Evaluation of eCognition for Assisted Target Detection and Recognition in SAR Imagery

Defence R&D Canada - Ottawa (DRDC Ottawa) has been assessing custom off-the-shelf (COTS) software products and developing custom tools that can assist image analysts (IAs) to complete their tasks more quickly and efficiently. In some cases, IAs are tasked with detecting targets of interest within large-area surveillance operations, and these detection tasks may be repeated with a regular flow of incoming imagery. SAR imagery is a data source of choice for large-scale detection operations due to its ability to image in all weather conditions without external illumination sources. An activity was undertaken to evaluate ecognition (a COTS product) for target detection in SAR imagery and as a means of facilitating workflow improvements in the course of target detection and surveillance activities. eCognition is segmentation-based classifier that uses fuzzy reasoning techniques, enabling users to represent fuzzy rules in a human reasoning-like form. It offers the promise of being a usable tool that fulfills the objective of providing improvements over the traditional practice of manual inspection for target detection and surveillance. This paper reports on the evaluation process and findings when using ecognition for ship detection and land vehicle detection in SAR imagery.