Method for Cost Optimized Volumetric Object Monitoring Systems

Through close collaboration with industrial partners researchers will provide the industry with a tool for predicting future activity within a volume. Although this project focuses on monitoring wind farms and flying birds’ activities, the result are applicable in other fields. 

Abstract

The ability to effectively identify flying objects in a given volume and characterize their activities can lead to improved prediction of future activity. Such prediction is applicable, for example, in improving collision avoidance systems for wind farms. Employing a visual monitoring system consisting of many camera nodes that form visual sensor network (VSN) will provide more reliable datasets compared to a human approach. Effective characterization can be achieved through robust models of the activities based on adequate datasets provided by the VSN. In this project we will investigate the requirements for VSN node architecture and deployment topology for remote outdoor monitoring. Our aim is to find a cost optimized design for the VSN nodes and topology constrained limited resources. These resources include energy, communication and data storage. In this project, we will investigate multi-camera node architecture (MCNA) by exploring trade-offs among there sources and node cost. We will also investigate how MCNA can lead to a cost optimized node deployment topology. Through close collaboration with industry partners we will be able to fulfil the project objectives and provide the industry with a tool for predicting future activity within a volume. Although this proposal focuses on monitoring wind farms and flying birds’ activities, the result are applicable in other fields.

Background

Research in surveillance and computer vision usually focuses on limited area and field of view under controlled conditions (such as illumination, placements and occlusions) and resources (energy, communication, computation and storage). Large volume outdoor monitoring (LVOM) has limited resources due to remote location requires high deployment cost. Addressing LVOM challenges is still an open research question. There is a need to investigate node design and node deployment topology in order to optimize the costs. These challenges can be seen from three perspectives namely: monitoring-, node- and system requirements. In LVOM, the expected results include characterizing of flying objects and prediction of their future activity. This places high demands on the design, deployment, maintenance and cost of the monitoring system. This research project has application focus on monitoring of birds in wind farms with varying objectives

In large area outdoor monitoring, the expected results include characterizing of flying objects and prediction of their future activity. This places high demands on the design, deployment, maintenance and cost of the monitoring system. This research project has application focus on monitoring of birds in wind farms with varying objectives. In one application scenario there might be a need to understand how different bird species interact with existing wind farms. In another application scenario the aim might be to ensure better protection of eagles through early prediction of potential impact of a proposed wind farm on the local eagle population. Designing a VSN that will fulfil the monitoring objectives in these scenarios demands investigations at node and system level. It nee to fulfil pixel resolution and colour requirements at node level and group-, temporal- and spatial-behaviour analysis at system level. The monitoring system could potentially be re-used at different locations for different objectives. For example, short term (6-18 months) monitoring during site surveys (e.g. scenario 2) and long term (1-5 years) monitoring to measure the impact of a wind farm on bird populations (e.g. scenario 1).

To be cost optimized, the LVOM system must consider the cost of each node, the number of nodes, and deployment cost as well as maintenance and operational (energy, storage and communication) costs. During node design, we will investigate essential image processing algorithms (e.g. segmentations, object detection, classification and feature extraction) to optimize for performance and resources usage.

Objective

The objectives is to investigate the design of configurable platform based multi-camera node architecture (MCNA), iteratively design and refine the MCNA-based dome topology for large volume outdoor monitoring and compare its cost effectiveness with that of the FSCNA-based grid topology. To achieve the research objectives the following challenges should be addressed:

• How to achieve a cost optimized design of VSN that fulfils the LVOM objectives.
• How to design configurable platform based MCNA for resource constrained LVOM.

Project Partners

This project aims to develop a method to design a cost optimized MCNA-based VSN to characterize the interaction of flying objects in a given monitored volume. To achieve this researchers work closely with the participating partners. This research project is proposed based on the analysis of the interests, contributions and expectations of the participating partners with respect to large volume outdoor monitoring and how the research can be employed to address the challenges. The following is a short description of interests of the project partners:

Vattenfall – is interested in large volume outdoor monitoring consisting of existing wind farms with requirements on determining bird population, species classification and analysis of collision events within a given time interval.

InSitu – would like to find out if it is possible perform large volume outdoor monitoring, what the challenges involved are and how they can be addressed in order to achieve a cost optimal system.

Combitech – is interested in finding accurate and cost effective method for detection of ice throw in large area and use the information in de-icing techniques and safety in wind parks. In addition, they are interested detection and behavior of drones in a volume.