V algorithm that processes large images and automatically builds GIS topology while scanning image lines one by one. The new algorithm, termed Two-Arm Chains Edge Tracing (TACET), has several significant advantages. First, it converts all types of area objects of RS classification in only one processing cycle. Secondly, it constructs complete area topological relationship by recording the shared edge between two polygons only once. Finally, it is scalable when processing large images. The program based on the algorithm is faster in processing large RS images with comparison to commercial software such as ENVI.

In an ongoing research project on the ascendancy of statistical visual forms, we have been concerned with the transformations wrought by such images and their organisation as datasets in ‘re-drawing’ knowledge about empirical phenomena. Historians and science studies researchers have long established the generative rather than simply illustrative role of images and figures within scientific practice. More recently, the deployment and generation of images by scientific research and its communication via publication has been impacted by the tools, techniques, and practices of working with large (image) datasets. Against this background, we built a dataset of 10 million-plus images drawn from all preprint articles deposited in the open access repository arXiv from 1991 (its inception) until the end of 2018. In this article, we suggest ways – including algorithms drawn from machine learning that facilitate visually ’slicing’ through the image data and metadata – for exploring large datasets of statistical scientific images. By treating all forms of visual material found in scientific publications – whether diagrams, photographs, or instrument data – as bare images, we developed methods for tracking their movements across a range of scientific research. We suggest that such methods allow us different entry points into large scientific image datasets and that they initiate a new set of questions about how scientific representation might be operating at more-than-human scale.

Sub-pixel image registration on an embedded Nanosatellite Platform

Event Abstract Back to Event A multi-scale spatial and semantic knowledge framework for neuroanatomy Stephen Larson1*, Christopher Aprea1, Sarah Maynard1, Fahim Imam1, Mark H. Ellisman1 and Maryann Martone1 1 UCSD, United States Cellular networks in the brain are fundamentally multi-scale with relevant data derived from subcellular junctional connectivity, cytoarchitectural local connectivity, and long-range topographical connectivity. Experimental limitations make it difficult to study all these scales simultaneously. Consequently, experimental methodologies tend to reveal only a limited aspect of nervous system organization. However, to generate hypotheses across scales, we must analyze the nervous system across spatial dimensions spanning several orders of magnitude. Experimental technologies are now able to reveal organization within these scales, yet the development of tools to synthesize these data into more coherent models of brain structure and function is lagging behind.An ideal knowledge framework for neuroanatomy would provide researchers the ability to 1) upload their data into a common data environment where it can be easily discovered and retrieved, 2) contextualize their data by superimposing and combining related data from multiple scales, and 3) create hypothetical, synthetic views of structures of interest by stitching data together or generalizing patterns found in data. The framework would allow researchers to deal with their data in both spatially, placing it in register in a location in a common brain space, and semantically, annotating and tagging it with standard labels drawn from shared ontologies for neuroscience. These two dimensions of data, spatial and semantic, create a key basis for organizing the heterogenous, scattered knowledge of the brain's structure into a cohesive whole.We have taken a step towards the construction of such a knowledge framework by creating a platform that can serve as its foundation. This platform consists of the following elements 1) a spatial registration workflow for images that allows large 2D data to be registered to a brain atlas 2) a 3D game engine that allows real-time rendering and interaction with 2D and 3D multi scale data, 3) an online collaborative ontology manipulation, maintenance, and retrieval system populated from the Neuroscience Information Framework's NIF Standard Ontology, 4) a backend web services and data management layer that ties these elements together.We have created an existance proof prototype of this platform that allows data from brain regions to molecules to be represented in a common space. It allows researchers to upload their data into this framework in forms such as 3D meshes of subcellular scenes and of brain region territories, large 2D image data sets from both EM and light level microscopy, NeuroML / Neurolucida neuronal reconstructions, and PDB protein structures. Researchers can annotate their data with tags from the ontology system, and manipulating their data within the prototype causes knowledge links to be made between the data in an exportable formal structure (OWL). We feel that this platform can scale up as the beginning of a shared knowledge environment for neuroanatomy that can handle the multiple scales and modalities of data within this complex domain. Conference: Neuroinformatics 2009, Pilsen, Czechia, 6 Sep - 8 Sep, 2009. Presentation Type: Poster Presentation Topic: Digital atlasing Citation: Larson S, Aprea C, Maynard S, Imam F, Ellisman MH and Martone M (2019). A multi-scale spatial and semantic knowledge framework for neuroanatomy. Front. Neuroinform. Conference Abstract: Neuroinformatics 2009. doi: 10.3389/conf.neuro.11.2009.08.015 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 21 May 2009; Published Online: 09 May 2019. * Correspondence: Stephen Larson, UCSD, San Diego, United States, stephen.larson@gmail.com Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Stephen Larson Christopher Aprea Sarah Maynard Fahim Imam Mark H Ellisman Maryann Martone Google Stephen Larson Christopher Aprea Sarah Maynard Fahim Imam Mark H Ellisman Maryann Martone Google Scholar Stephen Larson Christopher Aprea Sarah Maynard Fahim Imam Mark H Ellisman Maryann Martone PubMed Stephen Larson Christopher Aprea Sarah Maynard Fahim Imam Mark H Ellisman Maryann Martone Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

