Unsupervised Satellite Image Time Series Clustering Using Object-Based Approaches and 3D Convolutional Autoencoder

Nowadays, satellite image time series (SITS) analysis has become an indispensable part of many research projects as the quantity of freely available remote sensed data increases every day. However, with the growing image resolution, pixel-level SITS analysis approaches have been replaced by more efficient ones leveraging object-based data representations. Unfortunately, the segmentation of a full time series may be a complicated task as some objects undergo important variations from one image to another and can also appear and disappear. In this paper, we propose an algorithm that performs both segmentation and clustering of SITS. It is achieved by using a compressed SITS representation obtained with a multi-view 3D convolutional autoencoder. First, a unique segmentation map is computed for the whole SITS. Then, the extracted spatio-temporal objects are clustered using their encoded descriptors. The proposed approach was evaluated on two real-life datasets and outperformed the state-of-the-art methods.

Integrated field-agent based modelling using the LUE scientific data base

<p>The heterogeneous nature of environmental systems poses a challenge to researchers constructing environmental models. Many simulation models of integrated systems need to incorporate phenomena that are represented as spatially and temporally continuous fields as well as phenomena that are modelled as spatially and temporally bounded agents. Examples include moving animals (agents) interacting with vegetation (fields) or static water reservoirs (agents) as components of hydrological catchments (fields). However, phenomena bounded in space and time have particular properties mainly because they require representation of multiple (sometimes mobile) objects that each exist in a small subdomain of the space-time domain of interest. Moreover, these subdomains of objects may overlap in space and time such as interleaving branches due to tree crown growth. Efficient storage and access of different types of phenomena requires an approach that integrates representation of fields and objects in a single data model.</p><p>We develop the open-source LUE data model that explicitly stores and separates domain information, i.e. where phenomena exist in the space-time domain, and property information, i.e. what attribute value the phenomenon has at a particular space-time location, for a particular object. Notable functionalities are support for multiple spatio-temporal objects, time domains, objects linked to multiple space and time domains, and relations between objects. The design of LUE is based on the conceptual data model of de Bakker (2017) and implemented as a physical data model using HDF5 and C++ (de Jong, 2019). Our LUE data model is part of a new modelling language implemented in Python, allowing for operations accepting both fields and agents as arguments, and therefore resembling and extending the map algebra approach to field-agent modelling.</p><p>We present the conceptual and physical data models and illustrate the usage by implementing a spatial agent-based model simulating changes in human nutrition. We thereby consider the interaction between personal demand and supply of healthy food of nearby stores as well as the influence of agent's social network.</p><p><br>References:</p><p>de Bakker, M. P., de Jong, K., Schmitz, O., & Karssenberg, D. (2017). Design and demonstration of a data model to integrate agent-based and field-based modelling. Environmental Modelling & Software, 89, 172–189. https://doi.org/10.1016/j.envsoft.2016.11.016</p><p>de Jong, K., & Karssenberg, D. (2019). A physical data model for spatio-temporal objects. Environmental Modelling & Software. https://doi.org/10.1016/j.envsoft.2019.104553</p><p>LUE source code repository: https://github.com/pcraster/lue/</p>

RESEARCH ON CONSTRUCTION METHOD OF SPATIO-TEMPORAL DATA MODEL BASED ON BEHAVIOR-EVENT

Starting from the object-oriented idea, this paper analyses the existing event-based models and the logical relationship between behavioral cognition and events, and discusses the continuity of behavioral cognition on the time axis from the perspective of temporal and spatial cognition. A geospatial data model based on behavioral-event is proposed. The physical structure and logical structure of the model are mainly designed, and the four-dimensional model of “time, space, attribute and event” is constructed on the axis. The organic combination of the four models can well describe the internal mechanism and rules of geographical objects. The expression of data model based on behavior-event not only elaborates the basic information of geospatial objects, but also records the changes of related events caused by the changes of geographic Entities' behavior, and expresses the relationship between spatial and temporal objects before and after the changes of behavior cognition. This paper also designs an effective method to organize spatio-temporal data, so as to realize the effective management and analysis of spatio-temporal data and meet the requirements of storage, processing and mining of large spatio-temporal data.

TIME AND ITS INDETERMINACY IN ROMAN INGARDEN’S CONCEPT OF THE LITERARY WORK OF ART

The time of the literary work of art is a schematized aspect of the represented objectivities of the literary work. Roman Ingarden’s analysis of the time of the literary work extends upon Husserl’s phenomenology of time consciousness but nevertheless remains consistent with it, insofar as within a literary work as well as actuality, an objective time or literary objective time is constituted from an experience of temporal objects (or in the case of the work of art, represented objectivities). The time of the literary work of art is experienced through its represented objectivities and depends on an already existent concept of objective time. (We conceive of the time of the literary work of art as a variant of actual time.) The time of the literary work, however, gains an indeterminacy, as the author of the literary work is able to manipulate our experience of time within the literary work by predicting how the reader will concretize its events, characters, and objects with respect to time. I use Alain Robbe-Grillet’s representation of the centipede in La Jalousie as an example of the possible manipulations of time, explicable through Ingarden’s analysis of time in the literary work of art.

Instant Replay

In a first part, the contribution goes through different competing and/or complementary concepts of ontography as they appear in phenomenology, object-oriented ontology, science and technology studies, and semiotics. From this comparative examination, the text develops a notion of ontography in contrast to that of ontology. It highlights the ontography of temporal objects and adds a specific media-philosophical approach to it by concentrating on the operations and tools of ontographical writing, registering or drawing being in time. In a second part, ontography is analysed as a techno-aesthetic operation on television, namely as the core of the slow motion replay. Four spectacular examples of instant replay are taken from the history of television that make the writing of an extended and malleable presence through television visible. From that, the contribution develops a hypothesis about the ontographical functioning of the medium of television and of its approach to the world.

Spatio-Temporal Range Search using K-DSTH Indexing Structure

A large amount of data which includes spatial and temporal information related to different fields like geography, satellite, medical or multimedia is generated and collected at an extraordinary scale. Such data is produced by satellites, mobile devices, emerging applications like social networking sites, photo sharing sites and many more. As the whole world is aware of the importance of such spatio-temporal data, a great amount of research work is evolving around efficient storage structures and algorithms to handle a variety of spatio-temporal queries. In this paper, the authors are introducing a novel spatio-temporal indexing structure k-dStH which is an extension of k-dSTHash indexing structure. It uses master hash table, local hash table and B-tree additionally which are based on timestamp values. The researchers also introduce an algorithm k-dStHSpaTempRangeSrch based on the proposed indexing structure to find spatio-temporal objects in given spatial range at particular temporal value. The performance analysis shows that the algorithm proposed by the authors is far more efficient as compared to brute force technique of searching for the spatio-temporal objects.

Real-time interactive streams and temporal objects language

We introduce a real-time interactive streams language based on themodel of temporal objects in order to (1) study various typing featuresfor timed streams and (2) define a hierarchical calculus for streamprocessus. The present document deals only with formalisation, andintroduces the typing features.