Grants and Contributions

This project enhances Complex System Inc. (CSI) trustworthy artificial intelligence (AI) for IoT networks to ensure service continuity, secure municipal operations, and resist cyberattacks.

This project aims to develop a trustworthy AI software based on a data-driven predictive-3D technology from a single 2D satellite image. It will provide new high-resolution 3D maps and change-detection to aid in assessing land, agriculture, and provide resilience to topography changes.

Funding will support research to National Defence's Fast Forward: Forecasting global emerging threats Challenge. The Department of National Defence (DND) and the Canadian Armed Forces (CAF) are seeking innovative forecasting models that leverage the power of publicly available information, statistical methods, and machine learning models to assist DND with forecasting flashpoints and emerging threats around the world.

As part of the international consortium project “SIREN: Sampols 2.0: a disruptive solution for noise pollution monitoring”, this project will focus on the development of a system to monitor noise in large areas in real-time. While the current Sampols solution has been deployed in a few European cities and industrial sites, several new features for intelligent noise analysis based on machine learning will be developed which can allow decision makers to take corrective actions on noise pollution and opens new applications at a large scale with potential distribution to the Canadian market. Tech

This International Co-Innovation Action Program (ICAP) project will prepare the firm to enter into future international co-innovation work. Complex System Inc (CSI) is a multi-sensor analytic software and AI technology company.

The proliferation of small satellites in the New Space Era has spurred the development of computational systems deployed on agile CubeSats for onboard detection of cloud and floods. There is a need for scalable space IoT frameworks for satellite edge computing (SEC): In addition to growing real-time data requirements for earth observation applications, capability to perform or offload compute on devices is critical for 1) remote areas, mountains, and sea, where ground resource and connectivity is limited 2) Delay-sensitive applications with multi-domain data for situation awareness. CSI proposes to develop an edge-AI software framework for SEC. It will increase capability of AI deployed on satellites as edge nodes while adapting to new tasks. The framework will assess models and performance to minimize bandwidth. The proposed work is demonstrated on use case in satellite processing, updating models for new scenarios onboard, and has potential for multiple satellites.

Earth observation with large spatial coverage, high temporal revisits make an attractive option for continuous monitoring of Earth. The objective is to develop a software framework that leverages AI and processing tools to automatically search and discover events or objects of interest. To focus on overcoming the data limitation problem, with an innovative approach that combines few-shot learning paradigm with synthetic data generation and a use case for rare events, objects, or anomalies. A generative model will also be developed to learn the transformation between different sensor types to generate synthetic satellite images. To address the high dimensionality of satellite image data, we propose a patch/region-based approach to generation.

Participation in the Canadian ocean technology partnering mission to the EUREKA Global Innovation Summit (EGIS) in Portugal – June 20-24, 2022. Delegation led by the National Research Council of Canada Industrial Research Assistance Program (NRC IRAP) to develop international partnerships and potential project ideas under the EUREKA, Eurostars, and Cluster Programs.

Damage detection of roads around the city is a time-consuming manual visual inspection process prone to human errors. CSI solution comprises of an imaging system to enhance the detection and classification of the damages on roads.

In this project, the firm will extend capabilities of detection, tracking and identification from maritime environment to the land area. Land application poses challenge on detection of small and potentially faster moving objects. The firm proposes a motion-based approach for detection of land vehicles from satellite videos and a context-based detection method for land where every region of the land is semantically segmented to identify the land use patterns. The firm proposes to create a database of well annotated data points that can be used to correct the detection and classification performance based on uncertainty measure.