The branches extending from the cell body of neurons, the dendrites, receive more than 90% of the synaptic contacts made into that neuron. In many neurons of the mammalian brain, excitatory synapses involve specialized structures called dendritic spines that protrude from the dendrites and contain the molecules and organelles involved in the postsynaptic processing of the synaptic information. Neuron morphology, as captured in part by the structure of these spines, is illustrative of neuronal function and can be instrumental in better understanding the dysfunction seen in neurodegenerative conditions such as Alzheimer's and Parkinson's disease. Hence researchers have shown great interest in quantitatively studying dendritic spine morphology and density both statically and as a function of time. Such studies are typically carried out through the analysis of data collected from a range of microscopy modalities including confocal laser scanning microscopy (CLSM) and two-photon laser scanning microscopy (2PLSM). Due to the size and complexity of these data sets, manually analyzing the morphological changes of dendritic spines is very time consuming. In the thesis, we describe robust, automated approaches for dendritic spine detection and measurement that are especially suitable to the batch processing of large neuronal image data sets. Our work is roughly divided into three related components. First, we focus on an image processing pipeline we have developed for the neuroinformatics processing system released from our lab called Neuron Image Quantitator (NeuronIQ), an integrated system for automatic dendrite spine detection, quantification, and analysis. Second, to further improve detection results and solve a collection of related ""hard problems"" (such as disconnected spine segmentation) faced by existing automatic or semi-automatic methods, a post-processing segmentation algorithm based on a Maximum a Posteriori - orientated Markov random field (MAP-OMRF) is discussed in detail. Finally, we will present an efficient particle filter-based algorithm that is capable of tracking morphological changes in the spines over time. Possible future topics will be discussed at the end of the thesis.

Problems with crop production and plants are common in India, affecting rural farmers, the agriculture industry, and the economy as a whole. Leaf plays a significant role in Plants and Crops because, depending on its condition, it provides information in advance about the quantity and quality of agricultural yield. This study has proposed a Deep Neural Network (DNN) algorithm based disease classification, pesticide recommendation and pre-processing of leaf images obtained from a plant village dataset. After expanding the plant leaf image’s data set through geometric manipulation, the extended dataset is separated into a train dataset and a test dataset. After that, the large ImageNet image data set is used to train the ResNet-50 models. In this study, the pre-learned data model constraints are conveyed to the lesser model image dataset of plant infections to train the leaf diseases, pest recognition, and pesticides recommendation model founded on the concept of transfer learning. On a number of diseased leaves, this method is tested. The results of experiments demonstrate that using the transfer learning-based CNN method will improve the higher accuracy and performance efficiency for plant sprig image detection. It is capable of rapidly and precisely diagnosing crop diseases, reducing pesticide and fertilizer use, and increasing crop yield and quality.

The abundance of medical resources has encouraged the development of systems that allow for efficient searches of information in large medical image data sets. State‐of‐the‐art image retrieval models are classified into three categories: content‐based (visual) models, textual models, and combined models. Content‐based models use visual features to answer image queries, textual image retrieval models use word matching to answer textual queries, and combined image retrieval models, use both textual and visual features to answer queries. Nevertheless, most of previous works in this field have used the same image retrieval model independently of the query type. In this article, we define a list of generic and specific medical query features and exploit them in an association rule mining technique to discover correlations between query features and image retrieval models. Based on these rules, we propose to use an associative classifier (NaiveClass) to find the best suitable retrieval model given a new textual query. We also propose a second associative classifier (SmartClass) to select the most appropriate default class for the query. Experiments are performed on Medical ImageCLEF queries from 2008 to 2012 to evaluate the impact of the proposed query features on the classification performance. The results show that combining our proposed specific and generic query features is effective in query classification.

