Machine learning approaches for identifying prey handling activity in otariid pinnipeds

Systems developed in wearable devices with sensors onboard are widely used to collect data of humans and animals activities with the perspective of an on-board automatic classification of data. An interesting application of these systems is to support animals' behaviour monitoring gathered by sensors' data analysis. This is a challenging area and in particular with fixed memories capabilities because the devices should be able to operate autonomously for long periods before being retrieved by human operators, and being able to classify activities onboard can significantly improve their autonomy. In this paper, we focus on the identification of prey handling activity in seals (when the animal start attaching and biting the prey), which is one of the main movement that identifies a successful foraging activity. Data taken into consideration are streams of 3D accelerometers and depth sensors values collected by devices attached directly on seals. To analyse these data, we propose an automatic model based on Machine Learning (ML) algorithms. In particular, we compare the performance (in terms of accuracy and F1score) of three ML algorithms: Input Delay Neural Networks, Support Vector Machines, and Echo State Networks. We attend to the final aim of developing an automatic classifier on-board. For this purpose, in this paper, the comparison is performed concerning the performance obtained by each ML approach developed and its memory footprint. In the end, we highlight the advantage of using an ML algorithm, in terms of feasibility in wild animals' monitoring.

Estimating Fire Weather Indices via Semantic Reasoning over Wireless Sensor Network Data Streams

Wildfires are frequent, devastating events in Australia that regularly cause significant loss of life and widespread property damage. Fire weather indices are a widely-adopted method for measuring fire danger and they play a significant role in issuing bushfire warnings and in anticipating demand for bushfire management resources. Existing systems that calculate fire weather indices are limited due to low spatial and temporal resolution. Localized wireless sensor networks, on the other hand, gather continuous sensor data measuring variables such as air temperature, relative humidity, rainfall and wind speed at high resolutions. However, using wireless sensor networks to estimate fire weather indices is a challenge due to data quality issues, lack of standard data formats and lack of agreement on thresholds and methods for calculating fire weather indices. Within the scope of this paper, we propose a standardized approach to calculating Fire Weather Indices (a.k.a. fire danger ratings) and overcome a number of the challenges by applying Semantic Web Technologies to the processing of data streams from a wireless sensor network deployed in the Springbrook region of South East Queensland. This paper describes the underlying ontologies, the semantic reasoning and the Semantic Fire Weather Index (SFWI) system that we have developed to enable domain experts to specify and adapt rules for calculating Fire Weather Indices. We also describe the Web-based mapping interface that we have developed, that enables users to improve their understanding of how fire weather indices vary over time within a particular region.Finally, we discuss our evaluation results that indicate that the proposed system outperforms state-of-the-art techniques in terms of accuracy, precision and query performance.