Image-Based Multi-Sensor Data Representation and Fusion Via 2D Non-Linear Convolution

Sensor data fusion is the process of combining data collected from multi sensors ofhomogeneous or heterogeneous modalities to perform inferences that may not be possible usinga single sensor. This process encompasses several stages to arrive at a sound reliable decisionmaking end result. These stages include: senor-signal preprocessing, sub-object refinement,object refinement, situation refinement, threat refinement and process refinement. Every stagedraws from different domains to achieve its requirements and goals. Popular methods for sensordata fusion include: ad-hock and heuristic-based, classical hypothesis-based, Bayesianinference, fuzzy inference, neural networks, etc. in this work, we introduce a new data fusionmodel that contributes to the area of multi-senor/source data fusion. The new fusion model relieson image processing theory to map stimuli from sensors onto an energy map and uses non-linearconvolution to combine the energy responses on the map onto a single fused response map. Thisresponse map is then fed into a process of transformations to extract an inference that estimatesthe output state response as a normalized amplitude level. This new data fusion model is helpfulto identify sever events in the monitored environment. An efficiency comparison with similarfuzzy-logic fusion model revealed that our proposed model is superior in time complexity asvalidated theoretically and experimentally.