Increasingly, MRI is used outside the diagnostic setting to study brain physiology, or the biological mechanisms of brain maturation and ageing. In an era with an increasingly ageing population, the study of age-related brain changes is important to understand risk factors and early signs of neurodegenerative disease, such as Alzheimer’s disease. To detect subtle changes in brain structure, large image banks are typically needed, for example to examine the impact of rare or weak genetic traits. Those studies often require substantial efforts to collect and then are used to address a limited set of questions framed by the original investigators, even though they would allow additional questions to be considered by others if disclosed publically and well documented. One could argue that publicly funded agencies of such studies would get better value for the money if the data were used by more than one research group, a trend that is now being translated into new policies regarding ownership of research material in various countries. The ADNI project (www.adni-info.org) was a big step forward, where the ultimate goal was to create a publically accessible data set, which to date has already been used in hundreds of published studies. One advantage of such a prospective initiative is that the homogeneity of images and metadata is better than for retrospectively disclosed repositories, with often only partly overlapping data dictionaries. The review by Dickie et al. [1] in the current issue systematically reviews open access of publically accessible databanks of the ageing brain. Among a large number of potential databases, only a few were found to have public access; among those, nine fulfilled the inclusion criteria, and far less than 1,000 complete MR imaging data sets were available in subjects over the age of 60. Furthermore, access is often not easily obtained (even for ADNI user have to submit a proposal and obtain approval); metadata (e.g., clinical) are often presented incompletely (and vary between studies), intermediate results (e.g., segmentations or statistical maps) are typically lacking and basic MRI descriptors (e.g., the amount of white matter lesions) not available. Therefore, despite the excellent original intentions, the yield of public databases so far has been modest. What would be needed to enhance to unique possibilities in this field? First of all, the need for a public call to disclose more data publically should be articulated. The paper by Dickie et al. [1] helps to revive this public appeal. Secondly, there will be a need for a consortium that proactively approaches potential data donors. For example, the MRI data from existing studies such as the Cardiovascular Health Study (www.chs-nhlbi.org) or the Rotterdam Scan Study are not available, although they were funded by government money. Thirdly, for major funding requests, public funding agencies should strongly prefer applications that have both the plan and commitment to make the data publicly accessible within 2 or 3 years of acquisition (making the sponsors rather than the investigators the owners of the data). Fourthly, an independent forum may be needed to share the data publically in a regulated environment, ensuring subject confidentiality and data integrity. For example, the FP7 project entitled neuGRID (www.neugrid.eu) provides such a portal, and at the same time allows users to take advantage of the virtual laboratory and computing facilities typically needed to handle such large image data sets [2]; similar endeavours exist in North America (www.outgrid.eu). Fifthly, to ensure academic independence, users of the data should only have to report published results to the original investigators. Sixthly, there should be standards of reporting data and storing of intermediate results that can be used by subsequent users, and avoid replication of for example segmentation and registration efforts. Finally, metadata such as clinical, laboratory, genetic and other information should be unified. Standard terminology in various domains is being developed by the NINDS (http://www.commondataelements.ninds.nih.gov). Future expansion beyond the study of normal ageing could include disease states, such as Alzheimer's, or into other imaging techniques, such as nuclear medicine. Beyond the collection of anatomic brain MRI scans, there is increasing interest in more advanced MRI data, such as resting-state functional MRI and diffusion tensor imaging (DTI) to unravel the “human connectome” (http://www.humanconnectomeproject.org and humanconnectome.org). An example of PET/SPECT data sharing in Alzheimer’s disease already exists (https://www.eu-decide.eu). Data sharing could also be expanded to other disease areas where international collaborations already exist, such as for multiple sclerosis (www.magnims.eu) and Huntington's disease (http://hdresearch.ucl.ac.uk/our-results/track-hd), where currently such data are usually not publicly accessible. The scale of economy that is needed in brain imaging research calls not only for larger studies in ageing and dementia, but also for better use of data by third parties by making such data publicly accessible. Ideally, this is considered at the study outset, allowing interoperability of images and metadata with existing repositories.

This paper presents an image understanding approach to mine remotely sensed image data from different source dates for environmental change detection. It is focused on the immediate needs for knowledge discovery from large sets of image data for environmental monitoring. In contrast to the traditional approaches for change detection, we introduce a wavelet-based hierarchical scheme which integrates fuzzy set theory and image understanding techniques for knowledge discovery of the remote image data. The proposed approach includes algorithms for hierarchical change detection, region representations and classification. The effectiveness of the proposed algorithms is demonstrated throughout the completion of three tasks, namely hierarchial detection of change by fuzzy post classification comparisons, localization of change by B-spline based region representation, and categorization of change by hierarchial texture classification.

