Framework for detection of anomalies in mass moving objects by non-technical users utilizing contextual & spatio-temporal data

Increasing utility of wireless sensors and their decreasing cost with technological innovation has increased their presence in many mobile & wearable devices and many devices are able to sense and relate a variety of sensor data including GPS location to be used for decision making. However, much potential of this data remains unexploited due to inherent complexity of manipulating such information, which bars the majority of users benefiting from this. One potential application is the detection of anomalies by users from a non-technical background through definition of rules. As such, a methodology/framework, which could enable users to define rules/query information from sensors fused with other contextual data in a user-friendly manner would be highly useful in deriving value from this data. This research attempts to formulate a low-cost framework, which enables users without a strong technical background to make potential use of this. The research is performed using a case study; identification of anomalies in the behaviours and violations of rules by marine vessels in Sri Lankan waters. However, the framework is generalized to enable application to other contexts also. The framework also attempts to identify a possible method and a process to enable users enter geo spatial rules in an accurate as well as user friendly manner while allowing to enter rules of different levels of complexities accurately. The research attempts to solve this problem through the use of a map interface which enables users enter geo spatial data either plotting on a screen or as pairs of coordinates. Use of auto query forms with dynamically changing queries was proposed to enter conditions of rules and the interrelationships. The research was carried out using one prototype and two systems in three stages, using the input of one step to next. Both methods of spatial rule entry were deemed accepted by users considering the ease of use in direct plotting as well as accuracy in formal coordinate entry in the two methods. While auto query form was found as a suitable method with reasonable accuracy with accuracy increasing with the educational background of users and decreasing on several occasions with increased complexity (multiple conditions or ambiguity of underlying concept) of rules, the display of the rule query in user readable form at the bottom of the interface as well as enabling the user to dynamically view query results to refine the query showed increased accuracy.