Mike Botts – January SensorML and Processing September 2009 Mike Botts Botts Innovative Research, Inc.
Mike Botts – January What is SensorML? XML encoding for describing sensor processes –Including sensor tasking, measurement, and post-processing of observations –Detectors, actuators, sensors, etc. are modeled as processes Open Standard – –Approved by Open Geospatial Consortium in 2007 –Supported by Open Source software (COTS development starting) Not just a metadata language – enables on-demand execution of algorithms Describes –Sensor Systems –Processing algorithms and workflows
Mike Botts – January Why is SensorML Important? Importance: –Discovery of sensors and processes / plug-n-play sensors – SensorML is the means by which sensors and processes make themselves and their capabilities known; describes inputs, outputs and taskable parameters –Observation lineage – SensorML provides history of measurement and processing of observations; supports quality knowledge of observations –On-demand processing – SensorML supports on-demand derivation of higher-level information (e.g. geolocation or products) without a priori knowledge of the sensor system –Intelligent, autonomous sensor network – SensorML enables the development of taskable, adaptable sensor networks, and enables higher-level problem solving anticipated from the Semantic Web
Mike Botts – January SensorML Processes Physical ProcessesNon-Physical Processes Atomic Processes Composite Processes Processes that are considered Indivisible either by design or necessity Processes that are composed of other processes connected in some logical manner Processes where physical location or physical interface of the process is not important (e.g. a fast-Fourier process) Processes where physical location or physical interface of the process is important (e.g. a sensor system)
Mike Botts – January Example Atomic Processes Transducers (detectors, actuators, samplers, etc.)detectors Spatial transforms (static and dynamic) –Vector, matrix, quaternion operators –“Sensor models” scanners, frame cameras, SARframe cameras polynomial models (e.g. RPC, RSM)RPC tie point model –Orbital modelsOrbital models –Geospatial transformations (Map projection, datum, coordinate system)coordinate system Digital Signal Processing / image processing modules Decimators, interpolators, synchronizers, etc. Data readers, writers, and access services Derivable Information (e.g. wind chill)wind chill) Human analysts To browse ProcessModel
Mike Botts – January Example Composite Processes Sensor Systems, PlatformsSensor Systems Observation lineage –from tasking to measurement to processing to analysis Executable on-demand process chains: –geolocation and orthorectification –algorithms for higher-level products e.g. fire recognition, flood water classification, etc. –Image processing, digital signal processing Uploadable command instructions or executable processes
Mike Botts – January SensorML Process Chains
Mike Botts – January NASA Projects: SensorML-Enabled On-demand Processing (e.g. georeferencing and product algorithms) AMSR-E SSM/I CloudsatLIS TMI TMI & MODIS footprints MAS Geolocation of satellite and airborne sensors using SensorML
Mike Botts – January SensorML – Sensor Systems Mike Botts, Alexandre Robin, Tony Cook Sensor 1 Scanner System - Aircraft Sensor 2 IMU Sensor 3 GPS IR radiation Attitude Location Digital Numbers Pitch, Roll, Yaw Tuples Lat, Lon, Alt Tuples
Mike Botts – January AIRDAS UAV Geolocation Process Chain Demo AIRDAS data stream (consisting of navigation data and 4-band thermal-IR scan-line data) AIRDAS data stream geolocated using SensorML-defined process chain (software has no a priori knowledge of sensor system)
Mike Botts – January Supports description of Lineage for an Observation Observation SensorML Within an Observation, SensorML can describe how that Observation came to be using the “procedure” property
Mike Botts – January On-demand processing of sensor data Observation SensorML processes can be executed on-demand to generate Observations from low-level sensor data (without a priori knowledge of sensor system) SensorML
Mike Botts – January On-demand processing of higher-level products Observation SensorML SensorML processes can be executed on- demand to generate higher-level Observations from low-level Observations (e.g. discoverable georeferencing algorithms or classification algorithms) Observation
Mike Botts – January Clients can discover, download, and execute SensorML process chains For example, Space Time Toolkit is designed around a SensorML front-end and a Styler back- end that renders graphics to the screen SensorML OpenGL SensorML-enabled Client (e.g. STT) Stylers SLD SOS
Mike Botts – January Incorporation of SensorML into Space Time Toolkit Space Time Toolkit being retooled to be SensorML process chain executor + stylers
Mike Botts – January Space Time Toolkit Sample Applications -2-
Mike Botts – January SensorML can support generation of Observations within a Sensor Observation Service (SOS) Observation SensorML For example, SensorML has been used to support on-demand generation of nadir tracks and footprints for satellite and airborne sensors within SOS web services SOS Web Service request
Mike Botts – January Conclusions SensorML is not just for sensors SensorML provides a robust means of describing a process (both physical and non-physical) – including methodology SensorML process chains provide an implementation-agnostic way to describe workflows or algorithms SensorML process chains can include and mix processes that are implemented locally and those implemented on web services SensorML for processing has been tested and demonstrated in operational environments Propose that SensorML processes be at least one of the means for a WPS to describe the process