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] Camera traps are a popular tool to sample animal populations because they are noninvasive, detect a variety of species, and can record many thousands of animal detections per deployment. Cameras are typically set to take bursts of multiple images for each detection, and are deployed in arrays of dozens or hundreds of sites, often resulting in millions of images per study. The task of converting images to animal detection records from such large image collections is daunting, and made worse by situations that generate copious empty pictures from false triggers (e.g. camera malfunction or moving vegetation) or pictures of humans. We offer the first widely available computer vision tool for processing camera trap images. Our results show that the tool is accurate and results in substantial time savings for processing large image datasets, thus improving our ability to monitor wildlife across large scales with camera traps. In this dissertation, we have developed new image/video processing and computer vision algorithms for efficient and accurate object detection and sequence-level classiffication from natural scene camera-trap images. This work addresses the following five major tasks: (1) Human-animal detection. We develop a fast and accurate scheme for human-animal detection from highly cluttered camera-trap images using joint background modeling and deep learning classification. Specifically, first, We develop an effective background modeling and subtraction scheme to generate region proposals for the foreground objects. We then develop a cross-frame image patch verification to reduce the number of foreground object proposals. Finally, We perform complexity-accuracy analysis of deep convolutional neural networks (DCNN) to develop a fast deep learning classification scheme to classify these region proposals into three categories: human, animals, and background patches. The optimized DCNN is able to maintain high level of accuracy while reducing the computational complexity by 14 times. Our experimental results demonstrate that the proposed method outperforms existing methods on the camera-trap dataset. (2) Object segmentation from natural scene. We first design and train a fast DCNN for animal-human-background object classification, which is used to analyze the input image to generate multi-layer feature maps, representing the responses of different image regions to the animal-human-background classifier. From these feature maps, we construct the so-called deep objectness graph for accurate animal-human object segmentation with graph cut. The segmented object regions from each image in the sequence are then verfied and fused in the temporal domain using background modeling. Our experimental results demonstrate that our proposed method outperforms existing state-of-the-art methods on the camera-trap dataset with highly cluttered natural scenes. (3) DCNN domain background modeling. We replaced the background model with a new more efficient deep learning based model. The input frames are segmented into regions through the deep objectness graph then the region boundaries of the input frames are multiplied by each other to obtain the regions of movement patches. We construct the background representation using the temporal information of the co-located patches. We propose to fuse the subtraction and foreground/background pixel classiffcation of two representation : a) chromaticity and b) deep pixel information. (4) Sequence-level object classiffcation. We proposed a new method for sequence-level video recognition with application to animal species recognition from camera trap images. First, using background modeling and cross-frame patch verification, we developed a scheme to generate candidate object regions or object proposals in the spatiotemporal domain. Second, we develop a dynamic programming optimization approach to identify the best temporal subset of object proposals. Third, we aggregate and fuse the features of these selected object proposals for efficient sequence-level animal species classification.

InfraGAN: A GAN architecture to transfer visible images to infrared domain

While recognizing an image taken from an arbitrary angle in a real environment, it may be challenging to obtain a good recognition result on occasion. That is, the recognition accuracy is affected by changing the viewpoint (object viewing direction). This phenomenon could come from the fact that people photograph target objects on well explainable aspect. It has been known empirically that the recognition rate varies depending on the viewpoint so far, but this topic remains underdeveloped. Therefore, in this research, we aim to clarify and visualize the existence of viewpoint biases in existing large object recognition image datasets. We also quantify the viewpoint biases by a definition of viewpoint bias index.

Gaussian-based filtering techniques have been popularly applied to edge detection. However, how to effectively tune parameters of Gaussian filters and how to effectively combine different Gaussian filters are still open issues. In this study, a new genetic programming (GP) approach is proposed to automatically tune parameters of Gaussian filters and automatically combine different types of Gaussian filters to extract edge features. In general, it is time-consuming for GP to evolve edge detectors using a large training image dataset. To efficiently evolve edge detectors from a large training image dataset, we propose sampling techniques (randomly selecting training images) to evolve Gaussian-based edge detectors. A sampling technique only using part of a set of images obtains similar performance to the training data using all of these images. The evolved edge detectors from the proposed sampling technique perform better than the Gaussian gradient and rotation invariant surround suppression. Based on the analysis of GP evolving edge detectors, it is suggested that combining Gaussian filters should be based on different types of Gaussian filters, and the Gaussian gradient should be considered as a major filter in these combinations.

Applications analytics are growing at a rapid pace. Techniques for image processing, extraction, compression, and recognition are established and are in use. However, the presence of large data sets of images on the web is increasing exponentially. In this paper, we propose a mechanism that retrieves images from a large data set and arrive at a cluster for analysing behavioural trends. The proposed system has been implemented using distributed cloud technology and a crawler script. We arrived at interesting results and have done a comparative analysis of various social, geographical and technological trends across four well-known locations on the globe.